diff --git a/docs/developing-prometheus-rules-and-grafana-dashboards.md b/docs/developing-prometheus-rules-and-grafana-dashboards.md
index 99cf271c03406f93b14805192cf5252899b912d7..e0dbf7528117e76de5481bdfd33ed5955381d7cc 100644
--- a/docs/developing-prometheus-rules-and-grafana-dashboards.md
+++ b/docs/developing-prometheus-rules-and-grafana-dashboards.md
@@ -252,30 +252,32 @@ local kp = (import 'kube-prometheus/kube-prometheus.libsonnet') + {
_config+:: {
namespace: 'monitoring',
},
- grafanaDashboards+:: {
- 'my-dashboard.json':
- dashboard.new('My Dashboard')
- .addTemplate(
- {
- current: {
- text: 'Prometheus',
- value: 'Prometheus',
+ grafana+:: {
+ dashboards+:: {
+ 'my-dashboard.json':
+ dashboard.new('My Dashboard')
+ .addTemplate(
+ {
+ current: {
+ text: 'Prometheus',
+ value: 'Prometheus',
+ },
+ hide: 0,
+ label: null,
+ name: 'datasource',
+ options: [],
+ query: 'prometheus',
+ refresh: 1,
+ regex: '',
+ type: 'datasource',
},
- hide: 0,
- label: null,
- name: 'datasource',
- options: [],
- query: 'prometheus',
- refresh: 1,
- regex: '',
- type: 'datasource',
- },
- )
- .addRow(
- row.new()
- .addPanel(graphPanel.new('My Panel', span=6, datasource='$datasource')
- .addTarget(prometheus.target('vector(1)')))
- ),
+ )
+ .addRow(
+ row.new()
+ .addPanel(graphPanel.new('My Panel', span=6, datasource='$datasource')
+ .addTarget(prometheus.target('vector(1)')))
+ ),
+ },
},
};
@@ -298,9 +300,14 @@ local kp = (import 'kube-prometheus/kube-prometheus.libsonnet') + {
_config+:: {
namespace: 'monitoring',
},
- grafanaDashboards+:: {
+ grafanaDashboards+:: { // monitoring-mixin compatibility
'my-dashboard.json': (import 'example-grafana-dashboard.json'),
},
+ grafana+:: {
+ dashboards+:: { // use this method to import your dashboards to Grafana
+ 'my-dashboard.json': (import 'example-grafana-dashboard.json'),
+ },
+ },
};
{ ['00namespace-' + name]: kp.kubePrometheus[name] for name in std.objectFields(kp.kubePrometheus) } +
@@ -319,8 +326,10 @@ local kp = (import 'kube-prometheus/kube-prometheus.libsonnet') + {
_config+:: {
namespace: 'monitoring',
},
- rawGrafanaDashboards+:: {
- 'my-dashboard.json': (importstr 'example-grafana-dashboard.json'),
+ grafana+:: {
+ rawDashboards+:: {
+ 'my-dashboard.json': (importstr 'example-grafana-dashboard.json'),
+ },
},
};
diff --git a/jsonnet/kube-prometheus/jsonnetfile.json b/jsonnet/kube-prometheus/jsonnetfile.json
index c00bc857b2a725d61da3b4d59bdecbca10453c2d..50005ea29f0c19f1449f8ff4603fee600f9c7af0 100644
--- a/jsonnet/kube-prometheus/jsonnetfile.json
+++ b/jsonnet/kube-prometheus/jsonnetfile.json
@@ -26,7 +26,7 @@
"subdir": "jsonnet/prometheus-operator"
}
},
- "version": "release-0.39"
+ "version": "release-0.40"
},
{
"source": {
diff --git a/jsonnetfile.lock.json b/jsonnetfile.lock.json
index 6a8f854ed86b8237bb6a8639bb417ff987625376..4eb52890ff9438203938520c8a4e4d03dbfcc375 100644
--- a/jsonnetfile.lock.json
+++ b/jsonnetfile.lock.json
@@ -4,7 +4,7 @@
{
"source": {
"git": {
- "remote": "https://github.com/brancz/kubernetes-grafana",
+ "remote": "https://github.com/brancz/kubernetes-grafana.git",
"subdir": "grafana"
}
},
@@ -14,47 +14,47 @@
{
"source": {
"git": {
- "remote": "https://github.com/coreos/etcd",
+ "remote": "https://github.com/coreos/etcd.git",
"subdir": "Documentation/etcd-mixin"
}
},
- "version": "1166b1f195efae31439c7b3c913b4ef02e7df889",
- "sum": "Ko3qhNfC2vN/houLh6C0Ryacjv70gl0DVPGU/PQ4OD0="
+ "version": "d8c8f903eee10b8391abaef7758c38b2cd393c55",
+ "sum": "pk7mLpdUrHuJKkj2vhD6LGMU7P+oYYooBXAeZyZa398="
},
{
"source": {
"git": {
- "remote": "https://github.com/coreos/prometheus-operator",
+ "remote": "https://github.com/coreos/prometheus-operator.git",
"subdir": "jsonnet/prometheus-operator"
}
},
- "version": "d0a871b710de7b764c05ced98dbd1eb32a681790",
- "sum": "cIOKRTNBUOl3a+QsaA/NjClmZAhyVJHlDFReKlXJBAs="
+ "version": "e31c69f9b5c6555e0f4a5c1f39d0f03182dd6b41",
+ "sum": "WggWVWZ+CBEUThQCztSaRELbtqdXf9s3OFzf06HbYNA="
},
{
"source": {
"git": {
- "remote": "https://github.com/grafana/grafonnet-lib",
+ "remote": "https://github.com/grafana/grafonnet-lib.git",
"subdir": "grafonnet"
}
},
- "version": "906768d46973e022594d3f03d82c5a51d86de2cc",
- "sum": "J3Vp0EVbxTObr6KydLXsi4Rc0ssNVAEuwLc0NQ+4wqU="
+ "version": "8fb95bd89990e493a8534205ee636bfcb8db67bd",
+ "sum": "tDuuSKE9f4Ew2bjBM33Rs6behLEAzkmKkShSt+jpAak="
},
{
"source": {
"git": {
- "remote": "https://github.com/grafana/jsonnet-libs",
+ "remote": "https://github.com/grafana/jsonnet-libs.git",
"subdir": "grafana-builder"
}
},
- "version": "cb6bc2780a39afbbf9d4ee64fec8d1152023aee9",
+ "version": "881db2241f0c5007c3e831caf34b0c645202b4ab",
"sum": "slxrtftVDiTlQK22ertdfrg4Epnq97gdrLI63ftUfaE="
},
{
"source": {
"git": {
- "remote": "https://github.com/ksonnet/ksonnet-lib",
+ "remote": "https://github.com/ksonnet/ksonnet-lib.git",
"subdir": ""
}
},
@@ -65,62 +65,62 @@
{
"source": {
"git": {
- "remote": "https://github.com/kubernetes-monitoring/kubernetes-mixin",
+ "remote": "https://github.com/kubernetes-monitoring/kubernetes-mixin.git",
"subdir": ""
}
},
- "version": "3cc34f995c31ed6e1e92024fed1912d63569c39f",
