diff --git a/jsonnetfile.lock.json b/jsonnetfile.lock.json
index 0009cf7e44b8a709ebf304d0d9624ef16bc7b637..b1d29d7623635696cabe6199744254f257e59e1e 100644
--- a/jsonnetfile.lock.json
+++ b/jsonnetfile.lock.json
@@ -8,7 +8,7 @@
"subdir": "jsonnet/kube-prometheus"
}
},
- "version": "bdf84bf1865e66d76c027eb2cea7bf776acd18a4"
+ "version": "90b8632fb37be04ff73542218d980ba54c53295b"
},
{
"name": "ksonnet",
@@ -28,7 +28,7 @@
"subdir": ""
}
},
- "version": "af494738e1709998696ffbce9296063a20c80692"
+ "version": "d8f135ba007b4ec7ac58be9371042d19e1ae4dea"
},
{
"name": "grafonnet",
@@ -78,7 +78,7 @@
"subdir": "Documentation/etcd-mixin"
}
},
- "version": "9ff762857712c9c5c2037e6ddd9a692a488224bf"
+ "version": "948e276ca73d3eb09391829d8ac317dbda8c07a1"
}
]
}
diff --git a/manifests/grafana-dashboardDefinitions.yaml b/manifests/grafana-dashboardDefinitions.yaml
index 1a51fd48ca5ff9f6871012ba1b93d42594f9d45d..dea85646ff02cc139b7971a7eeb8b45911762365 100644
--- a/manifests/grafana-dashboardDefinitions.yaml
+++ b/manifests/grafana-dashboardDefinitions.yaml
@@ -2,26 +2,117 @@ apiVersion: v1
items:
- apiVersion: v1
data:
- k8s-cluster-rsrc-use.json: |-
+ apiserver.json: |-
{
+ "__inputs": [
+
+ ],
+ "__requires": [
+
+ ],
"annotations": {
"list": [
]
},
- "editable": true,
+ "editable": false,
"gnetId": null,
"graphTooltip": 0,
"hideControls": false,
+ "id": null,
"links": [
],
- "refresh": "10s",
+ "refresh": "",
"rows": [
{
"collapse": false,
- "height": "250px",
+ "collapsed": false,
"panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(up{job=\"apiserver\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Up",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ },
{
"aliasColors": {
@@ -30,42 +121,68 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 1,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
- "stack": true,
+ "span": 5,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:cluster_cpu_utilisation:ratio{cluster=\"$cluster\"}",
+ "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "2xx",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "3xx",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "4xx",
+ "refId": "C"
+ },
+ {
+ "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "5xx",
+ "refId": "D"
}
],
"thresholds": [
@@ -73,7 +190,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Utilisation",
+ "title": "RPC Rate",
"tooltip": {
"shared": false,
"sort": 0,
@@ -91,20 +208,20 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "ops",
"label": null,
"logBase": 1,
- "max": 1,
- "min": 0,
+ "max": null,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "ops",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
},
@@ -116,42 +233,47 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 2,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
"legend": {
+ "alignAsTable": "true",
"avg": false,
- "current": false,
+ "current": "true",
"max": false,
"min": false,
- "show": true,
+ "rightSide": "true",
+ "show": "true",
"total": false,
- "values": false
+ "values": "true"
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
- "stack": true,
+ "span": 5,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_cpu_saturation_load1:{cluster=\"$cluster\"} / scalar(sum(min(kube_pod_info{cluster=\"$cluster\"}) by (node)))",
+ "expr": "histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\", instance=~\"$instance\"}[5m])) by (verb, le))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "{{verb}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -159,7 +281,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Saturation (Load1)",
+ "title": "Request duration 99th quantile",
"tooltip": {
"shared": false,
"sort": 0,
@@ -177,20 +299,20 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "s",
"label": null,
"logBase": 1,
- "max": 1,
- "min": 0,
+ "max": null,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
}
@@ -198,13 +320,14 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "CPU",
- "titleSize": "h6"
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
"collapse": false,
- "height": "250px",
+ "collapsed": false,
"panels": [
{
"aliasColors": {
@@ -214,42 +337,47 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 3,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 5,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
- "show": true,
+ "rightSide": false,
+ "show": false,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
"span": 6,
- "stack": true,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:cluster_memory_utilisation:ratio{cluster=\"$cluster\"}",
+ "expr": "sum(rate(workqueue_adds_total{job=\"apiserver\", instance=~\"$instance\"}[5m])) by (instance, name)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "{{instance}} {{name}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -257,7 +385,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Utilisation",
+ "title": "Work Queue Add Rate",
"tooltip": {
"shared": false,
"sort": 0,
@@ -275,20 +403,20 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "ops",
"label": null,
"logBase": 1,
- "max": 1,
+ "max": null,
"min": 0,
"show": true
},
{
- "format": "short",
+ "format": "ops",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
},
@@ -300,42 +428,47 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 4,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 6,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
- "show": true,
+ "rightSide": false,
+ "show": false,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
"span": 6,
- "stack": true,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\"}",
+ "expr": "sum(rate(workqueue_depth{job=\"apiserver\", instance=~\"$instance\"}[5m])) by (instance, name)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "{{instance}} {{name}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -343,7 +476,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Saturation (Swap I/O)",
+ "title": "Work Queue Depth",
"tooltip": {
"shared": false,
"sort": 0,
@@ -361,7 +494,7 @@ items:
},
"yaxes": [
{
- "format": "Bps",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -373,23 +506,11 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Memory",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -398,42 +519,47 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 5,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 7,
"legend": {
+ "alignAsTable": "true",
"avg": false,
- "current": false,
+ "current": "true",
"max": false,
"min": false,
- "show": true,
+ "rightSide": "true",
+ "show": "true",
"total": false,
- "values": false
+ "values": "true"
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
- "stack": true,
+ "span": 12,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})",
+ "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{job=\"apiserver\", instance=~\"$instance\"}[5m])) by (instance, name, le))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "{{instance}} {{name}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -441,7 +567,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk IO Utilisation",
+ "title": "Work Queue Latency",
"tooltip": {
"shared": false,
"sort": 0,
@@ -459,23 +585,36 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "s",
"label": null,
"logBase": 1,
- "max": 1,
- "min": 0,
+ "max": null,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
{
"aliasColors": {
@@ -484,42 +623,47 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 6,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 8,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
- "stack": true,
+ "span": 4,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})",
+ "expr": "etcd_helper_cache_entry_total{job=\"apiserver\", instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "{{instance}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -527,7 +671,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk IO Saturation",
+ "title": "ETCD Cache Entry Total",
"tooltip": {
"shared": false,
"sort": 0,
@@ -545,10 +689,10 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "short",
"label": null,
"logBase": 1,
- "max": 1,
+ "max": null,
"min": 0,
"show": true
},
@@ -557,23 +701,11 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Disk",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -582,42 +714,54 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 7,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 9,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
- "stack": true,
+ "span": 4,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\"}",
+ "expr": "sum(rate(etcd_helper_cache_hit_total{job=\"apiserver\",instance=~\"$instance\"}[5m])) by (intance)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "{{instance}} hit",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(etcd_helper_cache_miss_total{job=\"apiserver\",instance=~\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} miss",
+ "refId": "B"
}
],
"thresholds": [
@@ -625,7 +769,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Net Utilisation (Transmitted)",
+ "title": "ETCD Cache Hit/Miss Rate",
"tooltip": {
"shared": false,
"sort": 0,
@@ -643,7 +787,7 @@ items:
},
"yaxes": [
{
- "format": "Bps",
+ "format": "ops",
"label": null,
"logBase": 1,
"max": null,
@@ -651,12 +795,12 @@ items:
"show": true
},
{
- "format": "short",
+ "format": "ops",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
},
@@ -668,42 +812,54 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 8,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 10,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
- "stack": true,
+ "span": 4,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\"}",
+ "expr": "histogram_quantile(0.99,sum(rate(etcd_request_cache_get_duration_seconds_bucket{job=\"apiserver\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "{{instance}} get",
+ "refId": "A"
+ },
+ {
+ "expr": "histogram_quantile(0.99,sum(rate(etcd_request_cache_add_duration_seconds_bucket{job=\"apiserver\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} miss",
+ "refId": "B"
}
],
"thresholds": [
@@ -711,7 +867,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Net Saturation (Dropped)",
+ "title": "ETCD Cache Duration 99th Quantile",
"tooltip": {
"shared": false,
"sort": 0,
@@ -729,7 +885,7 @@ items:
},
"yaxes": [
{
- "format": "Bps",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
@@ -737,12 +893,12 @@ items:
"show": true
},
{
- "format": "short",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
}
@@ -750,13 +906,14 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
"collapse": false,
- "height": "250px",
+ "collapsed": false,
"panels": [
{
"aliasColors": {
@@ -766,42 +923,47 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 9,
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 11,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 12,
- "stack": true,
+ "span": 4,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"} - node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)) by (pod,namespace)\n/ scalar(sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)))\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\"}\n",
+ "expr": "process_resident_memory_bytes{job=\"apiserver\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{node}}",
- "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
- "step": 10
+ "legendFormat": "{{instance}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -809,7 +971,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk Capacity",
+ "title": "Memory",
"tooltip": {
"shared": false,
"sort": 0,
@@ -827,146 +989,25 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "bytes",
"label": null,
"logBase": 1,
- "max": 1,
- "min": 0,
+ "max": null,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Storage",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
- "kubernetes-mixin"
- ],
- "templating": {
- "list": [
- {
- "current": {
- "text": "Prometheus",
- "value": "Prometheus"
},
- "hide": 0,
- "label": null,
- "name": "datasource",
- "options": [
-
- ],
- "query": "prometheus",
- "refresh": 1,
- "regex": "",
- "type": "datasource"
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 2,
- "includeAll": false,
- "label": "cluster",
- "multi": false,
- "name": "cluster",
- "options": [
-
- ],
- "query": "label_values(kube_node_info, cluster)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "",
- "title": "Kubernetes / USE Method / Cluster",
- "uid": "a6e7d1362e1ddbb79db21d5bb40d7137",
- "version": 0
- }
- kind: ConfigMap
- metadata:
- name: grafana-dashboard-k8s-cluster-rsrc-use
- namespace: monitoring
-- apiVersion: v1
- data:
- k8s-node-rsrc-use.json: |-
- {
- "annotations": {
- "list": [
-
- ]
- },
- "editable": true,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [
-
- ],
- "refresh": "10s",
- "rows": [
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {
+ {
+ "aliasColors": {
},
"bars": false,
@@ -974,12 +1015,17 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 1,
+ "gridPos": {
+
+ },
+ "id": 12,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -989,26 +1035,26 @@ items:
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_cpu_utilisation:avg1m{cluster=\"$cluster\", node=\"$node\"}",
+ "expr": "rate(process_cpu_seconds_total{job=\"apiserver\",instance=~\"$instance\"}[5m])",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Utilisation",
- "legendLink": null,
- "step": 10
+ "legendFormat": "{{instance}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -1016,7 +1062,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Utilisation",
+ "title": "CPU usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -1034,7 +1080,7 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -1046,8 +1092,8 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
},
@@ -1060,12 +1106,17 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 2,
+ "gridPos": {
+
+ },
+ "id": 13,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -1075,26 +1126,26 @@ items:
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_cpu_saturation_load1:{cluster=\"$cluster\", node=\"$node\"}",
+ "expr": "go_goroutines{job=\"apiserver\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Saturation",
- "legendLink": null,
- "step": 10
+ "legendFormat": "{{instance}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -1102,7 +1153,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Saturation (Load1)",
+ "title": "Goroutines",
"tooltip": {
"shared": false,
"sort": 0,
@@ -1120,11 +1171,11 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
@@ -1133,7 +1184,7 @@ items:
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
}
@@ -1141,99 +1192,213 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "CPU",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 1,
- "id": 3,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
- ],
- "nullPointMode": "null as zero",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "node:node_memory_utilisation:{cluster=\"$cluster\", node=\"$node\"}",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Memory",
- "legendLink": null,
- "step": 10
- }
- ],
- "thresholds": [
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "instance",
+ "options": [
- ],
- "timeFrom": null,
- "timeShift": null,
- "title": "Memory Utilisation",
- "tooltip": {
- "shared": false,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": [
+ ],
+ "query": "label_values(apiserver_request_total{job=\"apiserver\"}, instance)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
- ]
- },
- "yaxes": [
- {
- "format": "percentunit",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": 0,
- "show": true
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / API server",
+ "uid": "09ec8aa1e996d6ffcd6817bbaff4db1b",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-apiserver
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ controller-manager.json: |-
+ {
+ "__inputs": [
+
+ ],
+ "__requires": [
+
+ ],
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
},
{
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
+ "name": "range to text",
+ "value": 2
}
- ]
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(up{job=\"kube-controller-manager\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Up",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
},
{
"aliasColors": {
@@ -1244,41 +1409,46 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 4,
+ "gridPos": {
+
+ },
+ "id": 3,
"legend": {
+ "alignAsTable": "true",
"avg": false,
- "current": false,
+ "current": "true",
"max": false,
"min": false,
- "show": true,
+ "rightSide": "true",
+ "show": "true",
"total": false,
- "values": false
+ "values": "true"
},
"lines": true,
"linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 10,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\", node=\"$node\"}",
+ "expr": "sum(rate(workqueue_adds_total{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Swap IO",
- "legendLink": null,
- "step": 10
+ "legendFormat": "{{instance}} {{name}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -1286,7 +1456,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Saturation (Swap I/O)",
+ "title": "Work Queue Add Rate",
"tooltip": {
"shared": false,
"sort": 0,
@@ -1304,20 +1474,20 @@ items:
},
"yaxes": [
{
- "format": "Bps",
+ "format": "ops",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "ops",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
}
@@ -1325,13 +1495,14 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "Memory",
- "titleSize": "h6"
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
"collapse": false,
- "height": "250px",
+ "collapsed": false,
"panels": [
{
"aliasColors": {
@@ -1342,41 +1513,46 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 5,
+ "gridPos": {
+
+ },
+ "id": 4,
"legend": {
+ "alignAsTable": "true",
"avg": false,
- "current": false,
+ "current": "true",
"max": false,
"min": false,
- "show": true,
+ "rightSide": "true",
+ "show": "true",
"total": false,
- "values": false
+ "values": "true"
},
"lines": true,
"linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 12,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\", node=\"$node\"}",
+ "expr": "sum(rate(workqueue_depth{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Utilisation",
- "legendLink": null,
- "step": 10
+ "legendFormat": "{{instance}} {{name}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -1384,7 +1560,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk IO Utilisation",
+ "title": "Work Queue Depth",
"tooltip": {
"shared": false,
"sort": 0,
@@ -1402,7 +1578,7 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -1414,11 +1590,24 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
{
"aliasColors": {
@@ -1428,41 +1617,46 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 6,
+ "gridPos": {
+
+ },
+ "id": 5,
"legend": {
+ "alignAsTable": "true",
"avg": false,
- "current": false,
+ "current": "true",
"max": false,
"min": false,
- "show": true,
+ "rightSide": "true",
+ "show": "true",
"total": false,
- "values": false
+ "values": "true"
},
"lines": true,
"linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 12,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\", node=\"$node\"}",
+ "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name, le))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Saturation",
- "legendLink": null,
- "step": 10
+ "legendFormat": "{{instance}} {{name}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -1470,7 +1664,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk IO Saturation",
+ "title": "Work Queue Latency",
"tooltip": {
"shared": false,
"sort": 0,
@@ -1488,20 +1682,20 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
}
@@ -1509,13 +1703,14 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "Disk",
- "titleSize": "h6"
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
"collapse": false,
- "height": "250px",
+ "collapsed": false,
"panels": [
{
"aliasColors": {
@@ -1526,12 +1721,17 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 7,
+ "gridPos": {
+
+ },
+ "id": 6,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -1541,26 +1741,47 @@ items:
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\", node=\"$node\"}",
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Utilisation",
- "legendLink": null,
- "step": 10
+ "legendFormat": "2xx",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "3xx",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "4xx",
+ "refId": "C"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "5xx",
+ "refId": "D"
}
],
"thresholds": [
@@ -1568,7 +1789,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Net Utilisation (Transmitted)",
+ "title": "Kube API Request Rate",
"tooltip": {
"shared": false,
"sort": 0,
@@ -1586,20 +1807,20 @@ items:
},
"yaxes": [
{
- "format": "Bps",
+ "format": "ops",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "ops",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
},
@@ -1612,12 +1833,17 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 8,
+ "gridPos": {
+
+ },
+ "id": 7,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -1627,26 +1853,26 @@ items:
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 8,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\", node=\"$node\"}",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Saturation",
- "legendLink": null,
- "step": 10
+ "legendFormat": "{{verb}} {{url}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -1654,7 +1880,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Net Saturation (Dropped)",
+ "title": "Post Request Latency 99th Quantile",
"tooltip": {
"shared": false,
"sort": 0,
@@ -1672,7 +1898,7 @@ items:
},
"yaxes": [
{
- "format": "Bps",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
@@ -1680,12 +1906,12 @@ items:
"show": true
},
{
- "format": "short",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
}
@@ -1693,13 +1919,14 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "Net",
- "titleSize": "h6"
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
"collapse": false,
- "height": "250px",
+ "collapsed": false,
"panels": [
{
"aliasColors": {
@@ -1710,26 +1937,32 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 9,
+ "gridPos": {
+
+ },
+ "id": 8,
"legend": {
+ "alignAsTable": "true",
"avg": false,
- "current": false,
+ "current": "true",
"max": false,
"min": false,
- "show": true,
+ "rightSide": "true",
+ "show": "true",
"total": false,
- "values": false
+ "values": "true"
},
"lines": true,
"linewidth": 1,
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
@@ -1739,12 +1972,11 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "node:node_filesystem_usage:{cluster=\"$cluster\"}\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\", node=\"$node\"}\n",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{device}}",
- "legendLink": null,
- "step": 10
+ "legendFormat": "{{verb}} {{url}}",
+ "refId": "A"
}
],
"thresholds": [
@@ -1752,7 +1984,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk Utilisation",
+ "title": "Get Request Latency 99th Quantile",
"tooltip": {
"shared": false,
"sort": 0,
@@ -1770,7 +2002,7 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
@@ -1778,12 +2010,12 @@ items:
"show": true
},
{
- "format": "short",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
}
@@ -1791,206 +2023,78 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "Disk",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
- "kubernetes-mixin"
- ],
- "templating": {
- "list": [
- {
- "current": {
- "text": "Prometheus",
- "value": "Prometheus"
- },
- "hide": 0,
- "label": null,
- "name": "datasource",
- "options": [
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
- ],
- "query": "prometheus",
- "refresh": 1,
- "regex": "",
- "type": "datasource"
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 2,
- "includeAll": false,
- "label": "cluster",
- "multi": false,
- "name": "cluster",
- "options": [
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
- ],
- "query": "label_values(kube_node_info, cluster)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "node",
- "multi": false,
- "name": "node",
- "options": [
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
- ],
- "query": "label_values(kube_node_info{cluster=\"$cluster\"}, node)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "process_resident_memory_bytes{job=\"kube-controller-manager\",instance=~\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "",
- "title": "Kubernetes / USE Method / Node",
- "uid": "4ac4f123aae0ff6dbaf4f4f66120033b",
- "version": 0
- }
- kind: ConfigMap
- metadata:
- name: grafana-dashboard-k8s-node-rsrc-use
- namespace: monitoring
-- apiVersion: v1
- data:
- k8s-resources-cluster.json: |-
- {
- "annotations": {
- "list": [
-
- ]
- },
- "editable": true,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [
-
- ],
- "refresh": "10s",
- "rows": [
- {
- "collapse": false,
- "height": "100px",
- "panels": [
- {
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 1,
- "format": "percentunit",
- "id": 1,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [
-
- ],
- "nullPointMode": "null as zero",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
-
- ],
- "spaceLength": 10,
- "span": 2,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "1 - avg(rate(node_cpu_seconds_total{mode=\"idle\", cluster=\"$cluster\"}[1m]))",
- "format": "time_series",
- "instant": true,
- "intervalFactor": 2,
- "refId": "A"
- }
],
- "thresholds": "70,80",
"timeFrom": null,
"timeShift": null,
- "title": "CPU Utilisation",
+ "title": "Memory",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "singlestat",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -2002,20 +2106,20 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
},
@@ -2028,13 +2132,17 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "format": "percentunit",
- "id": 2,
+ "gridPos": {
+
+ },
+ "id": 10,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -2044,37 +2152,40 @@ items:
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 2,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) / sum(node:node_num_cpu:sum{cluster=\"$cluster\"})",
+ "expr": "rate(process_cpu_seconds_total{job=\"kube-controller-manager\",instance=~\"$instance\"}[5m])",
"format": "time_series",
- "instant": true,
"intervalFactor": 2,
+ "legendFormat": "{{instance}}",
"refId": "A"
}
],
- "thresholds": "70,80",
+ "thresholds": [
+
+ ],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Requests Commitment",
+ "title": "CPU usage",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "singlestat",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -2098,8 +2209,8 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
- "show": false
+ "min": 0,
+ "show": true
}
]
},
@@ -2112,13 +2223,17 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "format": "percentunit",
- "id": 3,
+ "gridPos": {
+
+ },
+ "id": 11,
"legend": {
+ "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -2128,37 +2243,40 @@ items:
"links": [
],
- "nullPointMode": "null as zero",
+ "nullPointMode": "null",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
+ "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 2,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) / sum(node:node_num_cpu:sum{cluster=\"$cluster\"})",
+ "expr": "go_goroutines{job=\"kube-controller-manager\",instance=~\"$instance\"}",
"format": "time_series",
- "instant": true,
"intervalFactor": 2,
+ "legendFormat": "{{instance}}",
"refId": "A"
}
],
- "thresholds": "70,80",
+ "thresholds": [
+
+ ],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Limits Commitment",
+ "title": "Goroutines",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "singlestat",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -2174,7 +2292,7 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
@@ -2183,66 +2301,189 @@ items:
"logBase": 1,
"max": null,
"min": null,
- "show": false
+ "show": true
}
]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
},
- {
- "aliasColors": {
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 1,
- "format": "percentunit",
- "id": 4,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
- ],
- "nullPointMode": "null as zero",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "instance",
+ "options": [
- ],
- "spaceLength": 10,
- "span": 2,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "1 - sum(:node_memory_MemFreeCachedBuffers_bytes:sum{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})",
- "format": "time_series",
- "instant": true,
- "intervalFactor": 2,
- "refId": "A"
- }
- ],
- "thresholds": "70,80",
+ ],
+ "query": "label_values(process_cpu_seconds_total{job=\"kube-controller-manager\"}, instance)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Controller Manager",
+ "uid": "72e0e05bef5099e5f049b05fdc429ed4",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-controller-manager
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ k8s-cluster-rsrc-use.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [
+
+ ],
+ "refresh": "10s",
+ "rows": [
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 10,
+ "id": 1,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [
+
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node:cluster_cpu_utilisation:ratio{cluster=\"$cluster\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Utilisation",
+ "title": "CPU Utilisation",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "singlestat",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -2254,10 +2495,10 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "percentunit",
"label": null,
"logBase": 1,
- "max": null,
+ "max": 1,
"min": 0,
"show": true
},
@@ -2279,9 +2520,8 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
- "format": "percentunit",
- "id": 5,
+ "fill": 10,
+ "id": 2,
"legend": {
"avg": false,
"current": false,
@@ -2292,7 +2532,7 @@ items:
"values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
@@ -2305,28 +2545,31 @@ items:
],
"spaceLength": 10,
- "span": 2,
- "stack": false,
+ "span": 6,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})",
+ "expr": "node:node_cpu_saturation_load1:{cluster=\"$cluster\"} / scalar(sum(min(kube_pod_info{cluster=\"$cluster\"}) by (node)))",
"format": "time_series",
- "instant": true,
"intervalFactor": 2,
- "refId": "A"
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
+ "step": 10
}
],
- "thresholds": "70,80",
+ "thresholds": [
+
+ ],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Requests Commitment",
+ "title": "CPU Saturation (Load1)",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "singlestat",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -2338,10 +2581,10 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "percentunit",
"label": null,
"logBase": 1,
- "max": null,
+ "max": 1,
"min": 0,
"show": true
},
@@ -2354,7 +2597,19 @@ items:
"show": false
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "CPU",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
{
"aliasColors": {
@@ -2363,9 +2618,8 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
- "format": "percentunit",
- "id": 6,
+ "fill": 10,
+ "id": 3,
"legend": {
"avg": false,
"current": false,
@@ -2376,7 +2630,7 @@ items:
"values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
@@ -2389,28 +2643,31 @@ items:
],
"spaceLength": 10,
- "span": 2,
- "stack": false,
+ "span": 6,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})",
+ "expr": "node:cluster_memory_utilisation:ratio{cluster=\"$cluster\"}",
"format": "time_series",
- "instant": true,
"intervalFactor": 2,
- "refId": "A"
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
+ "step": 10
}
],
- "thresholds": "70,80",
+ "thresholds": [
+
+ ],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Limits Commitment",
+ "title": "Memory Utilisation",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "singlestat",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -2422,10 +2679,10 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "percentunit",
"label": null,
"logBase": 1,
- "max": null,
+ "max": 1,
"min": 0,
"show": true
},
@@ -2438,19 +2695,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Headlines",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -2460,7 +2705,7 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 10,
- "id": 7,
+ "id": 4,
"legend": {
"avg": false,
"current": false,
@@ -2484,16 +2729,16 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)",
+ "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{namespace}}",
- "legendLink": null,
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
"step": 10
}
],
@@ -2502,7 +2747,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Usage",
+ "title": "Memory Saturation (Swap I/O)",
"tooltip": {
"shared": false,
"sort": 0,
@@ -2520,7 +2765,7 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "Bps",
"label": null,
"logBase": 1,
"max": null,
@@ -2542,7 +2787,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "CPU",
+ "title": "Memory",
"titleSize": "h6"
},
{
@@ -2557,8 +2802,8 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
- "id": 8,
+ "fill": 10,
+ "id": 5,
"legend": {
"avg": false,
"current": false,
@@ -2569,7 +2814,7 @@ items:
"values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
@@ -2582,238 +2827,16 @@ items:
],
"spaceLength": 10,
- "span": 12,
- "stack": false,
+ "span": 6,
+ "stack": true,
"steppedLine": false,
- "styles": [
- {
- "alias": "Time",
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "pattern": "Time",
- "type": "hidden"
- },
+ "targets": [
{
- "alias": "Pods",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 0,
- "link": true,
- "linkTooltip": "Drill down to pods",
- "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1",
- "pattern": "Value #A",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "Workloads",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 0,
- "link": true,
- "linkTooltip": "Drill down to workloads",
- "linkUrl": "/d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1",
- "pattern": "Value #B",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "CPU Usage",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #C",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "CPU Requests",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #D",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "CPU Requests %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #E",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "CPU Limits",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #F",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "CPU Limits %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #G",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "Namespace",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": true,
- "linkTooltip": "Drill down to pods",
- "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell",
- "pattern": "namespace",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "pattern": "/.*/",
- "thresholds": [
-
- ],
- "type": "string",
- "unit": "short"
- }
- ],
- "targets": [
- {
- "expr": "count(mixin_pod_workload{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 10
- },
- {
- "expr": "count(avg(mixin_pod_workload{cluster=\"$cluster\"}) by (workload, namespace)) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "B",
- "step": 10
- },
- {
- "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "C",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "D",
- "step": 10
- },
- {
- "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "E",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "F",
- "step": 10
- },
- {
- "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
+ "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})",
+ "format": "time_series",
"intervalFactor": 2,
- "legendFormat": "",
- "refId": "G",
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
"step": 10
}
],
@@ -2822,14 +2845,13 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Quota",
+ "title": "Disk IO Utilisation",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "transform": "table",
- "type": "table",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -2841,10 +2863,10 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "percentunit",
"label": null,
"logBase": 1,
- "max": null,
+ "max": 1,
"min": 0,
"show": true
},
@@ -2857,19 +2879,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "CPU Quota",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -2879,7 +2889,7 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 10,
- "id": 9,
+ "id": 6,
"legend": {
"avg": false,
"current": false,
@@ -2903,16 +2913,16 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)",
+ "expr": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{namespace}}",
- "legendLink": null,
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
"step": 10
}
],
@@ -2921,7 +2931,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage (w/o cache)",
+ "title": "Disk IO Saturation",
"tooltip": {
"shared": false,
"sort": 0,
@@ -2939,10 +2949,10 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "percentunit",
"label": null,
"logBase": 1,
- "max": null,
+ "max": 1,
"min": 0,
"show": true
},
@@ -2961,7 +2971,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Memory",
+ "title": "Disk",
"titleSize": "h6"
},
{
@@ -2976,8 +2986,8 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
- "id": 10,
+ "fill": 10,
+ "id": 7,
"legend": {
"avg": false,
"current": false,
@@ -2988,7 +2998,7 @@ items:
"values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
@@ -3001,238 +3011,16 @@ items:
],
"spaceLength": 10,
- "span": 12,
- "stack": false,
+ "span": 6,
+ "stack": true,
"steppedLine": false,
- "styles": [
- {
- "alias": "Time",
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "pattern": "Time",
- "type": "hidden"
- },
- {
- "alias": "Pods",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 0,
- "link": true,
- "linkTooltip": "Drill down to pods",
- "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1",
- "pattern": "Value #A",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "Workloads",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 0,
- "link": true,
- "linkTooltip": "Drill down to workloads",
- "linkUrl": "/d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1",
- "pattern": "Value #B",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "Memory Usage",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #C",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "bytes"
- },
- {
- "alias": "Memory Requests",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #D",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "bytes"
- },
- {
- "alias": "Memory Requests %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #E",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "Memory Limits",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #F",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "bytes"
- },
- {
- "alias": "Memory Limits %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #G",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "Namespace",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": true,
- "linkTooltip": "Drill down to pods",
- "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell",
- "pattern": "namespace",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "pattern": "/.*/",
- "thresholds": [
-
- ],
- "type": "string",
- "unit": "short"
- }
- ],
"targets": [
{
- "expr": "count(mixin_pod_workload{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 10
- },
- {
- "expr": "count(avg(mixin_pod_workload{cluster=\"$cluster\"}) by (workload, namespace)) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "B",
- "step": 10
- },
- {
- "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "C",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "D",
- "step": 10
- },
- {
- "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "E",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "F",
- "step": 10
- },
- {
- "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)",
- "format": "table",
- "instant": true,
+ "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\"}",
+ "format": "time_series",
"intervalFactor": 2,
- "legendFormat": "",
- "refId": "G",
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
"step": 10
}
],
@@ -3241,14 +3029,13 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Requests by Namespace",
+ "title": "Net Utilisation (Transmitted)",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "transform": "table",
- "type": "table",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -3260,7 +3047,7 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "Bps",
"label": null,
"logBase": 1,
"max": null,
@@ -3276,22 +3063,206 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Memory Requests",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
- "kubernetes-mixin"
- ],
- "templating": {
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 10,
+ "id": 8,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [
+
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Net Saturation (Dropped)",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Network",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 10,
+ "id": 9,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [
+
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"} - node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)) by (pod,namespace)\n/ scalar(sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)))\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\"}\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{node}}",
+ "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use",
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk Capacity",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": 1,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Storage",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
"list": [
{
"current": {
@@ -3324,7 +3295,7 @@ items:
"options": [
],
- "query": "label_values(node_cpu_seconds_total, cluster)",
+ "query": "label_values(kube_node_info, cluster)",
"refresh": 1,
"regex": "",
"sort": 2,
@@ -3368,17 +3339,17 @@ items:
]
},
"timezone": "",
- "title": "Kubernetes / Compute Resources / Cluster",
- "uid": "efa86fd1d0c121a26444b636a3f509a8",