- "sum": "r5Fg4KgiBtsFPCCHtM3Cb4CEgnizLyK97srDNAcjr+Y="
+ "version": "b61c5a34051f8f57284a08fe78ad8a45b430252b",
+ "sum": "7Hx/5eNm7ubLTsdrpk3b2+e/FLR3XOa4HCukmbRUCAY="
},
{
"source": {
"git": {
- "remote": "https://github.com/kubernetes-monitoring/kubernetes-mixin",
+ "remote": "https://github.com/kubernetes-monitoring/kubernetes-mixin.git",
"subdir": "lib/promgrafonnet"
}
},
- "version": "3cc34f995c31ed6e1e92024fed1912d63569c39f",
+ "version": "b61c5a34051f8f57284a08fe78ad8a45b430252b",
"sum": "VhgBM39yv0f4bKv8VfGg4FXkg573evGDRalip9ypKbc="
},
{
"source": {
"git": {
- "remote": "https://github.com/kubernetes/kube-state-metrics",
+ "remote": "https://github.com/kubernetes/kube-state-metrics.git",
"subdir": "jsonnet/kube-state-metrics"
}
},
- "version": "52fe3a268bd78c8f32a03361e28fdf23c41512c5",
+ "version": "d667979ed55ad1c4db44d331b51d646f5b903aa7",
"sum": "cJjGZaLBjcIGrLHZLjRPU9c3KL+ep9rZTb9dbALSKqA="
},
{
"source": {
"git": {
- "remote": "https://github.com/kubernetes/kube-state-metrics",
+ "remote": "https://github.com/kubernetes/kube-state-metrics.git",
"subdir": "jsonnet/kube-state-metrics-mixin"
}
},
- "version": "52fe3a268bd78c8f32a03361e28fdf23c41512c5",
- "sum": "E1GGavnf9PCWBm4WVrxWnc0FIj72UcbcweqGioWrOdU="
+ "version": "d667979ed55ad1c4db44d331b51d646f5b903aa7",
+ "sum": "o5avaguRsfFwYFNen00ZEsub1x4i8Z/ZZ2QoEjFMff8="
},
{
"source": {
"git": {
- "remote": "https://github.com/prometheus/node_exporter",
+ "remote": "https://github.com/prometheus/node_exporter.git",
"subdir": "docs/node-mixin"
}
},
- "version": "d4d2e1db98152ab6c94dc9a12a997950e0be2416",
- "sum": "ZwrC0+4o1xD6+oPBu1p+rBXLlf6pMBD9rT8ygyl2aW0="
+ "version": "08ce3c6dd430deb51798826701a395e460620d60",
+ "sum": "3jFV2qsc/GZe2GADswTYqxxP2zGOiANTj73W/VNFGqc="
},
{
"source": {
"git": {
- "remote": "https://github.com/prometheus/prometheus",
+ "remote": "https://github.com/prometheus/prometheus.git",
"subdir": "documentation/prometheus-mixin"
}
},
- "version": "209d4bb8a1491f4535cc6d991681e7dc03bb1d56",
- "sum": "kRb3XBTe/AALDcaTFfyuiKqzhxtLvihBkVkvJ5cUd/I=",
+ "version": "8d3c2f6829d73be15a6684f9324917e72fbf1a31",
+ "sum": "lEzhZ8gllSfAO4kmXeTwl4W0anapIeFd5GCaCNuDe18=",
"name": "prometheus"
},
{
diff --git a/manifests/grafana-dashboardDefinitions.yaml b/manifests/grafana-dashboardDefinitions.yaml
index ca9ea0033078325dbf160ef17a7d322587188594..8033c7d295bde48f2ce6917e93a213799fbb801c 100644
--- a/manifests/grafana-dashboardDefinitions.yaml
+++ b/manifests/grafana-dashboardDefinitions.yaml
@@ -22,6 +22,24 @@ items:
"id": null,
"links": [
+ ],
+ "panels": [
+ {
+ "content": "The SLO (service level objective) and other metrics displayed on this dashboard are for informational purposes only.",
+ "datasource": null,
+ "description": "The SLO (service level objective) and other metrics displayed on this dashboard are for informational purposes only.",
+ "gridPos": {
+ "h": 2,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "mode": "markdown",
+ "span": 12,
+ "title": "Notice",
+ "type": "text"
+ }
],
"refresh": "10s",
"rows": [
@@ -40,6 +58,7 @@ items:
],
"datasource": "$datasource",
"decimals": 3,
+ "description": "How many percent of requests (both read and write) in 30 days have been answered successfully and fast enough?",
"format": "percentunit",
"gauge": {
"maxValue": 100,
@@ -51,7 +70,7 @@ items:
"gridPos": {
},
- "id": 2,
+ "id": 3,
"interval": null,
"links": [
@@ -99,7 +118,7 @@ items:
}
],
"thresholds": "",
- "title": "Availability (30d) > 99.000",
+ "title": "Availability (30d) > 99.000%",
"tooltip": {
"shared": false
},
@@ -123,11 +142,12 @@ items:
"dashes": false,
"datasource": "$datasource",
"decimals": 3,
- "fill": 1,
+ "description": "How much error budget is left looking at our 0.990% availability gurantees?",
+ "fill": 10,
"gridPos": {
},
- "id": 3,
+ "id": 4,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -136,6 +156,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -171,7 +192,7 @@ items:
],
"timeFrom": null,
"timeShift": null,
- "title": "ErrorBudget (30d) > 99.000",
+ "title": "ErrorBudget (30d) > 99.000%",
"tooltip": {
"shared": false,
"sort": 0,
@@ -232,6 +253,7 @@ items:
],
"datasource": "$datasource",
"decimals": 3,
+ "description": "How many percent of read requests (LIST,GET) in 30 days have been answered successfully and fast enough?",
"format": "percentunit",
"gauge": {
"maxValue": 100,
@@ -243,7 +265,7 @@ items:
"gridPos": {
},
- "id": 4,
+ "id": 5,
"interval": null,
"links": [
@@ -314,11 +336,12 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
+ "description": "How many read requests (LIST,GET) per second do the apiservers get by code?",
+ "fill": 10,
"gridPos": {
},
- "id": 5,
+ "id": 6,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -327,6 +350,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -342,18 +366,33 @@ items:
"renderer": "flot",
"repeat": null,
"seriesOverrides": [
-
+ {
+ "alias": "/2../i",
+ "color": "#56A64B"
+ },
+ {
+ "alias": "/3../i",
+ "color": "#F2CC0C"
+ },
+ {
+ "alias": "/4../i",
+ "color": "#3274D9"
+ },
+ {
+ "alias": "/5../i",
+ "color": "#E02F44"
+ }
],
"spaceLength": 10,
"span": 3,