+ "title": "Kubernetes / USE Method / Cluster",
+ "uid": "a6e7d1362e1ddbb79db21d5bb40d7137",
"version": 0
}
kind: ConfigMap
metadata:
- name: grafana-dashboard-k8s-resources-cluster
+ name: grafana-dashboard-k8s-cluster-rsrc-use
namespace: monitoring
- apiVersion: v1
data:
- k8s-resources-namespace.json: |-
+ k8s-node-rsrc-use.json: |-
{
"annotations": {
"list": [
@@ -3406,7 +3377,7 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
+ "fill": 1,
"id": 1,
"legend": {
"avg": false,
@@ -3418,7 +3389,7 @@ items:
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
@@ -3431,15 +3402,15 @@ items:
],
"spaceLength": 10,
- "span": 12,
- "stack": true,
+ "span": 6,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod_name)",
+ "expr": "node:node_cpu_utilisation:avg1m{cluster=\"$cluster\", node=\"$node\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{pod_name}}",
+ "legendFormat": "Utilisation",
"legendLink": null,
"step": 10
}
@@ -3449,7 +3420,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Usage",
+ "title": "CPU Utilisation",
"tooltip": {
"shared": false,
"sort": 0,
@@ -3467,7 +3438,7 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "percentunit",
"label": null,
"logBase": 1,
"max": null,
@@ -3483,19 +3454,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "CPU Usage",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -3529,184 +3488,16 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": false,
"steppedLine": false,
- "styles": [
- {
- "alias": "Time",
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "pattern": "Time",
- "type": "hidden"
- },
+ "targets": [
{
- "alias": "CPU Usage",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #A",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "CPU Requests",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #B",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "CPU Requests %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #C",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "CPU Limits",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #D",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "CPU Limits %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #E",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "Pod",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": true,
- "linkTooltip": "Drill down",
- "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell",
- "pattern": "pod",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "pattern": "/.*/",
- "thresholds": [
-
- ],
- "type": "string",
- "unit": "short"
- }
- ],
- "targets": [
- {
- "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "B",
- "step": 10
- },
- {
- "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "C",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "D",
- "step": 10
- },
- {
- "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
- "format": "table",
- "instant": true,
+ "expr": "node:node_cpu_saturation_load1:{cluster=\"$cluster\", node=\"$node\"}",
+ "format": "time_series",
"intervalFactor": 2,
- "legendFormat": "",
- "refId": "E",
+ "legendFormat": "Saturation",
+ "legendLink": null,
"step": 10
}
],
@@ -3715,14 +3506,13 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Quota",
+ "title": "CPU Saturation (Load1)",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "transform": "table",
- "type": "table",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -3734,7 +3524,7 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "percentunit",
"label": null,
"logBase": 1,
"max": null,
@@ -3756,7 +3546,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "CPU Quota",
+ "title": "CPU",
"titleSize": "h6"
},
{
@@ -3771,7 +3561,7 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
+ "fill": 1,
"id": 3,
"legend": {
"avg": false,
@@ -3783,7 +3573,7 @@ items:
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
@@ -3796,15 +3586,15 @@ items:
],
"spaceLength": 10,
- "span": 12,
- "stack": true,
+ "span": 6,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"}) by (pod_name)",
+ "expr": "node:node_memory_utilisation:{cluster=\"$cluster\", node=\"$node\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{pod_name}}",
+ "legendFormat": "Memory",
"legendLink": null,
"step": 10
}
@@ -3814,7 +3604,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage (w/o cache)",
+ "title": "Memory Utilisation",
"tooltip": {
"shared": false,
"sort": 0,
@@ -3832,7 +3622,7 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "percentunit",
"label": null,
"logBase": 1,
"max": null,
@@ -3848,19 +3638,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Memory Usage",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -3894,265 +3672,384 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": false,
"steppedLine": false,
- "styles": [
- {
- "alias": "Time",
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "pattern": "Time",
- "type": "hidden"
- },
+ "targets": [
{
- "alias": "Memory Usage",
- "colorMode": null,
- "colors": [
+ "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\", node=\"$node\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Swap IO",
+ "legendLink": null,
+ "step": 10
+ }
+ ],
+ "thresholds": [
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #A",
- "thresholds": [
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Saturation (Swap I/O)",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
- ],
- "type": "number",
- "unit": "bytes"
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
},
{
- "alias": "Memory Requests",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #B",
- "thresholds": [
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Memory",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
- ],
- "type": "number",
- "unit": "bytes"
- },
- {
- "alias": "Memory Requests %",
- "colorMode": null,
- "colors": [
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "id": 5,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #C",
- "thresholds": [
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
- ],
- "type": "number",
- "unit": "percentunit"
- },
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
{
- "alias": "Memory Limits",
- "colorMode": null,
- "colors": [
+ "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\", node=\"$node\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Utilisation",
+ "legendLink": null,
+ "step": 10
+ }
+ ],
+ "thresholds": [
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #D",
- "thresholds": [
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk IO Utilisation",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
- ],
- "type": "number",
- "unit": "bytes"
- },
+ ]
+ },
+ "yaxes": [
{
- "alias": "Memory Limits %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #E",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "percentunit"
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
},
{
- "alias": "Memory Usage (RSS)",
- "colorMode": null,
- "colors": [
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "aliasColors": {
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #F",
- "thresholds": [
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "id": 6,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
- ],
- "type": "number",
- "unit": "bytes"
- },
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
{
- "alias": "Memory Usage (Cache)",
- "colorMode": null,
- "colors": [
+ "expr": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\", node=\"$node\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Saturation",
+ "legendLink": null,
+ "step": 10
+ }
+ ],
+ "thresholds": [
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #G",
- "thresholds": [
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk IO Saturation",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
- ],
- "type": "number",
- "unit": "bytes"
- },
+ ]
+ },
+ "yaxes": [
{
- "alias": "Memory Usage (Swap",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #H",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "bytes"
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
},
{
- "alias": "Pod",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": true,
- "linkTooltip": "Drill down",
- "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell",
- "pattern": "pod",
- "thresholds": [
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Disk",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "",
- "colorMode": null,
- "colors": [
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "id": 7,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "pattern": "/.*/",
- "thresholds": [
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
- ],
- "type": "string",
- "unit": "short"
- }
],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
"targets": [
{
- "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "B",
- "step": 10
- },
- {
- "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\"}) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "C",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "D",
- "step": 10
- },
- {
- "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\"}) by (pod)",
- "format": "table",
- "instant": true,
+ "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\", node=\"$node\"}",
+ "format": "time_series",
"intervalFactor": 2,
- "legendFormat": "",
- "refId": "E",
+ "legendFormat": "Utilisation",
+ "legendLink": null,
"step": 10
- },
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Net Utilisation (Transmitted)",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
{
- "expr": "sum(label_replace(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "F",
- "step": 10
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
},
{
- "expr": "sum(label_replace(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "G",
- "step": 10
- },
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "id": 8,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
{
- "expr": "sum(label_replace(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)",
- "format": "table",
- "instant": true,
+ "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\", node=\"$node\"}",
+ "format": "time_series",
"intervalFactor": 2,
- "legendFormat": "",
- "refId": "H",
+ "legendFormat": "Saturation",
+ "legendLink": null,
"step": 10
}
],
@@ -4161,14 +4058,13 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Quota",
+ "title": "Net Saturation (Dropped)",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "transform": "table",
- "type": "table",
+ "type": "graph",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -4180,7 +4076,7 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "Bps",
"label": null,
"logBase": 1,
"max": null,
@@ -4202,14 +4098,112 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Memory Quota",
+ "title": "Net",
"titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
- "kubernetes-mixin"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "id": 9,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node:node_filesystem_usage:{cluster=\"$cluster\"}\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\", node=\"$node\"}\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "legendLink": null,
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk Utilisation",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Disk",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
],
"templating": {
"list": [
@@ -4244,7 +4238,7 @@ items:
"options": [
],
- "query": "label_values(kube_pod_info, cluster)",
+ "query": "label_values(kube_node_info, cluster)",
"refresh": 1,
"regex": "",
"sort": 2,
@@ -4265,13 +4259,13 @@ items:
"datasource": "$datasource",
"hide": 0,
"includeAll": false,
- "label": "namespace",
+ "label": "node",
"multi": false,
- "name": "namespace",
+ "name": "node",
"options": [
],
- "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
+ "query": "label_values(kube_node_info{cluster=\"$cluster\"}, node)",
"refresh": 1,
"regex": "",
"sort": 2,
@@ -4315,17 +4309,17 @@ items:
]
},
"timezone": "",
- "title": "Kubernetes / Compute Resources / Namespace (Pods)",
- "uid": "85a562078cdf77779eaa1add43ccec1e",
+ "title": "Kubernetes / USE Method / Node",
+ "uid": "4ac4f123aae0ff6dbaf4f4f66120033b",
"version": 0
}
kind: ConfigMap
metadata:
- name: grafana-dashboard-k8s-resources-namespace
+ name: grafana-dashboard-k8s-node-rsrc-use
namespace: monitoring
- apiVersion: v1
data:
- k8s-resources-pod.json: |-
+ k8s-resources-cluster.json: |-
{
"annotations": {
"list": [
@@ -4343,7 +4337,7 @@ items:
"rows": [
{
"collapse": false,
- "height": "250px",
+ "height": "100px",
"panels": [
{
"aliasColors": {
@@ -4353,7 +4347,8 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
+ "fill": 1,
+ "format": "percentunit",
"id": 1,
"legend": {
"avg": false,
@@ -4365,7 +4360,7 @@ items:
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
@@ -4378,31 +4373,28 @@ items:
],
"spaceLength": 10,
- "span": 12,
- "stack": true,
+ "span": 2,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", cluster=\"$cluster\"}) by (container_name)",
+ "expr": "1 - avg(rate(node_cpu_seconds_total{mode=\"idle\", cluster=\"$cluster\"}[1m]))",
"format": "time_series",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "{{container_name}}",
- "legendLink": null,
- "step": 10
+ "refId": "A"
}
],
- "thresholds": [
-
- ],
+ "thresholds": "70,80",
"timeFrom": null,
"timeShift": null,
- "title": "CPU Usage",
+ "title": "CPU Utilisation",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "graph",
+ "type": "singlestat",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -4430,19 +4422,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "CPU Usage",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -4452,6 +4432,7 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
+ "format": "percentunit",
"id": 2,
"legend": {
"avg": false,
@@ -4476,200 +4457,112 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 2,
"stack": false,
"steppedLine": false,
- "styles": [
- {
- "alias": "Time",
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "pattern": "Time",
- "type": "hidden"
- },
+ "targets": [
{
- "alias": "CPU Usage",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #A",
- "thresholds": [
+ "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_cpu_cores{cluster=\"$cluster\"})",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 2,
+ "refId": "A"
+ }
+ ],
+ "thresholds": "70,80",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Requests Commitment",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "singlestat",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
- ],
- "type": "number",
- "unit": "short"
- },
+ ]
+ },
+ "yaxes": [
{
- "alias": "CPU Requests",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #B",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
},
{
- "alias": "CPU Requests %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #C",
- "thresholds": [
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "aliasColors": {
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "CPU Limits",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #D",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "CPU Limits %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #E",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "Container",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "container",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "",
- "colorMode": null,
- "colors": [
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "format": "percentunit",
+ "id": 3,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "pattern": "/.*/",
- "thresholds": [
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
- ],
- "type": "string",
- "unit": "short"
- }
],
+ "spaceLength": 10,
+ "span": 2,
+ "stack": false,
+ "steppedLine": false,
"targets": [
{
- "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "B",
- "step": 10
- },
- {
- "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "C",
- "step": 10
- },
- {
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "D",
- "step": 10
- },
- {
- "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
- "format": "table",
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_cpu_cores{cluster=\"$cluster\"})",
+ "format": "time_series",
"instant": true,
"intervalFactor": 2,
- "legendFormat": "",
- "refId": "E",
- "step": 10
+ "refId": "A"
}
],
- "thresholds": [
-
- ],
+ "thresholds": "70,80",
"timeFrom": null,
"timeShift": null,
- "title": "CPU Quota",
+ "title": "CPU Limits Commitment",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "transform": "table",
- "type": "table",
+ "type": "singlestat",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -4697,19 +4590,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "CPU Quota",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -4718,8 +4599,9 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 10,
- "id": 3,
+ "fill": 1,
+ "format": "percentunit",
+ "id": 4,
"legend": {
"avg": false,
"current": false,
@@ -4730,7 +4612,7 @@ items:
"values": false
},
"lines": true,
- "linewidth": 0,
+ "linewidth": 1,
"links": [
],
@@ -4743,47 +4625,28 @@ items:
],
"spaceLength": 10,
- "span": 12,
- "stack": true,
+ "span": 2,
+ "stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{container_name}} (RSS)",
- "legendLink": null,
- "step": 10
- },
- {
- "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{container_name}} (Cache)",
- "legendLink": null,
- "step": 10
- },
- {
- "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)",
+ "expr": "1 - sum(:node_memory_MemFreeCachedBuffers_bytes:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})",
"format": "time_series",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "{{container_name}} (Swap)",
- "legendLink": null,
- "step": 10
+ "refId": "A"
}
],
- "thresholds": [
-
- ],
+ "thresholds": "70,80",
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage",
+ "title": "Memory Utilisation",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "graph",
+ "type": "singlestat",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -4795,7 +4658,7 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -4811,19 +4674,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Memory Usage",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -4833,7 +4684,8 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 4,
+ "format": "percentunit",
+ "id": 5,
"legend": {
"avg": false,
"current": false,
@@ -4857,146 +4709,333 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 2,
"stack": false,
"steppedLine": false,
- "styles": [
- {
- "alias": "Time",
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "pattern": "Time",
- "type": "hidden"
- },
+ "targets": [
{
- "alias": "Memory Usage",
- "colorMode": null,
- "colors": [
+ "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 2,
+ "refId": "A"
+ }
+ ],
+ "thresholds": "70,80",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Requests Commitment",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "singlestat",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #A",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "bytes"
- },
+ ]
+ },
+ "yaxes": [
{
- "alias": "Memory Requests",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #B",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "bytes"
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
},
{
- "alias": "Memory Requests %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #C",
- "thresholds": [
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "aliasColors": {
- ],
- "type": "number",
- "unit": "percentunit"
- },
- {
- "alias": "Memory Limits",
- "colorMode": null,
- "colors": [
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "format": "percentunit",
+ "id": 6,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #D",
- "thresholds": [
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
- ],
- "type": "number",
- "unit": "bytes"
- },
+ ],
+ "spaceLength": 10,
+ "span": 2,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
{
- "alias": "Memory Limits %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #E",
- "thresholds": [
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 2,
+ "refId": "A"
+ }
+ ],
+ "thresholds": "70,80",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Limits Commitment",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "singlestat",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
- ],
- "type": "number",
- "unit": "percentunit"
- },
+ ]
+ },
+ "yaxes": [
{
- "alias": "Memory Usage (RSS)",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #F",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "bytes"
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
},
{
- "alias": "Memory Usage (Cache)",
- "colorMode": null,
- "colors": [
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Headlines",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #G",
- "thresholds": [
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 10,
+ "id": 7,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [
- ],
- "type": "number",
- "unit": "bytes"
- },
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
{
- "alias": "Memory Usage (Swap",
- "colorMode": null,