- "stack": false,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(code_resource:apiserver_request_total:rate5m{verb=\"read\"})",
+ "expr": "sum by (code) (code_resource:apiserver_request_total:rate5m{verb=\"read\"})",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "",
+ "legendFormat": "{{ code }}",
"refId": "A"
}
],
@@ -405,11 +444,12 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
+ "description": "How many percent of read requests (LIST,GET) per second are returned with errors (5xx)?",
"fill": 1,
"gridPos": {
},
- "id": 6,
+ "id": 7,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -418,6 +458,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -475,7 +516,7 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
@@ -483,7 +524,7 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
}
]
@@ -496,11 +537,12 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
+ "description": "How many seconds is the 99th percentile for reading (LIST|GET) a given resource?",
"fill": 1,
"gridPos": {
},
- "id": 7,
+ "id": 8,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -509,6 +551,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -603,6 +646,7 @@ items:
],
"datasource": "$datasource",
"decimals": 3,
+ "description": "How many percent of write requests (POST|PUT|PATCH|DELETE) in 30 days have been answered successfully and fast enough?",
"format": "percentunit",
"gauge": {
"maxValue": 100,
@@ -614,7 +658,7 @@ items:
"gridPos": {
},
- "id": 8,
+ "id": 9,
"interval": null,
"links": [
@@ -685,11 +729,12 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
+ "description": "How many write requests (POST|PUT|PATCH|DELETE) per second do the apiservers get by code?",
+ "fill": 10,
"gridPos": {
},
- "id": 9,
+ "id": 10,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -698,6 +743,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -713,18 +759,33 @@ items:
"renderer": "flot",
"repeat": null,
"seriesOverrides": [
-
+ {
+ "alias": "/2../i",
+ "color": "#56A64B"
+ },
+ {
+ "alias": "/3../i",
+ "color": "#F2CC0C"
+ },
+ {
+ "alias": "/4../i",
+ "color": "#3274D9"
+ },
+ {
+ "alias": "/5../i",
+ "color": "#E02F44"
+ }
],
"spaceLength": 10,
"span": 3,
- "stack": false,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(code_resource:apiserver_request_total:rate5m{verb=\"write\"})",
+ "expr": "sum by (code) (code_resource:apiserver_request_total:rate5m{verb=\"write\"})",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "",
+ "legendFormat": "{{ code }}",
"refId": "A"
}
],
@@ -776,11 +837,12 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
+ "description": "How many percent of write requests (POST|PUT|PATCH|DELETE) per second are returned with errors (5xx)?",
"fill": 1,
"gridPos": {
},
- "id": 10,
+ "id": 11,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -789,6 +851,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -846,7 +909,7 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
@@ -854,7 +917,7 @@ items:
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
}
]
@@ -867,11 +930,12 @@ items:
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
+ "description": "How many seconds is the 99th percentile for writing (POST|PUT|PATCH|DELETE) a given resource?",
"fill": 1,
"gridPos": {
},
- "id": 11,
+ "id": 12,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -880,6 +944,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -963,90 +1028,6 @@ items:
"collapse": false,
"collapsed": false,
"panels": [
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "#299c46",
- "rgba(237, 129, 40, 0.89)",
- "#d44a3a"
- ],
- "datasource": "$datasource",
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
-
- },
- "id": 12,
- "interval": null,
- "links": [
-
- ],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 2,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(up{job=\"apiserver\", cluster=\"$cluster\"})",
- "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": {
@@ -1060,222 +1041,6 @@ items:
},
"id": 13,
- "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(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"2..\", cluster=\"$cluster\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "2xx",
- "refId": "A"
- },
- {
- "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"3..\", cluster=\"$cluster\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "3xx",
- "refId": "B"
- },
- {
- "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"4..\", cluster=\"$cluster\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "4xx",
- "refId": "C"
- },
- {
- "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"5..\", cluster=\"$cluster\"}[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": null,
- "show": true
- },
- {
- "format": "ops",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 1,
- "gridPos": {
-
- },
- "id": 14,
- "legend": {
- "alignAsTable": true,
- "avg": false,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 1,
- "links": [
-
- ],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "seriesOverrides": [
-
- ],
- "spaceLength": 10,
- "span": 5,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\", instance=~\"$instance\", verb!=\"WATCH\", cluster=\"$cluster\"}[5m])) by (verb, le))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{verb}}",
- "refId": "A"
- }
- ],
- "thresholds": [
-
- ],
- "timeFrom": null,
- "timeShift": null,
- "title": "Request duration 99th quantile",
- "tooltip": {
- "shared": false,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": [
-
- ]
- },
- "yaxes": [
- {
- "format": "s",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "s",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "panels": [
- {
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "fill": 1,
- "gridPos": {
-
- },
- "id": 15,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -1284,6 +1049,7 @@ items:
"min": false,
"rightSide": false,
"show": false,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -1302,7 +1068,7 @@ items:
],
"spaceLength": 10,
- "span": 6,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
@@ -1366,7 +1132,7 @@ items:
"gridPos": {
},
- "id": 16,
+ "id": 14,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -1375,6 +1141,7 @@ items:
"min": false,
"rightSide": false,
"show": false,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -1393,7 +1160,7 @@ items:
],
"spaceLength": 10,
- "span": 6,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
@@ -1457,7 +1224,7 @@ items:
"gridPos": {
},
- "id": 17,
+ "id": 15,
"legend": {
"alignAsTable": true,
"avg": false,
@@ -1466,6 +1233,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -1484,7 +1252,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 4,
"stack": false,
"steppedLine": false,
"targets": [
@@ -1561,7 +1329,7 @@ items:
"gridPos": {
},
- "id": 18,
+ "id": 16,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -1570,6 +1338,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -1652,7 +1421,7 @@ items:
"gridPos": {
},
- "id": 19,
+ "id": 17,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -1661,6 +1430,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -1750,7 +1520,7 @@ items:
"gridPos": {
},
- "id": 20,
+ "id": 18,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -1759,6 +1529,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -1861,7 +1632,7 @@ items:
"gridPos": {
},
- "id": 21,
+ "id": 19,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -1870,6 +1641,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -1952,7 +1724,7 @@ items:
"gridPos": {
},
- "id": 22,
+ "id": 20,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -1961,6 +1733,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -2043,7 +1816,7 @@ items:
"gridPos": {
},
- "id": 23,
+ "id": 21,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -2052,6 +1825,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -2239,7 +2013,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / API server",
"uid": "09ec8aa1e996d6ffcd6817bbaff4db1b",
"version": 0
@@ -2327,6 +2101,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -2428,6 +2203,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -2867,6 +2643,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -2968,6 +2745,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -3099,6 +2877,7 @@ items:
"min": true,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -3198,6 +2977,7 @@ items:
"min": true,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -3308,6 +3088,7 @@ items:
"min": true,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -3407,6 +3188,7 @@ items:
"min": true,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -3526,6 +3308,7 @@ items:
"min": true,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -3625,6 +3408,7 @@ items:
"min": true,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -3724,6 +3508,7 @@ items:
"min": true,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -3827,6 +3612,7 @@ items:
"min": true,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -4052,7 +3838,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Networking / Cluster",
"uid": "ff635a025bcfea7bc3dd4f508990a3e9",
"version": 0
@@ -4195,6 +3981,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -4299,6 +4086,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -4403,6 +4191,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -4507,6 +4296,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -4619,6 +4409,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -4723,6 +4514,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -4827,6 +4619,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -4918,6 +4711,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -5009,6 +4803,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -5169,7 +4964,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Controller Manager",
"uid": "72e0e05bef5099e5f049b05fdc429ed4",
"version": 0
@@ -5211,6 +5006,7 @@ items:
"fill": 1,
"format": "percentunit",
"id": 1,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -6019,7 +5815,7 @@ items:
],
"targets": [
{
- "expr": "count(mixin_pod_workload{cluster=\"$cluster\"}) by (namespace)",
+ "expr": "sum(kube_pod_owner{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6438,7 +6234,7 @@ items:
],
"targets": [
{
- "expr": "count(mixin_pod_workload{cluster=\"$cluster\"}) by (namespace)",