- "colors": [
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{namespace}}",
+ "legendLink": null,
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "CPU",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "id": 8,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "styles": [
+ {
+ "alias": "Time",
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "pattern": "Time",
+ "type": "hidden"
+ },
+ {
+ "alias": "Pods",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "link": true,
+ "linkTooltip": "Drill down to pods",
+ "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1",
+ "pattern": "Value #A",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Workloads",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "link": true,
+ "linkTooltip": "Drill down to workloads",
+ "linkUrl": "/d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1",
+ "pattern": "Value #B",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Usage",
+ "colorMode": null,
+ "colors": [
],
"dateFormat": "YYYY-MM-DD HH:mm:ss",
@@ -5004,15 +5043,15 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "Value #H",
+ "pattern": "Value #C",
"thresholds": [
],
"type": "number",
- "unit": "bytes"
+ "unit": "short"
},
{
- "alias": "Container",
+ "alias": "CPU Requests",
"colorMode": null,
"colors": [
@@ -5022,7 +5061,79 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "container",
+ "pattern": "Value #D",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Requests %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #E",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "CPU Limits",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #F",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Limits %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #G",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "Namespace",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": true,
+ "linkTooltip": "Drill down to pods",
+ "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell",
+ "pattern": "namespace",
"thresholds": [
],
@@ -5047,7 +5158,7 @@ items:
],
"targets": [
{
- "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)",
+ "expr": "count(mixin_pod_workload{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5056,7 +5167,7 @@ items:
"step": 10
},
{
- "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
+ "expr": "count(avg(mixin_pod_workload{cluster=\"$cluster\"}) by (workload, namespace)) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5065,7 +5176,7 @@ items:
"step": 10
},
{
- "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5074,7 +5185,7 @@ items:
"step": 10
},
{
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\"}) by (container)",
+ "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5083,7 +5194,7 @@ items:
"step": 10
},
{
- "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5092,7 +5203,7 @@ items:
"step": 10
},
{
- "expr": "sum(label_replace(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)",
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5101,22 +5212,13 @@ items:
"step": 10
},
{
- "expr": "sum(label_replace(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
"legendFormat": "",
"refId": "G",
"step": 10
- },
- {
- "expr": "sum(label_replace(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "H",
- "step": 10
}
],
"thresholds": [
@@ -5124,7 +5226,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Quota",
+ "title": "CPU Quota",
"tooltip": {
"shared": false,
"sort": 0,
@@ -5165,172 +5267,9 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Memory Quota",
+ "title": "CPU Quota",
"titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
- "kubernetes-mixin"
- ],
- "templating": {
- "list": [
- {
- "current": {
- "text": "Prometheus",
- "value": "Prometheus"
- },
- "hide": 0,
- "label": null,
- "name": "datasource",
- "options": [
-
- ],
- "query": "prometheus",
- "refresh": 1,
- "regex": "",
- "type": "datasource"
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 2,
- "includeAll": false,
- "label": "cluster",
- "multi": false,
- "name": "cluster",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info, cluster)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "namespace",
- "multi": false,
- "name": "namespace",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "pod",
- "multi": false,
- "name": "pod",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info{cluster=\"$cluster\", namespace=\"$namespace\"}, pod)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "",
- "title": "Kubernetes / Compute Resources / Pod",
- "uid": "6581e46e4e5c7ba40a07646395ef7b23",
- "version": 0
- }
- kind: ConfigMap
- metadata:
- name: grafana-dashboard-k8s-resources-pod
- namespace: monitoring
-- apiVersion: v1
- data:
- k8s-resources-workload.json: |-
- {
- "annotations": {
- "list": [
-
- ]
- },
- "editable": true,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [
-
- ],
- "refresh": "10s",
- "rows": [
+ },
{
"collapse": false,
"height": "250px",
@@ -5344,7 +5283,7 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 10,
- "id": 1,
+ "id": 9,
"legend": {
"avg": false,
"current": false,
@@ -5373,10 +5312,10 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{pod}}",
+ "legendFormat": "{{namespace}}",
"legendLink": null,
"step": 10
}
@@ -5386,7 +5325,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Usage",
+ "title": "Memory Usage (w/o cache)",
"tooltip": {
"shared": false,
"sort": 0,
@@ -5404,7 +5343,7 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
@@ -5426,7 +5365,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "CPU Usage",
+ "title": "Memory",
"titleSize": "h6"
},
{
@@ -5442,7 +5381,7 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 2,
+ "id": 10,
"legend": {
"avg": false,
"current": false,
@@ -5477,7 +5416,43 @@ items:
"type": "hidden"
},
{
- "alias": "CPU Usage",
+ "alias": "Pods",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "link": true,
+ "linkTooltip": "Drill down to pods",
+ "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1",
+ "pattern": "Value #A",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Workloads",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "link": true,
+ "linkTooltip": "Drill down to workloads",
+ "linkUrl": "/d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1",
+ "pattern": "Value #B",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Memory Usage",
"colorMode": null,
"colors": [
@@ -5487,15 +5462,15 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "Value #A",
+ "pattern": "Value #C",
"thresholds": [
],
"type": "number",
- "unit": "short"
+ "unit": "bytes"
},
{
- "alias": "CPU Requests",
+ "alias": "Memory Requests",
"colorMode": null,
"colors": [
@@ -5505,15 +5480,15 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "Value #B",
+ "pattern": "Value #D",
"thresholds": [
],
"type": "number",
- "unit": "short"
+ "unit": "bytes"
},
{
- "alias": "CPU Requests %",
+ "alias": "Memory Requests %",
"colorMode": null,
"colors": [
@@ -5523,7 +5498,7 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "Value #C",
+ "pattern": "Value #E",
"thresholds": [
],
@@ -5531,7 +5506,7 @@ items:
"unit": "percentunit"
},
{
- "alias": "CPU Limits",
+ "alias": "Memory Limits",
"colorMode": null,
"colors": [
@@ -5541,15 +5516,15 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "Value #D",
+ "pattern": "Value #F",
"thresholds": [
],
"type": "number",
- "unit": "short"
+ "unit": "bytes"
},
{
- "alias": "CPU Limits %",
+ "alias": "Memory Limits %",
"colorMode": null,
"colors": [
@@ -5559,7 +5534,7 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "Value #E",
+ "pattern": "Value #G",
"thresholds": [
],
@@ -5567,7 +5542,7 @@ items:
"unit": "percentunit"
},
{
- "alias": "Pod",
+ "alias": "Namespace",
"colorMode": null,
"colors": [
@@ -5575,9 +5550,9 @@ items:
"dateFormat": "YYYY-MM-DD HH:mm:ss",
"decimals": 2,
"link": true,
- "linkTooltip": "Drill down",
- "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell",
- "pattern": "pod",
+ "linkTooltip": "Drill down to pods",
+ "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell",
+ "pattern": "namespace",
"thresholds": [
],
@@ -5602,7 +5577,7 @@ items:
],
"targets": [
{
- "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "count(mixin_pod_workload{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5611,7 +5586,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "count(avg(mixin_pod_workload{cluster=\"$cluster\"}) by (workload, namespace)) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5620,7 +5595,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5629,7 +5604,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5638,13 +5613,31 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
"legendFormat": "",
"refId": "E",
"step": 10
+ },
+ {
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "F",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "G",
+ "step": 10
}
],
"thresholds": [
@@ -5652,7 +5645,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Quota",
+ "title": "Requests by Namespace",
"tooltip": {
"shared": false,
"sort": 0,
@@ -5693,23 +5686,132 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "CPU Quota",
+ "title": "Memory Requests",
"titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {
-
- },
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
+
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(node_cpu_seconds_total, cluster)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Compute Resources / Cluster",
+ "uid": "efa86fd1d0c121a26444b636a3f509a8",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-k8s-resources-cluster
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ k8s-resources-namespace.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [
+
+ ],
+ "refresh": "10s",
+ "rows": [
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
"bars": false,
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
"fill": 10,
- "id": 3,
+ "id": 1,
"legend": {
"avg": false,
"current": false,
@@ -5738,7 +5840,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -5751,7 +5853,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage",
+ "title": "CPU Usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -5769,7 +5871,7 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -5791,7 +5893,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Memory Usage",
+ "title": "CPU Usage",
"titleSize": "h6"
},
{
@@ -5807,7 +5909,7 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 4,
+ "id": 2,
"legend": {
"avg": false,
"current": false,
@@ -5842,7 +5944,7 @@ items:
"type": "hidden"
},
{
- "alias": "Memory Usage",
+ "alias": "CPU Usage",
"colorMode": null,
"colors": [
@@ -5857,10 +5959,10 @@ items:
],
"type": "number",
- "unit": "bytes"
+ "unit": "short"
},
{
- "alias": "Memory Requests",
+ "alias": "CPU Requests",
"colorMode": null,
"colors": [
@@ -5875,10 +5977,10 @@ items:
],
"type": "number",
- "unit": "bytes"
+ "unit": "short"
},
{
- "alias": "Memory Requests %",
+ "alias": "CPU Requests %",
"colorMode": null,
"colors": [
@@ -5896,7 +5998,7 @@ items:
"unit": "percentunit"
},
{
- "alias": "Memory Limits",
+ "alias": "CPU Limits",
"colorMode": null,
"colors": [
@@ -5911,10 +6013,10 @@ items:
],
"type": "number",
- "unit": "bytes"
+ "unit": "short"
},
{
- "alias": "Memory Limits %",
+ "alias": "CPU Limits %",
"colorMode": null,
"colors": [
@@ -5967,7 +6069,7 @@ items:
],
"targets": [
{
- "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5976,7 +6078,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5985,7 +6087,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5994,7 +6096,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6003,7 +6105,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6017,7 +6119,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Quota",
+ "title": "CPU Quota",
"tooltip": {
"shared": false,
"sort": 0,
@@ -6058,233 +6160,43 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Memory Quota",
+ "title": "CPU Quota",
"titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
- "kubernetes-mixin"
- ],
- "templating": {
- "list": [
- {
- "current": {
- "text": "Prometheus",
- "value": "Prometheus"
- },
- "hide": 0,
- "label": null,
- "name": "datasource",
- "options": [
-
- ],
- "query": "prometheus",
- "refresh": 1,
- "regex": "",
- "type": "datasource"
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 2,
- "includeAll": false,
- "label": "cluster",
- "multi": false,
- "name": "cluster",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info, cluster)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "namespace",
- "multi": false,
- "name": "namespace",
- "options": [
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 10,
+ "id": 3,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [
- ],
- "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "workload",
- "multi": false,
- "name": "workload",
- "options": [
-
- ],
- "query": "label_values(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}, workload)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "type",
- "multi": false,
- "name": "type",
- "options": [
-
- ],
- "query": "label_values(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\"}, workload_type)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "",
- "title": "Kubernetes / Compute Resources / Workload",
- "uid": "a164a7f0339f99e89cea5cb47e9be617",
- "version": 0
- }
- kind: ConfigMap
- metadata:
- name: grafana-dashboard-k8s-resources-workload
- namespace: monitoring
-- apiVersion: v1
- data:
- k8s-resources-workloads-namespace.json: |-
- {
- "annotations": {
- "list": [
-
- ]
- },
- "editable": true,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [
-
- ],
- "refresh": "10s",
- "rows": [
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 10,
- "id": 1,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 0,
- "links": [
-
- ],
- "nullPointMode": "null as zero",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
],
"spaceLength": 10,
@@ -6293,10 +6205,10 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}) by (pod)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{workload}} - {{workload_type}}",
+ "legendFormat": "{{pod}}",
"legendLink": null,
"step": 10
}
@@ -6306,7 +6218,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Usage",
+ "title": "Memory Usage (w/o cache)",
"tooltip": {
"shared": false,
"sort": 0,
@@ -6324,7 +6236,7 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
@@ -6346,7 +6258,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "CPU Usage",
+ "title": "Memory Usage",
"titleSize": "h6"
},
{
@@ -6362,7 +6274,7 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 2,
+ "id": 4,
"legend": {
"avg": false,
"current": false,
@@ -6397,13 +6309,13 @@ items:
"type": "hidden"
},
{
- "alias": "Running Pods",
+ "alias": "Memory Usage",
"colorMode": null,
"colors": [
],
"dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 0,
+ "decimals": 2,
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
@@ -6412,10 +6324,10 @@ items:
],
"type": "number",
- "unit": "short"
+ "unit": "bytes"
},
{
- "alias": "CPU Usage",
+ "alias": "Memory Requests",
"colorMode": null,
"colors": [
@@ -6430,10 +6342,10 @@ items:
],
"type": "number",
- "unit": "short"
+ "unit": "bytes"
},
{
- "alias": "CPU Requests",
+ "alias": "Memory Requests %",
"colorMode": null,
"colors": [
@@ -6448,10 +6360,10 @@ items:
],
"type": "number",
- "unit": "short"
+ "unit": "percentunit"
},
{
- "alias": "CPU Requests %",
+ "alias": "Memory Limits",
"colorMode": null,
"colors": [
@@ -6466,10 +6378,10 @@ items:
],
"type": "number",
- "unit": "percentunit"
+ "unit": "bytes"
},
{
- "alias": "CPU Limits",
+ "alias": "Memory Limits %",
"colorMode": null,
"colors": [
@@ -6484,10 +6396,10 @@ items:
],
"type": "number",
- "unit": "short"
+ "unit": "percentunit"
},
{
- "alias": "CPU Limits %",
+ "alias": "Memory Usage (RSS)",
"colorMode": null,
"colors": [
@@ -6502,28 +6414,28 @@ items:
],
"type": "number",
- "unit": "percentunit"
+ "unit": "bytes"
},
{
- "alias": "Workload",
+ "alias": "Memory Usage (Cache)",
"colorMode": null,
"colors": [
],
"dateFormat": "YYYY-MM-DD HH:mm:ss",
"decimals": 2,
- "link": true,
+ "link": false,
"linkTooltip": "Drill down",
- "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2",
- "pattern": "workload",
+ "linkUrl": "",
+ "pattern": "Value #G",
"thresholds": [
],
"type": "number",
- "unit": "short"
+ "unit": "bytes"
},
{
- "alias": "Workload Type",
+ "alias": "Memory Usage (Swap",
"colorMode": null,
"colors": [
@@ -6533,7 +6445,25 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "workload_type",
+ "pattern": "Value #H",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Pod",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": true,
+ "linkTooltip": "Drill down",
+ "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell",
+ "pattern": "pod",
"thresholds": [
],
@@ -6558,7 +6488,7 @@ items:
],
"targets": [
{
- "expr": "count(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}) by (workload, workload_type)",
+ "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6567,7 +6497,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6576,7 +6506,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6585,7 +6515,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6594,7 +6524,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6603,13 +6533,31 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
"legendFormat": "",
"refId": "F",
"step": 10
+ },
+ {
+ "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "G",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "H",
+ "step": 10
}
],
"thresholds": [
@@ -6617,7 +6565,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Quota",
+ "title": "Memory Quota",
"tooltip": {
"shared": false,
"sort": 0,
@@ -6658,23 +6606,159 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "CPU Quota",
+ "title": "Memory Quota",
"titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 10,
- "id": 3,
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
+
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "namespace",
+ "multi": false,
+ "name": "namespace",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Compute Resources / Namespace (Pods)",
+ "uid": "85a562078cdf77779eaa1add43ccec1e",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-k8s-resources-namespace
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ k8s-resources-pod.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [
+
+ ],
+ "refresh": "10s",
+ "rows": [
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 10,
+ "id": 1,
"legend": {
"avg": false,
"current": false,
@@ -6703,10 +6787,10 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", cluster=\"$cluster\"}) by (container)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{workload}} - {{workload_type}}",
+ "legendFormat": "{{container}}",
"legendLink": null,
"step": 10
}
@@ -6716,7 +6800,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage",
+ "title": "CPU Usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -6734,7 +6818,7 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -6756,7 +6840,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Memory Usage",
+ "title": "CPU Usage",
"titleSize": "h6"
},
{
@@ -6772,7 +6856,7 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "id": 4,
+ "id": 2,
"legend": {
"avg": false,
"current": false,
@@ -6807,13 +6891,13 @@ items:
"type": "hidden"
},
{
- "alias": "Running Pods",
+ "alias": "CPU Usage",
"colorMode": null,
"colors": [
],
"dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 0,
+ "decimals": 2,
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
@@ -6825,7 +6909,7 @@ items:
"unit": "short"
},
{
- "alias": "Memory Usage",
+ "alias": "CPU Requests",
"colorMode": null,
"colors": [
@@ -6840,10 +6924,10 @@ items:
],
"type": "number",
- "unit": "bytes"
+ "unit": "short"
},
{
- "alias": "Memory Requests",
+ "alias": "CPU Requests %",
"colorMode": null,
"colors": [
@@ -6858,10 +6942,10 @@ items:
],
"type": "number",
- "unit": "bytes"
+ "unit": "percentunit"
},
{
- "alias": "Memory Requests %",
+ "alias": "CPU Limits",
"colorMode": null,
"colors": [
@@ -6876,10 +6960,10 @@ items:
],
"type": "number",
- "unit": "percentunit"
+ "unit": "short"
},
{
- "alias": "Memory Limits",
+ "alias": "CPU Limits %",
"colorMode": null,
"colors": [
@@ -6892,48 +6976,12 @@ items:
"pattern": "Value #E",
"thresholds": [
- ],
- "type": "number",
- "unit": "bytes"
- },
- {
- "alias": "Memory Limits %",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": false,
- "linkTooltip": "Drill down",
- "linkUrl": "",
- "pattern": "Value #F",
- "thresholds": [
-
],
"type": "number",
"unit": "percentunit"
},
{
- "alias": "Workload",
- "colorMode": null,
- "colors": [
-
- ],
- "dateFormat": "YYYY-MM-DD HH:mm:ss",
- "decimals": 2,
- "link": true,
- "linkTooltip": "Drill down",
- "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2",
- "pattern": "workload",
- "thresholds": [
-
- ],
- "type": "number",
- "unit": "short"
- },
- {
- "alias": "Workload Type",
+ "alias": "Container",
"colorMode": null,
"colors": [
@@ -6943,7 +6991,7 @@ items:
"link": false,
"linkTooltip": "Drill down",
"linkUrl": "",
- "pattern": "workload_type",
+ "pattern": "container",
"thresholds": [
],
@@ -6968,7 +7016,7 @@ items:
],
"targets": [
{
- "expr": "count(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}) by (workload, workload_type)",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6977,7 +7025,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n",
+ "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6986,7 +7034,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6995,7 +7043,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -7004,22 +7052,13 @@ items:
"step": 10
},
{
- "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
"legendFormat": "",
"refId": "E",
"step": 10
- },
- {
- "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
- "format": "table",
- "instant": true,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "F",
- "step": 10
}
],
"thresholds": [
@@ -7027,7 +7066,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Quota",
+ "title": "CPU Quota",
"tooltip": {
"shared": false,
"sort": 0,
@@ -7068,155 +7107,12 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Memory Quota",
+ "title": "CPU Quota",
"titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
- "kubernetes-mixin"
- ],
- "templating": {
- "list": [
- {
- "current": {
- "text": "Prometheus",
- "value": "Prometheus"
- },
- "hide": 0,
- "label": null,
- "name": "datasource",
- "options": [
-
- ],
- "query": "prometheus",
- "refresh": 1,
- "regex": "",
- "type": "datasource"
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 2,
- "includeAll": false,
- "label": "cluster",
- "multi": false,
- "name": "cluster",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info, cluster)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "namespace",
- "multi": false,
- "name": "namespace",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 2,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "",
- "title": "Kubernetes / Compute Resources / Namespace (Workloads)",
- "uid": "a87fb0d919ec0ea5f6543124e16c42a5",
- "version": 0
- }
- kind: ConfigMap
- metadata:
- name: grafana-dashboard-k8s-resources-workloads-namespace
- namespace: monitoring
-- apiVersion: v1
- data:
- nodes.json: |-
- {
- "__inputs": [
-
- ],
- "__requires": [
-
- ],
- "annotations": {
- "list": [
-
- ]
- },
- "editable": false,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "id": null,
- "links": [
-
- ],
- "refresh": "",
- "rows": [
+ },
{
"collapse": false,
- "collapsed": false,
+ "height": "250px",
"panels": [
{
"aliasColors": {
@@ -7226,68 +7122,58 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
- "gridPos": {
-
- },
- "id": 2,
+ "fill": 10,
+ "id": 3,
"legend": {
- "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
- "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
- "stack": false,
+ "span": 12,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "max(node_load1{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "load 1m",
- "refId": "A"
- },
- {
- "expr": "max(node_load5{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\"}) by (container)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "load 5m",
- "refId": "B"
+ "legendFormat": "{{container}} (RSS)",
+ "legendLink": null,
+ "step": 10
},
{
- "expr": "max(node_load15{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\"}) by (container)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "load 15m",
- "refId": "C"
+ "legendFormat": "{{container}} (Cache)",
+ "legendLink": null,
+ "step": 10
},
{
- "expr": "count(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", mode=\"user\"})",
+ "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\"}) by (container)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "logical cores",
- "refId": "D"
+ "legendFormat": "{{container}} (Swap)",
+ "legendLink": null,
+ "step": 10
}
],
"thresholds": [
@@ -7295,7 +7181,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "System load",
+ "title": "Memory Usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -7313,11 +7199,11 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
@@ -7326,10 +7212,22 @@ items:
"logBase": 1,
"max": null,
"min": null,
- "show": true
+ "show": false
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Memory Usage",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
{
"aliasColors": {
@@ -7339,17 +7237,12 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "gridPos": {
-
- },
- "id": 3,
+ "id": 4,
"legend": {
- "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
- "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -7359,339 +7252,275 @@ items:
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 12,
"stack": false,
"steppedLine": false,
- "targets": [
+ "styles": [
{
- "expr": "sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{cpu}}",
- "refId": "A"
- }
- ],
- "thresholds": [
+ "alias": "Time",
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "pattern": "Time",
+ "type": "hidden"
+ },
+ {
+ "alias": "Memory Usage",
+ "colorMode": null,
+ "colors": [
- ],
- "timeFrom": null,
- "timeShift": null,
- "title": "Usage Per Core",
- "tooltip": {
- "shared": false,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": [
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #A",
+ "thresholds": [
- ]
- },
- "yaxes": [
- {
- "format": "percentunit",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
+ ],
+ "type": "number",
+ "unit": "bytes"
},
{
- "format": "percentunit",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "panels": [
- {
- "aliasColors": {
+ "alias": "Memory Requests",
+ "colorMode": null,
+ "colors": [
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 1,
- "gridPos": {
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #B",
+ "thresholds": [
- },
- "id": 4,
- "legend": {
- "alignAsTable": "true",
- "avg": "true",
- "current": "true",
- "max": "false",
- "min": "false",
- "rightSide": "true",
- "show": "true",
- "total": "false",
- "values": "true"
- },
- "lines": true,
- "linewidth": 1,
- "links": [
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Memory Requests %",
+ "colorMode": null,
+ "colors": [
- ],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "seriesOverrides": [
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #C",
+ "thresholds": [
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
{
- "expr": "max (sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m])) ) * 100\n",
- "format": "time_series",
- "intervalFactor": 10,
- "legendFormat": "{{ cpu }}",
- "refId": "A"
- }
- ],
- "thresholds": [
+ "alias": "Memory Limits",
+ "colorMode": null,
+ "colors": [
- ],
- "timeFrom": null,
- "timeShift": null,
- "title": "CPU Utilization",
- "tooltip": {
- "shared": false,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": [
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #D",
+ "thresholds": [
- ]
- },
- "yaxes": [
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
{
- "format": "percent",
- "label": null,
- "logBase": 1,
- "max": 100,
- "min": 0,
- "show": true
+ "alias": "Memory Limits %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #E",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
},
{
- "format": "percent",
- "label": null,
- "logBase": 1,
- "max": 100,
- "min": 0,
- "show": true
- }
- ]
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "$datasource",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
+ "alias": "Memory Usage (RSS)",
+ "colorMode": null,
+ "colors": [
- },
- "id": 5,
- "interval": null,
- "links": [
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #F",
+ "thresholds": [
- ],
- "mappingType": 1,
- "mappingTypes": [
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
{
- "name": "value to text",
- "value": 1
+ "alias": "Memory Usage (Cache)",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #G",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
},
{
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
+ "alias": "Memory Usage (Swap",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #H",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
{
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "avg(sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m]))) * 100\n",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A"
- }
- ],
- "thresholds": "80, 90",
- "title": "CPU Usage",
- "tooltip": {
- "shared": false
- },
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "panels": [
- {
- "aliasColors": {
+ "alias": "Container",
+ "colorMode": null,
+ "colors": [
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 1,
- "gridPos": {
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "container",
+ "thresholds": [
- },
- "id": 6,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
- ],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "seriesOverrides": [
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "/.*/",
+ "thresholds": [
+ ],
+ "type": "string",
+ "unit": "short"
+ }
],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
"targets": [
{
- "expr": "max(\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n",
- "format": "time_series",
+ "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\"}) by (container)",
+ "format": "table",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "memory used",
- "refId": "A"
+ "legendFormat": "",
+ "refId": "A",
+ "step": 10
},
{
- "expr": "max(node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
- "format": "time_series",
+ "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
+ "format": "table",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "memory buffers",
- "refId": "B"
+ "legendFormat": "",
+ "refId": "B",
+ "step": 10
},
{
- "expr": "max(node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
- "format": "time_series",
+ "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
+ "format": "table",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "memory cached",
- "refId": "C"
+ "legendFormat": "",
+ "refId": "C",
+ "step": 10
},
{
- "expr": "max(node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
- "format": "time_series",
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\"}) by (container)",
+ "format": "table",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "memory free",
- "refId": "D"
+ "legendFormat": "",
+ "refId": "D",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\"}) by (container) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "E",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "F",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "G",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "H",
+ "step": 10
}
],
"thresholds": [
@@ -7699,13 +7528,14 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage",
+ "title": "Memory Quota",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "graph",
+ "transform": "table",
+ "type": "table",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -7717,119 +7547,197 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": true
+ "show": false
}
]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Memory Quota",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
},
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "$datasource",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
-
- },
- "id": 7,
- "interval": null,
- "links": [
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
- ],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "max(\n (\n (\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A"
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory Usage",
- "tooltip": {
- "shared": false
- },
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "namespace",
+ "multi": false,
+ "name": "namespace",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "pod",
+ "multi": false,
+ "name": "pod",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{cluster=\"$cluster\", namespace=\"$namespace\"}, pod)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Compute Resources / Pod",
+ "uid": "6581e46e4e5c7ba40a07646395ef7b23",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-k8s-resources-pod
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ k8s-resources-workload.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [
+
+ ],
+ "refresh": "10s",
+ "rows": [
{
"collapse": false,
- "collapsed": false,
+ "height": "250px",
"panels": [
{
"aliasColors": {
@@ -7839,68 +7747,42 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
- "gridPos": {
-
- },
- "id": 8,
+ "fill": 10,
+ "id": 1,
"legend": {
- "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
- "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
- {
- "alias": "read",
- "yaxis": 1
- },
- {
- "alias": "io time",
- "yaxis": 2
- }
+
],
"spaceLength": 10,
- "span": 6,
- "stack": false,
+ "span": 12,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "max(rate(node_disk_read_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "read",
- "refId": "A"
- },
- {
- "expr": "max(rate(node_disk_written_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "written",
- "refId": "B"
- },
- {
- "expr": "max(rate(node_disk_io_time_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))",
+ "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "io time",
- "refId": "C"
+ "legendFormat": "{{pod}}",
+ "legendLink": null,
+ "step": 10
}
],
"thresholds": [
@@ -7908,7 +7790,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk I/O",
+ "title": "CPU Usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -7926,23 +7808,35 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
- "format": "ms",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": true
+ "show": false
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "CPU Usage",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
{
"aliasColors": {
@@ -7952,17 +7846,12 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "gridPos": {
-
- },
- "id": 9,
+ "id": 2,
"legend": {
- "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
- "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -7972,33 +7861,194 @@ items:
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 12,
"stack": false,
"steppedLine": false,
+ "styles": [
+ {
+ "alias": "Time",
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "pattern": "Time",
+ "type": "hidden"
+ },
+ {
+ "alias": "CPU Usage",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #A",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Requests",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #B",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Requests %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #C",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "CPU Limits",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #D",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Limits %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #E",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "Pod",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": true,
+ "linkTooltip": "Drill down",
+ "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell",
+ "pattern": "pod",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "/.*/",
+ "thresholds": [
+
+ ],
+ "type": "string",
+ "unit": "short"
+ }
+ ],
"targets": [
{
- "expr": "max by (namespace, pod, device) ((node_filesystem_size_bytes{cluster=\"$cluster\", fstype=~\"ext[234]|btrfs|xfs|zfs\", instance=\"$instance\", job=\"node-exporter\"} - node_filesystem_avail_bytes{cluster=\"$cluster\", fstype=~\"ext[234]|btrfs|xfs|zfs\", instance=\"$instance\", job=\"node-exporter\"}) / node_filesystem_size_bytes{cluster=\"$cluster\", fstype=~\"ext[234]|btrfs|xfs|zfs\", instance=\"$instance\", job=\"node-exporter\"})",
- "format": "time_series",
+ "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "disk used",
- "refId": "A"
+ "legendFormat": "",
+ "refId": "A",
+ "step": 10
},
{
- "expr": "max by (namespace, pod, device) (node_filesystem_avail_bytes{cluster=\"$cluster\", fstype=~\"ext[234]|btrfs|xfs|zfs\", instance=\"$instance\", job=\"node-exporter\"} / node_filesystem_size_bytes{cluster=\"$cluster\", fstype=~\"ext[234]|btrfs|xfs|zfs\", instance=\"$instance\", job=\"node-exporter\"})",
- "format": "time_series",
+ "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "disk free",
- "refId": "B"
+ "legendFormat": "",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "C",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "D",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "E",
+ "step": 10
}
],
"thresholds": [
@@ -8006,13 +8056,14 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk Space Usage",
+ "title": "CPU Quota",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "graph",
+ "transform": "table",
+ "type": "table",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -8024,20 +8075,20 @@ items:
},
"yaxes": [
{
- "format": "percentunit",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
- "format": "percentunit",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": true
+ "show": false
}
]
}
@@ -8045,14 +8096,13 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
+ "showTitle": true,
+ "title": "CPU Quota",
+ "titleSize": "h6"
},
{
"collapse": false,
- "collapsed": false,
+ "height": "250px",
"panels": [
{
"aliasColors": {
@@ -8062,47 +8112,42 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
- "gridPos": {
-
- },
- "id": 10,
+ "fill": 10,
+ "id": 3,
"legend": {
- "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
- "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
- "stack": false,
+ "span": 12,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "max(rate(node_network_receive_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m]))",
+ "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{device}}",
- "refId": "A"
+ "legendFormat": "{{pod}}",
+ "legendLink": null,
+ "step": 10
}
],
"thresholds": [
@@ -8110,7 +8155,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Network Received",
+ "title": "Memory Usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -8132,39 +8177,46 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": true
+ "show": false
}
]
- },
- {
- "aliasColors": {
-
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Memory Usage",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
"bars": false,
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "gridPos": {
-
- },
- "id": 11,
+ "id": 4,
"legend": {
- "alignAsTable": false,
"avg": false,
"current": false,
"max": false,
"min": false,
- "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -8174,26 +8226,194 @@ items:
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 6,
+ "span": 12,
"stack": false,
"steppedLine": false,
+ "styles": [
+ {
+ "alias": "Time",
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "pattern": "Time",
+ "type": "hidden"
+ },
+ {
+ "alias": "Memory Usage",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #A",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Memory Requests",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #B",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Memory Requests %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #C",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "Memory Limits",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #D",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Memory Limits %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #E",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "Pod",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": true,
+ "linkTooltip": "Drill down",
+ "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell",
+ "pattern": "pod",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "/.*/",
+ "thresholds": [
+
+ ],
+ "type": "string",
+ "unit": "short"
+ }
+ ],
"targets": [
{
- "expr": "max(rate(node_network_transmit_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m]))",
- "format": "time_series",
+ "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
"intervalFactor": 2,
- "legendFormat": "{{device}}",
- "refId": "A"
+ "legendFormat": "",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "C",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "D",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "E",
+ "step": 10
}
],
"thresholds": [
@@ -8201,13 +8421,14 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Network Transmitted",
+ "title": "Memory Quota",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "graph",
+ "transform": "table",
+ "type": "table",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -8219,20 +8440,20 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
"min": null,
- "show": true
+ "show": false
}
]
}
@@ -8240,71 +8461,248 @@ items:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "panels": [
- {
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 1,
- "gridPos": {
+ "showTitle": true,
+ "title": "Memory Quota",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
- },
- "id": 12,
- "legend": {
- "alignAsTable": false,
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "namespace",
+ "multi": false,
+ "name": "namespace",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "workload",
+ "multi": false,
+ "name": "workload",
+ "options": [
+
+ ],