+ "expr": "sum(kube_pod_owner{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6564,6 +6360,7 @@ items:
"datasource": "$datasource",
"fill": 1,
"id": 11,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -7652,33 +7449,6 @@ items:
"regex": "",
"type": "datasource"
},
- {
- "allValue": null,
- "current": {
- "text": "prod",
- "value": "prod"
- },
- "datasource": "$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
- },
{
"allValue": null,
"current": {
@@ -7737,7 +7507,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Compute Resources / Cluster",
"uid": "efa86fd1d0c121a26444b636a3f509a8",
"version": 0
@@ -9007,6 +8777,7 @@ items:
"datasource": "$datasource",
"fill": 1,
"id": 9,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -9984,7 +9755,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Compute Resources / Namespace (Pods)",
"uid": "85a562078cdf77779eaa1add43ccec1e",
"version": 0
@@ -10053,7 +9824,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=~\"$node\"}) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -10282,7 +10053,7 @@ items:
],
"targets": [
{
- "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10291,7 +10062,7 @@ items:
"step": 10
},
{
- "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10300,7 +10071,7 @@ items:
"step": 10
},
{
- "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=\"$node\"}) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=~\"$node\"}) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10309,7 +10080,7 @@ items:
"step": 10
},
{
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10318,7 +10089,7 @@ items:
"step": 10
},
{
- "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=\"$node\"}) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=~\"$node\"}) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10418,7 +10189,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\", container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\", container!=\"\"}) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -10701,7 +10472,7 @@ items:
],
"targets": [
{
- "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10710,7 +10481,7 @@ items:
"step": 10
},
{
- "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10719,7 +10490,7 @@ items:
"step": 10
},
{
- "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{node=\"$node\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10728,7 +10499,7 @@ items:
"step": 10
},
{
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10737,7 +10508,7 @@ items:
"step": 10
},
{
- "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{node=\"$node\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{node=~\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10746,7 +10517,7 @@ items:
"step": 10
},
{
- "expr": "sum(node_namespace_pod_container:container_memory_rss{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_rss{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10755,7 +10526,7 @@ items:
"step": 10
},
{
- "expr": "sum(node_namespace_pod_container:container_memory_cache{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_cache{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10764,7 +10535,7 @@ items:
"step": 10
},
{
- "expr": "sum(node_namespace_pod_container:container_memory_swap{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_swap{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -10883,7 +10654,7 @@ items:
"hide": 0,
"includeAll": false,
"label": null,
- "multi": false,
+ "multi": true,
"name": "node",
"options": [
@@ -10931,7 +10702,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Compute Resources / Node (Pods)",
"uid": "200ac8fdbfbb74b39aff88118e4d1c2c",
"version": 0
@@ -11956,6 +11727,7 @@ items:
"datasource": "$datasource",
"fill": 10,
"id": 6,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -12054,6 +11826,7 @@ items:
"datasource": "$datasource",
"fill": 10,
"id": 7,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -12152,6 +11925,7 @@ items:
"datasource": "$datasource",
"fill": 10,
"id": 8,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -12250,6 +12024,7 @@ items:
"datasource": "$datasource",
"fill": 10,
"id": 9,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -12348,6 +12123,7 @@ items:
"datasource": "$datasource",
"fill": 10,
"id": 10,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -12446,6 +12222,7 @@ items:
"datasource": "$datasource",
"fill": 10,
"id": 11,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -12666,7 +12443,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Compute Resources / Pod",
"uid": "6581e46e4e5c7ba40a07646395ef7b23",
"version": 0
@@ -13437,6 +13214,7 @@ items:
"datasource": "$datasource",
"fill": 1,
"id": 5,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -14664,7 +14442,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Compute Resources / Workload",
"uid": "a164a7f0339f99e89cea5cb47e9be617",
"version": 0
@@ -15595,6 +15373,7 @@ items:
"datasource": "$datasource",
"fill": 1,
"id": 5,
+ "interval": "1m",
"legend": {
"avg": false,
"current": false,
@@ -16621,7 +16400,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__interval])\n* on (namespace,pod) \ngroup_left(workload,workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n",