+ "query": "label_values(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}, workload)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "prod",
+ "value": "prod"
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "type",
+ "multi": false,
+ "name": "type",
+ "options": [
+
+ ],
+ "query": "label_values(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\"}, workload_type)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Compute Resources / Workload",
+ "uid": "a164a7f0339f99e89cea5cb47e9be617",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-k8s-resources-workload
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ k8s-resources-workloads-namespace.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [
+
+ ],
+ "refresh": "10s",
+ "rows": [
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 10,
+ "id": 1,
+ "legend": {
"avg": false,
"current": false,
"max": false,
"min": false,
- "rightSide": false,
"show": true,
"total": false,
"values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 9,
- "stack": false,
+ "span": 12,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "max(\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "inodes used",
- "refId": "A"
- },
- {
- "expr": "max(node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "inodes free",
- "refId": "B"
+ "legendFormat": "{{workload}} - {{workload_type}}",
+ "legendLink": null,
+ "step": 10
}
],
"thresholds": [
@@ -8312,7 +8710,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Inodes Usage",
+ "title": "CPU Usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -8334,7 +8732,7 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
@@ -8343,250 +8741,24 @@ items:
"logBase": 1,
"max": null,
"min": null,
- "show": true
+ "show": false
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "CPU Usage",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
{
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "$datasource",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
-
- },
- "id": 13,
- "interval": null,
- "links": [
-
- ],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "max(\n (\n (\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A"
- }
- ],
- "thresholds": "80, 90",
- "title": "Inodes Usage",
- "tooltip": {
- "shared": false
- },
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
- "kubernetes-mixin"
- ],
- "templating": {
- "list": [
- {
- "current": {
- "text": "Prometheus",
- "value": "Prometheus"
- },
- "hide": 0,
- "label": null,
- "name": "datasource",
- "options": [
-
- ],
- "query": "prometheus",
- "refresh": 1,
- "regex": "",
- "type": "datasource"
- },
- {
- "allValue": null,
- "current": {
-
- },
- "datasource": "$datasource",
- "hide": 2,
- "includeAll": false,
- "label": "cluster",
- "multi": false,
- "name": "cluster",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info, cluster)",
- "refresh": 2,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
-
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": null,
- "multi": false,
- "name": "instance",
- "options": [
-
- ],
- "query": "label_values(node_boot_time_seconds{cluster=\"$cluster\", job=\"node-exporter\"}, instance)",
- "refresh": 2,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "",
- "title": "Kubernetes / Nodes",
- "uid": "fa49a4706d07a042595b664c87fb33ea",
- "version": 0
- }
- kind: ConfigMap
- metadata:
- name: grafana-dashboard-nodes
- namespace: monitoring
-- apiVersion: v1
- data:
- persistentvolumesusage.json: |-
- {
- "__inputs": [
-
- ],
- "__requires": [
-
- ],
- "annotations": {
- "list": [
-
- ]
- },
- "editable": false,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "id": null,
- "links": [
-
- ],
- "refresh": "",
- "rows": [
- {
- "collapse": false,
- "collapsed": false,
- "panels": [
- {
- "aliasColors": {
+ "aliasColors": {
},
"bars": false,
@@ -8594,67 +8766,269 @@ items:
"dashes": false,
"datasource": "$datasource",
"fill": 1,
- "gridPos": {
-
- },
"id": 2,
"legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": true,
- "min": true,
- "rightSide": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
"show": true,
"total": false,
- "values": true
+ "values": false
},
"lines": true,
"linewidth": 1,
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 9,
- "stack": true,
+ "span": 12,
+ "stack": false,
"steppedLine": false,
- "targets": [
+ "styles": [
{
- "expr": "(\n sum without(instance, node) (kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n",
- "format": "time_series",
- "intervalFactor": 1,
- "legendFormat": "Used Space",
- "refId": "A"
+ "alias": "Time",
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "pattern": "Time",
+ "type": "hidden"
},
{
- "expr": "sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n",
- "format": "time_series",
- "intervalFactor": 1,
- "legendFormat": "Free Space",
- "refId": "B"
- }
- ],
- "thresholds": [
+ "alias": "Running Pods",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #A",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Usage",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #B",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Requests",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #C",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Requests %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #D",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "CPU Limits",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #E",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "CPU Limits %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #F",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "Workload",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": true,
+ "linkTooltip": "Drill down",
+ "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2",
+ "pattern": "workload",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Workload Type",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "workload_type",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "/.*/",
+ "thresholds": [
+
+ ],
+ "type": "string",
+ "unit": "short"
+ }
+ ],
+ "targets": [
+ {
+ "expr": "count(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}) by (workload, workload_type)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "C",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "D",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "E",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "F",
+ "step": 10
+ }
+ ],
+ "thresholds": [
],
"timeFrom": null,
"timeShift": null,
- "title": "Volume Space Usage",
+ "title": "CPU Quota",
"tooltip": {
"shared": false,
"sort": 0,
"value_type": "individual"
},
- "type": "graph",
+ "transform": "table",
+ "type": "table",
"xaxis": {
"buckets": null,
"mode": "time",
@@ -8666,7 +9040,7 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -8674,168 +9048,71 @@ items:
"show": true
},
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
- "show": true
+ "min": null,
+ "show": false
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "CPU Quota",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
{
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "$datasource",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
-
- },
- "id": 3,
- "interval": null,
- "links": [
-
- ],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "(\n kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n -\n kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n)\n/\nkubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100\n",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A"
- }
- ],
- "thresholds": "80, 90",
- "title": "Volume Space Usage",
- "tooltip": {
- "shared": false
- },
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "panels": [
- {
- "aliasColors": {
-
+ "aliasColors": {
+
},
"bars": false,
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
- "gridPos": {
-
- },
- "id": 4,
+ "fill": 10,
+ "id": 3,
"legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": true,
- "min": true,
- "rightSide": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
"show": true,
"total": false,
- "values": true
+ "values": false
},
"lines": true,
- "linewidth": 1,
+ "linewidth": 0,
"links": [
],
- "nullPointMode": "null",
+ "nullPointMode": "null as zero",
"percentage": false,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "repeat": null,
"seriesOverrides": [
],
"spaceLength": 10,
- "span": 9,
+ "span": 12,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum without(instance, node) (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n",
- "format": "time_series",
- "intervalFactor": 1,
- "legendFormat": "Used inodes",
- "refId": "A"
- },
- {
- "expr": "(\n sum without(instance, node) (kubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n",
+ "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
"format": "time_series",
- "intervalFactor": 1,
- "legendFormat": " Free inodes",
- "refId": "B"
+ "intervalFactor": 2,
+ "legendFormat": "{{workload}} - {{workload_type}}",
+ "legendLink": null,
+ "step": 10
}
],
"thresholds": [
@@ -8843,7 +9120,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Volume inodes Usage",
+ "title": "Memory Usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -8861,7 +9138,7 @@ items:
},
"yaxes": [
{
- "format": "none",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
@@ -8869,107 +9146,334 @@ items:
"show": true
},
{
- "format": "none",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
- "show": true
+ "min": null,
+ "show": false
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Memory Usage",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
{
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
"datasource": "$datasource",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
+ "fill": 1,
+ "id": 4,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
"show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
-
+ "total": false,
+ "values": false
},
- "id": 5,
- "interval": null,
+ "lines": true,
+ "linewidth": 1,
"links": [
],
- "mappingType": 1,
- "mappingTypes": [
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "styles": [
{
- "name": "value to text",
- "value": 1
+ "alias": "Time",
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "pattern": "Time",
+ "type": "hidden"
},
{
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
+ "alias": "Running Pods",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #A",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
{
- "expr": "kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n/\nkubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100\n",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A"
- }
- ],
- "thresholds": "80, 90",
- "title": "Volume inodes Usage",
+ "alias": "Memory Usage",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #B",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Memory Requests",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #C",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Memory Requests %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #D",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "Memory Limits",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #E",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Memory Limits %",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "Value #F",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "Workload",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": true,
+ "linkTooltip": "Drill down",
+ "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2",
+ "pattern": "workload",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Workload Type",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "pattern": "workload_type",
+ "thresholds": [
+
+ ],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "/.*/",
+ "thresholds": [
+
+ ],
+ "type": "string",
+ "unit": "short"
+ }
+ ],
+ "targets": [
+ {
+ "expr": "count(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}) by (workload, workload_type)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "C",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "D",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "E",
+ "step": 10
+ },
+ {
+ "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "F",
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Quota",
"tooltip": {
- "shared": false
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
},
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
+ "transform": "table",
+ "type": "table",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
{
- "op": "=",
- "text": "N/A",
- "value": "null"
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
}
- ],
- "valueName": "current"
+ ]
}
],
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
+ "showTitle": true,
+ "title": "Memory Quota",
+ "titleSize": "h6"
}
],
"schemaVersion": 14,
@@ -8998,7 +9502,8 @@ items:
{
"allValue": null,
"current": {
-
+ "text": "prod",
+ "value": "prod"
},
"datasource": "$datasource",
"hide": 2,
@@ -9009,10 +9514,10 @@ items:
"options": [
],
- "query": "label_values(kubelet_volume_stats_capacity_bytes, cluster)",
- "refresh": 2,
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 1,
"regex": "",
- "sort": 0,
+ "sort": 2,
"tagValuesQuery": "",
"tags": [
@@ -9024,47 +9529,22 @@ items:
{
"allValue": null,
"current": {
-
+ "text": "prod",
+ "value": "prod"
},
"datasource": "$datasource",
"hide": 0,
"includeAll": false,
- "label": "Namespace",
+ "label": "namespace",
"multi": false,
"name": "namespace",
"options": [
],
- "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\"}, namespace)",
- "refresh": 2,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
-
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "PersistentVolumeClaim",
- "multi": false,
- "name": "volume",
- "options": [
-
- ],
- "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\"}, persistentvolumeclaim)",
- "refresh": 2,
+ "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
+ "refresh": 1,
"regex": "",
- "sort": 0,
+ "sort": 2,
"tagValuesQuery": "",
"tags": [
@@ -9076,7 +9556,7 @@ items:
]
},
"time": {
- "from": "now-7d",
+ "from": "now-1h",
"to": "now"
},
"timepicker": {
@@ -9105,17 +9585,17 @@ items:
]
},
"timezone": "",
- "title": "Kubernetes / Persistent Volumes",
- "uid": "919b92a8e8041bd567af9edab12c840c",
+ "title": "Kubernetes / Compute Resources / Namespace (Workloads)",
+ "uid": "a87fb0d919ec0ea5f6543124e16c42a5",
"version": 0
}
kind: ConfigMap
metadata:
- name: grafana-dashboard-persistentvolumesusage
+ name: grafana-dashboard-k8s-resources-workloads-namespace
namespace: monitoring
- apiVersion: v1
data:
- pods.json: |-
+ kubelet.json: |-
{
"__inputs": [
@@ -9125,20 +9605,7 @@ items:
],
"annotations": {
"list": [
- {
- "builtIn": 1,
- "datasource": "$datasource",
- "enable": true,
- "expr": "time() == BOOL timestamp(rate(kube_pod_container_status_restarts_total{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[2m]) > 0)",
- "hide": false,
- "iconColor": "rgba(215, 44, 44, 1)",
- "name": "Restarts",
- "showIn": 0,
- "tags": [
- "restart"
- ],
- "type": "rows"
- }
+
]
},
"editable": false,
@@ -9148,9 +9615,6784 @@ items:
"id": null,
"links": [
- ],
- "refresh": "",
- "rows": [
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(up{job=\"kubelet\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Up",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 3,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kubelet_running_pod_count{job=\"kubelet\", instance=~\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Running Pods",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 4,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kubelet_running_container_count{job=\"kubelet\", instance=~\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Running Container",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(volume_manager_total_volumes{job=\"kubelet\", instance=~\"$instance\", state=\"actual_state_of_world\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Actual Volume Count",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 6,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(volume_manager_total_volumes{job=\"kubelet\", instance=~\"$instance\",state=\"desired_state_of_world\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Desired Volume Count",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(rate(kubelet_node_config_error{job=\"kubelet\", instance=~\"$instance\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Config Error Count",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 8,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(kubelet_runtime_operations_total{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (operation_type, instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} {{operation_type}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Operation Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(kubelet_runtime_operations_errors_total{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_type)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} {{operation_type}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Operation Error Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 10,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_runtime_operations_duration_seconds_bucket{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_type, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} {{operation_type}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Operation duration 99th quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 11,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(kubelet_pod_start_duration_seconds_count{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} pod",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(kubelet_pod_worker_duration_seconds_count{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} worker",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Pod Start Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 12,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_start_duration_seconds_count{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} pod",
+ "refId": "A"
+ },
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_worker_duration_seconds_bucket{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} worker",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Pod Start Duration",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 13,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "hideEmpty": "true",
+ "hideZero": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(storage_operation_duration_seconds_count{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Storage Operation Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 14,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "hideEmpty": "true",
+ "hideZero": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(storage_operation_errors_total{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Storage Operation Error Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 15,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "hideEmpty": "true",
+ "hideZero": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": true,
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(storage_operation_duration_seconds_bucket{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Storage Operation Duration 99th quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 16,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(kubelet_cgroup_manager_duration_seconds_count{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_type)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{operation_type}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Cgroup manager operation rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 17,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_cgroup_manager_duration_seconds_bucket{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_type, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} {{operation_type}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Cgroup manager 99th quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "Pod lifecycle event generator",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 18,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(kubelet_pleg_relist_duration_seconds_count{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "PLEG relist rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 19,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_interval_seconds_bucket{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "PLEG relist interval",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 20,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_duration_seconds_bucket{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "PLEG relist duration",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 21,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "2xx",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "3xx",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "4xx",
+ "refId": "C"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "5xx",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "RPC Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 22,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, verb, url, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} {{verb}} {{url}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Request duration 99th quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 23,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "process_resident_memory_bytes{job=\"kubelet\",instance=~\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 24,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(process_cpu_seconds_total{job=\"kubelet\",instance=~\"$instance\"}[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 25,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "go_goroutines{job=\"kubelet\",instance=~\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Goroutines",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
+
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "instance",
+ "options": [
+
+ ],
+ "query": "label_values(kubelet_runtime_operations{job=\"kubelet\"}, instance)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Kubelet",
+ "uid": "3138fa155d5915769fbded898ac09fd9",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-kubelet
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ nodes.json: |-
+ {
+ "__inputs": [
+
+ ],
+ "__requires": [
+
+ ],
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max(node_load1{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 1m",
+ "refId": "A"
+ },
+ {
+ "expr": "max(node_load5{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 5m",
+ "refId": "B"
+ },
+ {
+ "expr": "max(node_load15{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 15m",
+ "refId": "C"
+ },
+ {
+ "expr": "count(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", mode=\"user\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "logical cores",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "System load",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{cpu}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Usage Per Core",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": "true",
+ "current": "true",
+ "max": "false",
+ "min": "false",
+ "rightSide": "true",
+ "show": "true",
+ "total": "false",
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max (sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m])) ) * 100\n",
+ "format": "time_series",
+ "intervalFactor": 10,
+ "legendFormat": "{{ cpu }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Utilization",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": 100,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": 100,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "$datasource",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "avg(sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m]))) * 100\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "CPU Usage",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max(\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory used",
+ "refId": "A"
+ },
+ {
+ "expr": "max(node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory buffers",
+ "refId": "B"
+ },
+ {
+ "expr": "max(node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory cached",
+ "refId": "C"
+ },
+ {
+ "expr": "max(node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory free",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "$datasource",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(\n (\n (\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory Usage",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 8,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "read",
+ "yaxis": 1
+ },
+ {
+ "alias": "io time",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max(rate(node_disk_read_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "read",
+ "refId": "A"
+ },
+ {
+ "expr": "max(rate(node_disk_written_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "written",
+ "refId": "B"
+ },
+ {
+ "expr": "max(rate(node_disk_io_time_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "io time",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk I/O",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "ms",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node:node_filesystem_usage:{cluster=\"$cluster\", instance=\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}} disk used",
+ "refId": "A"
+ },
+ {
+ "expr": "node:node_filesystem_usage:{cluster=\"$cluster\", instance=\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}} disk free",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk Space Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 10,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_receive_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Received",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 11,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_transmit_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Transmitted",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 12,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max(\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "inodes used",
+ "refId": "A"
+ },
+ {
+ "expr": "max(node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "inodes free",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Inodes Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "$datasource",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 13,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(\n (\n (\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Inodes Usage",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
+
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": null,
+ "multi": false,
+ "name": "instance",
+ "options": [
+
+ ],
+ "query": "label_values(node_boot_time_seconds{cluster=\"$cluster\", job=\"node-exporter\"}, instance)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Nodes",
+ "uid": "fa49a4706d07a042595b664c87fb33ea",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-nodes
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ persistentvolumesusage.json: |-
+ {
+ "__inputs": [
+
+ ],
+ "__requires": [
+
+ ],
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": true,
+ "min": true,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "(\n sum without(instance, node) (kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Used Space",
+ "refId": "A"
+ },
+ {
+ "expr": "sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Free Space",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Volume Space Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "$datasource",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 3,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(\n kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n -\n kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n)\n/\nkubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Volume Space Usage",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": true,
+ "min": true,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum without(instance, node) (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Used inodes",
+ "refId": "A"
+ },
+ {
+ "expr": "(\n sum without(instance, node) (kubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": " Free inodes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Volume inodes Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "none",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "$datasource",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n/\nkubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100\n",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Volume inodes Usage",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
+
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kubelet_volume_stats_capacity_bytes, cluster)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Namespace",
+ "multi": false,
+ "name": "namespace",
+ "options": [
+
+ ],
+ "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\"}, namespace)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "PersistentVolumeClaim",
+ "multi": false,
+ "name": "volume",
+ "options": [
+
+ ],
+ "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\"}, persistentvolumeclaim)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-7d",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Persistent Volumes",
+ "uid": "919b92a8e8041bd567af9edab12c840c",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-persistentvolumesusage
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ pods.json: |-
+ {
+ "__inputs": [
+
+ ],
+ "__requires": [
+
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "$datasource",
+ "enable": true,
+ "expr": "time() == BOOL timestamp(rate(kube_pod_container_status_restarts_total{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[2m]) > 0)",
+ "hide": false,
+ "iconColor": "rgba(215, 44, 44, 1)",
+ "name": "Restarts",
+ "showIn": 0,
+ "tags": [
+ "restart"
+ ],
+ "type": "rows"
+ }
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by(container) (container_memory_usage_bytes{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\", container!=\"POD\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Current: {{ container }}",
+ "refId": "A"
+ },
+ {
+ "expr": "sum by(container) (kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\", pod=\"$pod\", container=~\"$container\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Requested: {{ container }}",
+ "refId": "B"
+ },
+ {
+ "expr": "sum by(container) (kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\", pod=\"$pod\", container=~\"$container\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Limit: {{ container }}",
+ "refId": "C"
+ },
+ {
+ "expr": "sum by(container) (container_memory_cache{job=\"kubelet\", namespace=\"$namespace\", pod=~\"$pod\", container=~\"$container\", container!=\"POD\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Cache: {{ container }}",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (container) (irate(container_cpu_usage_seconds_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", image!=\"\", pod=\"$pod\", container=~\"$container\", container!=\"POD\"}[4m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Current: {{ container }}",
+ "refId": "A"
+ },
+ {
+ "expr": "sum by(container) (kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\", pod=\"$pod\", container=~\"$container\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Requested: {{ container }}",
+ "refId": "B"
+ },
+ {
+ "expr": "sum by(container) (kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\", pod=\"$pod\", container=~\"$container\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Limit: {{ container }}",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sort_desc(sum by (pod) (irate(container_network_receive_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[4m])))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "RX: {{ pod }}",
+ "refId": "A"
+ },
+ {
+ "expr": "sort_desc(sum by (pod) (irate(container_network_transmit_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[4m])))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "TX: {{ pod }}",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network I/O",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 5,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max by (container) (kube_pod_container_status_restarts_total{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Restarts: {{ container }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Total Restarts Per Container",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
+
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Namespace",
+ "multi": false,
+ "name": "namespace",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Pod",
+ "multi": false,
+ "name": "pod",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{cluster=\"$cluster\", namespace=~\"$namespace\"}, pod)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": "Container",
+ "multi": false,
+ "name": "container",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_container_info{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}, container)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Pods",
+ "uid": "ab4f13a9892a76a4d21ce8c2445bf4ea",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-pods
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ proxy.json: |-
+ {
+ "__inputs": [
+
+ ],
+ "__requires": [
+
+ ],
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(up{job=\"kube-proxy\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Up",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 5,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(kubeproxy_sync_proxy_rules_duration_seconds_count{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "rate",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Rules Sync Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 5,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99,rate(kubeproxy_sync_proxy_rules_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Rule Sync Latency 99th Quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 5,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(kubeproxy_network_programming_duration_seconds_count{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "rate",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Programming Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(kubeproxy_network_programming_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Programming Latency 99th Quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 7,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "2xx",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "3xx",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "4xx",
+ "refId": "C"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "5xx",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Kube API Request Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 8,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 8,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-proxy\",instance=~\"$instance\",verb=\"POST\"}[5m])) by (verb, url, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{verb}} {{url}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Post Request Latency 99th Quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{verb}} {{url}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Get Request Latency 99th Quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 10,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "process_resident_memory_bytes{job=\"kube-proxy\",instance=~\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 11,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(process_cpu_seconds_total{job=\"kube-proxy\",instance=~\"$instance\"}[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 12,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "go_goroutines{job=\"kube-proxy\",instance=~\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Goroutines",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+ "kubernetes-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [
+
+ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "instance",
+ "options": [
+
+ ],
+ "query": "label_values(kubeproxy_network_programming_duration_seconds_bucket{job=\"kube-proxy\"}, instance)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Proxy",
+ "uid": "632e265de029684c40b21cb76bca4f94",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-proxy
+ namespace: monitoring
+- apiVersion: v1
+ data:
+ scheduler.json: |-
+ {
+ "__inputs": [
+
+ ],
+ "__requires": [
+
+ ],
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(up{job=\"kube-scheduler\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Up",
+ "tooltip": {
+ "shared": false
+ },
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "min"
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 5,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(scheduler_e2e_scheduling_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} e2e",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(scheduler_binding_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} binding",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(rate(scheduler_scheduling_algorithm_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} scheduling algorithm",
+ "refId": "C"
+ },
+ {
+ "expr": "sum(rate(scheduler_volume_scheduling_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} volume",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Scheduling Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
+ "max": false,
+ "min": false,
+ "rightSide": "true",
+ "show": "true",
+ "total": false,
+ "values": "true"
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 5,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} e2e",
+ "refId": "A"
+ },
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(scheduler_binding_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} binding",
+ "refId": "B"
+ },
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} scheduling algorithm",
+ "refId": "C"
+ },
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(scheduler_volume_scheduling_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} volume",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Scheduling latency 99th Quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 5,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "2xx",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "3xx",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "4xx",
+ "refId": "C"
+ },
+ {
+ "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "5xx",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Kube API Request Rate",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [
+
+ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 8,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-scheduler\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{verb}} {{url}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Post Request Latency 99th Quantile",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
{
"collapse": false,
"collapsed": false,
@@ -9167,17 +16409,17 @@ items:
"gridPos": {
},
- "id": 2,
+ "id": 7,
"legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
+ "alignAsTable": "true",
+ "avg": false,
+ "current": "true",
"max": false,
"min": false,
- "rightSide": true,
- "show": true,
+ "rightSide": "true",
+ "show": "true",
"total": false,
- "values": false
+ "values": "true"
},
"lines": true,
"linewidth": 1,
@@ -9199,32 +16441,11 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum by(container_name) (container_memory_usage_bytes{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-scheduler\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Current: {{ container_name }}",
+ "legendFormat": "{{verb}} {{url}}",
"refId": "A"
- },
- {
- "expr": "sum by(container) (kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\", pod=\"$pod\", container=~\"$container\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Requested: {{ container }}",
- "refId": "B"
- },
- {
- "expr": "sum by(container) (kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\", pod=\"$pod\", container=~\"$container\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Limit: {{ container }}",
- "refId": "C"
- },
- {
- "expr": "sum by(container_name) (container_memory_cache{job=\"kubelet\", namespace=\"$namespace\", pod_name=~\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Cache: {{ container_name }}",
- "refId": "D"
}
],
"thresholds": [
@@ -9232,7 +16453,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage",
+ "title": "Get Request Latency 99th Quantile",
"tooltip": {
"shared": false,
"sort": 0,
@@ -9250,7 +16471,7 @@ items:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
@@ -9258,7 +16479,7 @@ items:
"show": true
},
{
- "format": "bytes",
+ "format": "s",
"label": null,
"logBase": 1,
"max": null,
@@ -9292,14 +16513,14 @@ items:
"gridPos": {
},
- "id": 3,
+ "id": 8,
"legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
"max": false,
"min": false,
- "rightSide": true,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -9319,30 +16540,16 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum by (container_name) (rate(container_cpu_usage_seconds_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", image!=\"\", pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"}[1m]))",
+ "expr": "process_resident_memory_bytes{job=\"kube-scheduler\", instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Current: {{ container_name }}",
+ "legendFormat": "{{instance}}",
"refId": "A"
- },
- {
- "expr": "sum by(container) (kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\", pod=\"$pod\", container=~\"$container\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Requested: {{ container }}",
- "refId": "B"
- },
- {
- "expr": "sum by(container) (kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\", pod=\"$pod\", container=~\"$container\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Limit: {{ container }}",
- "refId": "C"