+ "expr": "(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{workload}}",
@@ -16818,7 +16597,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Compute Resources / Namespace (Workloads)",
"uid": "a87fb0d919ec0ea5f6543124e16c42a5",
"version": 0
@@ -17394,6 +17173,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -17485,6 +17265,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -17589,6 +17370,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -17693,6 +17475,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -17791,6 +17574,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -17904,6 +17688,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -17997,6 +17782,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -18103,6 +17889,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -18207,6 +17994,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -18298,6 +18086,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -18403,6 +18192,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -18494,6 +18284,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -18598,6 +18389,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -18702,6 +18494,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -18827,6 +18620,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -18931,6 +18725,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -19022,6 +18817,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -19113,6 +18909,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -19299,7 +19096,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Kubelet",
"uid": "3138fa155d5915769fbded898ac09fd9",
"version": 0
@@ -19927,6 +19724,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -20026,6 +19824,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -20136,6 +19935,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -20235,6 +20035,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -20354,6 +20155,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -20453,6 +20255,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -20706,7 +20509,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Networking / Namespace (Pods)",
"uid": "8b7a8b326d7a6f1f04244066368c67af",
"version": 0
@@ -20794,6 +20597,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -20895,6 +20699,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -21334,6 +21139,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -21435,6 +21241,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -21566,6 +21373,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -21665,6 +21473,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -21775,6 +21584,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -21874,6 +21684,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -21993,6 +21804,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -22092,6 +21904,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -22377,7 +22190,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Networking / Namespace (Workload)",
"uid": "bbb2a765a623ae38130206c7d94a160f",
"version": 0
@@ -23325,7 +23138,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "USE Method / Cluster",
"uid": "3e97d1d02672cdd0861f4c97c64f89b2",
"version": 0
@@ -24300,7 +24113,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "USE Method / Node",
"uid": "fac67cfbe174d3ef53eb473d73d9212f",
"version": 0
@@ -24359,6 +24172,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -24451,6 +24265,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -24576,6 +24391,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -24785,6 +24601,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -24900,6 +24717,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -25018,6 +24836,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -25110,6 +24929,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -25271,7 +25091,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Nodes",
"uid": "fa49a4706d07a042595b664c87fb33ea",
"version": 0
@@ -25330,6 +25150,7 @@ items:
"min": true,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -25525,6 +25346,7 @@ items:
"min": true,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -25828,7 +25650,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Persistent Volumes",
"uid": "919b92a8e8041bd567af9edab12c840c",
"version": 0
@@ -26191,6 +26013,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -26290,6 +26113,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -26400,6 +26224,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -26499,6 +26324,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -26618,6 +26444,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -26717,6 +26544,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27002,7 +26830,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Networking / Pod",
"uid": "7a18067ce943a40ae25454675c19ff5c",
"version": 0