}
],
"thresholds": [
@@ -9350,7 +16557,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "CPU Usage",
+ "title": "Memory",
"tooltip": {
"shared": false,
"sort": 0,
@@ -9368,36 +16575,23 @@ items:
},
"yaxes": [
{
- "format": "short",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
- "format": "short",
+ "format": "bytes",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "panels": [
+ },
{
"aliasColors": {
@@ -9410,14 +16604,14 @@ items:
"gridPos": {
},
- "id": 4,
+ "id": 9,
"legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
"max": false,
"min": false,
- "rightSide": true,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -9437,23 +16631,16 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}[1m])))",
+ "expr": "rate(process_cpu_seconds_total{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "RX: {{ pod_name }}",
+ "legendFormat": "{{instance}}",
"refId": "A"
- },
- {
- "expr": "sort_desc(sum by (pod_name) (rate(container_network_transmit_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}[1m])))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "TX: {{ pod_name }}",
- "refId": "B"
}
],
"thresholds": [
@@ -9461,7 +16648,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Network I/O",
+ "title": "CPU usage",
"tooltip": {
"shared": false,
"sort": 0,
@@ -9495,20 +16682,7 @@ items:
"show": true
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "panels": [
+ },
{
"aliasColors": {
@@ -9521,14 +16695,14 @@ items:
"gridPos": {
},
- "id": 5,
+ "id": 10,
"legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
"max": false,
"min": false,
- "rightSide": true,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -9548,15 +16722,15 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "max by (container) (kube_pod_container_status_restarts_total{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})",
+ "expr": "go_goroutines{job=\"kube-scheduler\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Restarts: {{ container }}",
+ "legendFormat": "{{instance}}",
"refId": "A"
}
],
@@ -9565,7 +16739,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "Total Restarts Per Container",
+ "title": "Goroutines",
"tooltip": {
"shared": false,
"sort": 0,
@@ -9587,7 +16761,7 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
@@ -9595,7 +16769,7 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
}
]
@@ -9637,95 +16811,17 @@ items:
"allValue": null,
"current": {
- },
- "datasource": "$datasource",
- "hide": 2,
- "includeAll": false,
- "label": "cluster",
- "multi": false,
- "name": "cluster",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info, cluster)",
- "refresh": 2,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
-
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "Namespace",
- "multi": false,
- "name": "namespace",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)",
- "refresh": 2,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
-
- },
- "datasource": "$datasource",
- "hide": 0,
- "includeAll": false,
- "label": "Pod",
- "multi": false,
- "name": "pod",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info{cluster=\"$cluster\", namespace=~\"$namespace\"}, pod)",
- "refresh": 2,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
-
},
"datasource": "$datasource",
"hide": 0,
"includeAll": true,
- "label": "Container",
+ "label": null,
"multi": false,
- "name": "container",
+ "name": "instance",
"options": [
],
- "query": "label_values(kube_pod_container_info{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}, container)",
+ "query": "label_values(process_cpu_seconds_total{job=\"kube-scheduler\"}, instance)",
"refresh": 2,
"regex": "",
"sort": 0,
@@ -9769,13 +16865,13 @@ items:
]
},
"timezone": "",
- "title": "Kubernetes / Pods",
- "uid": "ab4f13a9892a76a4d21ce8c2445bf4ea",
+ "title": "Kubernetes / Scheduler",
+ "uid": "2e6b6a3b4bddf1427b3a55aa1311c656",
"version": 0
}
kind: ConfigMap
metadata:
- name: grafana-dashboard-pods
+ name: grafana-dashboard-scheduler
namespace: monitoring
- apiVersion: v1
data:
@@ -9866,7 +16962,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(rate(container_cpu_usage_seconds_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}[3m]))",
+ "expr": "sum(rate(container_cpu_usage_seconds_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}[3m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "",
@@ -9949,7 +17045,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(container_memory_usage_bytes{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}) / 1024^3",
+ "expr": "sum(container_memory_usage_bytes{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}) / 1024^3",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "",
@@ -10032,7 +17128,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(rate(container_network_transmit_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=\"$namespace\",pod_name=~\"$statefulset.*\"}[3m]))",
+ "expr": "sum(rate(container_network_transmit_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=\"$namespace\",pod=~\"$statefulset.*\"}[3m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "",
diff --git a/manifests/grafana-deployment.yaml b/manifests/grafana-deployment.yaml
index c977e450991afc4f4848a9f0ccd743d2ef7338ba..98719fbca5193a74fb3e9b13133bd8e43844efea 100644
--- a/manifests/grafana-deployment.yaml
+++ b/manifests/grafana-deployment.yaml
@@ -42,6 +42,12 @@ spec:
- mountPath: /etc/grafana/provisioning/dashboards
name: grafana-dashboards
readOnly: false
+ - mountPath: /grafana-dashboard-definitions/0/apiserver
+ name: grafana-dashboard-apiserver
+ readOnly: false
+ - mountPath: /grafana-dashboard-definitions/0/controller-manager
+ name: grafana-dashboard-controller-manager
+ readOnly: false
- mountPath: /grafana-dashboard-definitions/0/k8s-cluster-rsrc-use
name: grafana-dashboard-k8s-cluster-rsrc-use
readOnly: false
@@ -63,6 +69,9 @@ spec:
- mountPath: /grafana-dashboard-definitions/0/k8s-resources-workloads-namespace
name: grafana-dashboard-k8s-resources-workloads-namespace
readOnly: false
+ - mountPath: /grafana-dashboard-definitions/0/kubelet
+ name: grafana-dashboard-kubelet
+ readOnly: false
- mountPath: /grafana-dashboard-definitions/0/nodes
name: grafana-dashboard-nodes
readOnly: false
@@ -72,6 +81,12 @@ spec:
- mountPath: /grafana-dashboard-definitions/0/pods
name: grafana-dashboard-pods
readOnly: false
+ - mountPath: /grafana-dashboard-definitions/0/proxy
+ name: grafana-dashboard-proxy
+ readOnly: false
+ - mountPath: /grafana-dashboard-definitions/0/scheduler
+ name: grafana-dashboard-scheduler
+ readOnly: false
- mountPath: /grafana-dashboard-definitions/0/statefulset
name: grafana-dashboard-statefulset
readOnly: false
@@ -90,6 +105,12 @@ spec:
- configMap:
name: grafana-dashboards
name: grafana-dashboards
+ - configMap:
+ name: grafana-dashboard-apiserver
+ name: grafana-dashboard-apiserver
+ - configMap:
+ name: grafana-dashboard-controller-manager
+ name: grafana-dashboard-controller-manager
- configMap:
name: grafana-dashboard-k8s-cluster-rsrc-use
name: grafana-dashboard-k8s-cluster-rsrc-use
@@ -111,6 +132,9 @@ spec:
- configMap:
name: grafana-dashboard-k8s-resources-workloads-namespace
name: grafana-dashboard-k8s-resources-workloads-namespace
+ - configMap:
+ name: grafana-dashboard-kubelet
+ name: grafana-dashboard-kubelet
- configMap:
name: grafana-dashboard-nodes
name: grafana-dashboard-nodes
@@ -120,6 +144,12 @@ spec:
- configMap:
name: grafana-dashboard-pods
name: grafana-dashboard-pods
+ - configMap:
+ name: grafana-dashboard-proxy
+ name: grafana-dashboard-proxy
+ - configMap:
+ name: grafana-dashboard-scheduler
+ name: grafana-dashboard-scheduler
- configMap:
name: grafana-dashboard-statefulset
name: grafana-dashboard-statefulset
diff --git a/manifests/prometheus-rules.yaml b/manifests/prometheus-rules.yaml
index 92b131c9286e291dc8484d6b2f654fbe79f8ba43..1cb939cd71e8aee36d68d3cbc7bb756ac00c49f6 100644
--- a/manifests/prometheus-rules.yaml
+++ b/manifests/prometheus-rules.yaml
@@ -11,30 +11,44 @@ spec:
- name: k8s.rules
rules:
- expr: |
- sum(rate(container_cpu_usage_seconds_total{job="kubelet", image!="", container_name!=""}[5m])) by (namespace)
+ sum(rate(container_cpu_usage_seconds_total{job="kubelet", image!="", container!="POD"}[5m])) by (namespace)
record: namespace:container_cpu_usage_seconds_total:sum_rate
- expr: |
- sum(container_memory_usage_bytes{job="kubelet", image!="", container_name!=""}) by (namespace)
+ sum by (namespace, pod, container) (
+ rate(container_cpu_usage_seconds_total{job="kubelet", image!="", container!="POD"}[5m])
+ )
+ record: namespace_pod_container:container_cpu_usage_seconds_total:sum_rate
+ - expr: |
+ sum(container_memory_usage_bytes{job="kubelet", image!="", container!="POD"}) by (namespace)
record: namespace:container_memory_usage_bytes:sum
- expr: |
- sum by (namespace, pod_name, container_name) (
- rate(container_cpu_usage_seconds_total{job="kubelet", image!="", container_name!=""}[5m])
+ sum by (namespace, label_name) (
+ sum(rate(container_cpu_usage_seconds_total{job="kubelet", image!="", container!="POD"}[5m])) by (namespace, pod)
+ * on (namespace, pod)
+ group_left(label_name) kube_pod_labels{job="kube-state-metrics"}
)
- record: namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate
+ record: namespace:container_cpu_usage_seconds_total:sum_rate
- expr: |
- sum by(namespace) (
- kube_pod_container_resource_requests_memory_bytes{job="kube-state-metrics"}
- * on (endpoint, instance, job, namespace, pod, service)
- group_left(phase) (kube_pod_status_phase{phase=~"^(Pending|Running)$"} == 1)
+ sum by (namespace, label_name) (
+ sum(container_memory_usage_bytes{job="kubelet",image!="", container!="POD"}) by (pod, namespace)
+ * on (namespace, pod)
+ group_left(label_name) kube_pod_labels{job="kube-state-metrics"}
)
- record: namespace_name:kube_pod_container_resource_requests_memory_bytes:sum
+ record: namespace:container_memory_usage_bytes:sum
- expr: |
- sum by (namespace) (
- kube_pod_container_resource_requests_cpu_cores{job="kube-state-metrics"}
- * on (endpoint, instance, job, namespace, pod, service)
- group_left(phase) (kube_pod_status_phase{phase=~"^(Pending|Running)$"} == 1)
+ sum by (namespace, label_name) (
+ sum(kube_pod_container_resource_requests_memory_bytes{job="kube-state-metrics"} * on (endpoint, instance, job, namespace, pod, service) group_left(phase) (kube_pod_status_phase{phase=~"^(Pending|Running)$"} == 1)) by (namespace, pod)
+ * on (namespace, pod)
+ group_left(label_name) kube_pod_labels{job="kube-state-metrics"}
)
- record: namespace_name:kube_pod_container_resource_requests_cpu_cores:sum
+ record: namespace:kube_pod_container_resource_requests_memory_bytes:sum
+ - expr: |
+ sum by (namespace, label_name) (
+ sum(kube_pod_container_resource_requests_cpu_cores{job="kube-state-metrics"} * on (endpoint, instance, job, namespace, pod, service) group_left(phase) (kube_pod_status_phase{phase=~"^(Pending|Running)$"} == 1)) by (namespace, pod)
+ * on (namespace, pod)
+ group_left(label_name) kube_pod_labels{job="kube-state-metrics"}
+ )
+ record: namespace:kube_pod_container_resource_requests_cpu_cores:sum
- expr: |
sum(
label_replace(
@@ -71,67 +85,67 @@ spec:
- name: kube-scheduler.rules
rules:
- expr: |
- histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.99"
- record: cluster_quantile:scheduler_e2e_scheduling_latency:histogram_quantile
+ record: cluster_quantile:scheduler_e2e_scheduling_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.99"
- record: cluster_quantile:scheduler_scheduling_algorithm_latency:histogram_quantile
+ record: cluster_quantile:scheduler_scheduling_algorithm_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.99, sum(rate(scheduler_binding_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.99, sum(rate(scheduler_binding_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.99"
- record: cluster_quantile:scheduler_binding_latency:histogram_quantile
+ record: cluster_quantile:scheduler_binding_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.9, sum(rate(scheduler_e2e_scheduling_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.9, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.9"
- record: cluster_quantile:scheduler_e2e_scheduling_latency:histogram_quantile
+ record: cluster_quantile:scheduler_e2e_scheduling_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.9, sum(rate(scheduler_scheduling_algorithm_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.9, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.9"
- record: cluster_quantile:scheduler_scheduling_algorithm_latency:histogram_quantile
+ record: cluster_quantile:scheduler_scheduling_algorithm_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.9, sum(rate(scheduler_binding_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.9, sum(rate(scheduler_binding_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.9"
- record: cluster_quantile:scheduler_binding_latency:histogram_quantile
+ record: cluster_quantile:scheduler_binding_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.5, sum(rate(scheduler_e2e_scheduling_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.5, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.5"
- record: cluster_quantile:scheduler_e2e_scheduling_latency:histogram_quantile
+ record: cluster_quantile:scheduler_e2e_scheduling_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.5, sum(rate(scheduler_scheduling_algorithm_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.5, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.5"
- record: cluster_quantile:scheduler_scheduling_algorithm_latency:histogram_quantile
+ record: cluster_quantile:scheduler_scheduling_algorithm_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.5, sum(rate(scheduler_binding_latency_microseconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.5, sum(rate(scheduler_binding_duration_seconds_bucket{job="kube-scheduler"}[5m])) without(instance, pod))
labels:
quantile: "0.5"
- record: cluster_quantile:scheduler_binding_latency:histogram_quantile
+ record: cluster_quantile:scheduler_binding_duration_seconds:histogram_quantile
- name: kube-apiserver.rules
rules:
- expr: |
- histogram_quantile(0.99, sum(rate(apiserver_request_latencies_bucket{job="apiserver"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver"}[5m])) without(instance, pod))
labels:
quantile: "0.99"
- record: cluster_quantile:apiserver_request_latencies:histogram_quantile
+ record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.9, sum(rate(apiserver_request_latencies_bucket{job="apiserver"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.9, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver"}[5m])) without(instance, pod))
labels:
quantile: "0.9"
- record: cluster_quantile:apiserver_request_latencies:histogram_quantile
+ record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
- expr: |
- histogram_quantile(0.5, sum(rate(apiserver_request_latencies_bucket{job="apiserver"}[5m])) without(instance, pod)) / 1e+06
+ histogram_quantile(0.5, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver"}[5m])) without(instance, pod))
labels:
quantile: "0.5"
- record: cluster_quantile:apiserver_request_latencies:histogram_quantile
+ record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
- name: node.rules
rules:
- expr: sum(min(kube_pod_info) by (node))
@@ -593,11 +607,11 @@ spec:
tolerate node failure.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubecpuovercommit
expr: |
- sum(namespace_name:kube_pod_container_resource_requests_cpu_cores:sum)
+ sum(namespace:kube_pod_container_resource_requests_cpu_cores:sum)
/
- sum(node:node_num_cpu:sum)
+ sum(kube_node_status_allocatable_cpu_cores)
>
- (count(node:node_num_cpu:sum)-1) / count(node:node_num_cpu:sum)
+ (count(kube_node_status_allocatable_cpu_cores)-1) / count(kube_node_status_allocatable_cpu_cores)
for: 5m
labels:
severity: warning
@@ -607,13 +621,13 @@ spec:
tolerate node failure.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubememovercommit
expr: |
- sum(namespace_name:kube_pod_container_resource_requests_memory_bytes:sum)
+ sum(namespace:kube_pod_container_resource_requests_memory_bytes:sum)
/
- sum(node_memory_MemTotal_bytes)
+ sum(kube_node_status_allocatable_memory_bytes)
>
- (count(node:node_num_cpu:sum)-1)
+ (count(kube_node_status_allocatable_memory_bytes)-1)
/
- count(node:node_num_cpu:sum)
+ count(kube_node_status_allocatable_memory_bytes)
for: 5m
labels:
severity: warning
@@ -624,7 +638,7 @@ spec:
expr: |
sum(kube_resourcequota{job="kube-state-metrics", type="hard", resource="cpu"})
/
- sum(node:node_num_cpu:sum)
+ sum(kube_node_status_allocatable_cpu_cores)
> 1.5
for: 5m
labels:
@@ -636,7 +650,7 @@ spec:
expr: |
sum(kube_resourcequota{job="kube-state-metrics", type="hard", resource="memory"})
/
- sum(node_memory_MemTotal_bytes{job="node-exporter"})
+ sum(kube_node_status_allocatable_memory_bytes{job="node-exporter"})
> 1.5
for: 5m
labels:
@@ -657,12 +671,11 @@ spec:
- alert: CPUThrottlingHigh
annotations:
message: '{{ printf "%0.0f" $value }}% throttling of CPU in namespace {{ $labels.namespace
- }} for container {{ $labels.container_name }} in pod {{ $labels.pod_name
- }}.'
+ }} for container {{ $labels.container }} in pod {{ $labels.pod }}.'
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-cputhrottlinghigh
- expr: "100 * sum(increase(container_cpu_cfs_throttled_periods_total{container_name!=\"\",
- }[5m])) by (container_name, pod_name, namespace)\n /\nsum(increase(container_cpu_cfs_periods_total{}[5m]))
- by (container_name, pod_name, namespace)\n > 25 \n"
+ expr: "100 * sum(increase(container_cpu_cfs_throttled_periods_total{container!=\"\",
+ }[5m])) by (container, pod, namespace)\n /\nsum(increase(container_cpu_cfs_periods_total{}[5m]))
+ by (container, pod, namespace)\n > 25 \n"
for: 15m
labels:
severity: warning
@@ -769,7 +782,7 @@ spec:
for {{ $labels.verb }} {{ $labels.resource }}.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapilatencyhigh
expr: |
- cluster_quantile:apiserver_request_latencies:histogram_quantile{job="apiserver",quantile="0.99",subresource!="log",verb!~"^(?:LIST|WATCH|WATCHLIST|PROXY|CONNECT)$"} > 1
+ cluster_quantile:apiserver_request_duration_seconds:histogram_quantile{job="apiserver",quantile="0.99",subresource!="log",verb!~"^(?:LIST|WATCH|WATCHLIST|PROXY|CONNECT)$"} > 1
for: 10m
labels:
severity: warning
@@ -779,7 +792,7 @@ spec:
for {{ $labels.verb }} {{ $labels.resource }}.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapilatencyhigh
expr: |
- cluster_quantile:apiserver_request_latencies:histogram_quantile{job="apiserver",quantile="0.99",subresource!="log",verb!~"^(?:LIST|WATCH|WATCHLIST|PROXY|CONNECT)$"} > 4
+ cluster_quantile:apiserver_request_duration_seconds:histogram_quantile{job="apiserver",quantile="0.99",subresource!="log",verb!~"^(?:LIST|WATCH|WATCHLIST|PROXY|CONNECT)$"} > 4
for: 10m
labels:
severity: critical
@@ -788,9 +801,9 @@ spec:
message: API server is returning errors for {{ $value }}% of requests.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapierrorshigh
expr: |
- sum(rate(apiserver_request_count{job="apiserver",code=~"^(?:5..)$"}[5m]))
+ sum(rate(apiserver_request_total{job="apiserver",code=~"^(?:5..)$"}[5m]))
/
- sum(rate(apiserver_request_count{job="apiserver"}[5m])) * 100 > 3
+ sum(rate(apiserver_request_total{job="apiserver"}[5m])) * 100 > 3
for: 10m
labels:
severity: critical
@@ -799,9 +812,9 @@ spec:
message: API server is returning errors for {{ $value }}% of requests.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapierrorshigh
expr: |
- sum(rate(apiserver_request_count{job="apiserver",code=~"^(?:5..)$"}[5m]))
+ sum(rate(apiserver_request_total{job="apiserver",code=~"^(?:5..)$"}[5m]))
/
- sum(rate(apiserver_request_count{job="apiserver"}[5m])) * 100 > 1
+ sum(rate(apiserver_request_total{job="apiserver"}[5m])) * 100 > 1
for: 10m
labels:
severity: warning
@@ -811,9 +824,9 @@ spec:
{{ $labels.verb }} {{ $labels.resource }} {{ $labels.subresource }}.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapierrorshigh
expr: |
- sum(rate(apiserver_request_count{job="apiserver",code=~"^(?:5..)$"}[5m])) by (resource,subresource,verb)
+ sum(rate(apiserver_request_total{job="apiserver",code=~"^(?:5..)$"}[5m])) by (resource,subresource,verb)
/
- sum(rate(apiserver_request_count{job="apiserver"}[5m])) by (resource,subresource,verb) * 100 > 10
+ sum(rate(apiserver_request_total{job="apiserver"}[5m])) by (resource,subresource,verb) * 100 > 10
for: 10m
labels:
severity: critical
@@ -823,9 +836,9 @@ spec:
{{ $labels.verb }} {{ $labels.resource }} {{ $labels.subresource }}.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapierrorshigh
expr: |
- sum(rate(apiserver_request_count{job="apiserver",code=~"^(?:5..)$"}[5m])) by (resource,subresource,verb)
+ sum(rate(apiserver_request_total{job="apiserver",code=~"^(?:5..)$"}[5m])) by (resource,subresource,verb)
/
- sum(rate(apiserver_request_count{job="apiserver"}[5m])) by (resource,subresource,verb) * 100 > 5
+ sum(rate(apiserver_request_total{job="apiserver"}[5m])) by (resource,subresource,verb) * 100 > 5
for: 10m
labels:
severity: warning