@@ -27061,6 +26889,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27152,6 +26981,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27256,6 +27086,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27360,6 +27191,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27452,6 +27284,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27543,6 +27376,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27634,6 +27468,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27738,6 +27573,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27829,6 +27665,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -27933,6 +27770,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -28024,6 +27862,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -28128,6 +27967,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -28219,6 +28059,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -28310,6 +28151,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -28401,6 +28243,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -29975,6 +29818,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -30066,6 +29910,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -30170,6 +30015,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -30261,6 +30107,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -30365,6 +30212,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -30477,6 +30325,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -30581,6 +30430,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -30685,6 +30535,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -30776,6 +30627,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -30867,6 +30719,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -31027,7 +30880,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Proxy",
"uid": "632e265de029684c40b21cb76bca4f94",
"version": 0
@@ -31170,6 +31023,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -31282,6 +31136,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -31407,6 +31262,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -31519,6 +31375,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -31623,6 +31480,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"total": false,
"values": true
},
@@ -31727,6 +31585,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -31818,6 +31677,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -31909,6 +31769,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -32069,7 +31930,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Scheduler",
"uid": "2e6b6a3b4bddf1427b3a55aa1311c656",
"version": 0
@@ -32740,6 +32601,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -32979,7 +32841,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / StatefulSets",
"uid": "a31c1f46e6f727cb37c0d731a7245005",
"version": 0
@@ -33067,6 +32929,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -33168,6 +33031,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -33280,6 +33144,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -33381,6 +33246,7 @@ items:
"min": false,
"rightSide": true,
"show": true,
+ "sideWidth": null,
"sort": "current",
"sortDesc": true,
"total": false,
@@ -33512,6 +33378,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -33611,6 +33478,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -33721,6 +33589,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -33820,6 +33689,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -33939,6 +33809,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -34038,6 +33909,7 @@ items:
"min": false,
"rightSide": false,
"show": true,
+ "sideWidth": null,
"total": false,
"values": false
},
@@ -34355,7 +34227,7 @@ items:
"30d"
]
},
- "timezone": "",
+ "timezone": "UTC",
"title": "Kubernetes / Networking / Workload",
"uid": "728bf77cc1166d2f3133bf25846876cc",
"version": 0
diff --git a/manifests/prometheus-operator-serviceMonitor.yaml b/manifests/prometheus-operator-serviceMonitor.yaml
index 3d7f7e572cbada5c4ec639d1f474a3b9320045fc..39e48aa775489eea6bd1daf3f1656328377bc4d7 100644
--- a/manifests/prometheus-operator-serviceMonitor.yaml
+++ b/manifests/prometheus-operator-serviceMonitor.yaml
@@ -4,7 +4,7 @@ metadata:
labels:
app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator
- app.kubernetes.io/version: v0.39.0
+ app.kubernetes.io/version: v0.40.0
name: prometheus-operator
namespace: monitoring
spec:
@@ -19,4 +19,4 @@ spec:
matchLabels:
app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator
- app.kubernetes.io/version: v0.39.0
+ app.kubernetes.io/version: v0.40.0
diff --git a/manifests/prometheus-rules.yaml b/manifests/prometheus-rules.yaml
index 34b445e801944d9ed9055ec0d5ee272fadb9a857..141fc8ab15568415515838a17c74ff00d1c35610 100644
--- a/manifests/prometheus-rules.yaml
+++ b/manifests/prometheus-rules.yaml
@@ -74,7 +74,7 @@ spec:
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[1d]))
-
(
- sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="resource",le="0.1"}[1d])) +
+ sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[1d])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[1d])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[1d]))
)
@@ -95,7 +95,7 @@ spec:
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[1h]))
-
(
- sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="resource",le="0.1"}[1h])) +
+ sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[1h])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[1h])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[1h]))
)
@@ -116,7 +116,7 @@ spec:
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[2h]))
-
(
- sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="resource",le="0.1"}[2h])) +
+ sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[2h])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[2h])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[2h]))
)
@@ -137,7 +137,7 @@ spec:
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[30m]))
-
(
- sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="resource",le="0.1"}[30m])) +
+ sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30m])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[30m])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[30m]))
)
@@ -158,7 +158,7 @@ spec:
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[3d]))
-
(
- sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="resource",le="0.1"}[3d])) +
+ sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[3d])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[3d])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[3d]))
)
@@ -179,7 +179,7 @@ spec:
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[5m]))
-
(
- sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="resource",le="0.1"}[5m])) +
+ sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[5m])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[5m])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[5m]))
)
@@ -200,7 +200,7 @@ spec:
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[6h]))
-
(
- sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="resource",le="0.1"}[6h])) +
+ sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[6h])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[6h])) +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[6h]))
)
@@ -326,6 +326,51 @@ spec:
labels:
verb: write
record: apiserver_request:burnrate6h
+ - expr: |
+ sum by (code,resource) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[5m]))
+ labels:
+ verb: read
+ record: code_resource:apiserver_request_total:rate5m
+ - expr: |
+ sum by (code,resource) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))
+ labels:
+ verb: write
+ record: code_resource:apiserver_request_total:rate5m
+ - expr: |
+ histogram_quantile(0.99, sum by (le, resource) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET"}[5m]))) > 0
+ labels:
+ quantile: "0.99"
+ verb: read
+ record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ - expr: |
+ histogram_quantile(0.99, sum by (le, resource) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))) > 0
+ labels:
+ quantile: "0.99"
+ verb: write
+ record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ - expr: |
+ sum(rate(apiserver_request_duration_seconds_sum{subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod)
+ /
+ sum(rate(apiserver_request_duration_seconds_count{subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod)
+ record: cluster:apiserver_request_duration_seconds:mean5m
+ - expr: |
+ histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod))
+ labels:
+ quantile: "0.99"
+ record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ - expr: |
+ histogram_quantile(0.9, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod))
+ labels:
+ quantile: "0.9"
+ record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ - expr: |
+ histogram_quantile(0.5, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod))
+ labels:
+ quantile: "0.5"
+ record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ - interval: 3m
+ name: kube-apiserver-availability.rules
+ rules:
- expr: |
1 - (
(
@@ -339,13 +384,13 @@ spec:
sum(increase(apiserver_request_duration_seconds_count{verb=~"LIST|GET"}[30d]))
-
(
- sum(increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope="resource",le="0.1"}[30d])) +
+ sum(increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30d])) +
sum(increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope="namespace",le="0.5"}[30d])) +
sum(increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope="cluster",le="5"}[30d]))
)
) +
# errors
- sum(code:apiserver_request_total:increase30d{code=~"5.."})
+ sum(code:apiserver_request_total:increase30d{code=~"5.."} or vector(0))
)
/
sum(code:apiserver_request_total:increase30d)
@@ -358,13 +403,13 @@ spec:
-
(
# too slow
- sum(increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="resource",le="0.1"}[30d])) +
+ sum(increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30d])) +
sum(increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[30d])) +
sum(increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[30d]))
)
+
# errors
- sum(code:apiserver_request_total:increase30d{verb="read",code=~"5.."})
+ sum(code:apiserver_request_total:increase30d{verb="read",code=~"5.."} or vector(0))
)
/
sum(code:apiserver_request_total:increase30d{verb="read"})
@@ -381,7 +426,7 @@ spec:
)
+
# errors
- sum(code:apiserver_request_total:increase30d{verb="write",code=~"5.."})
+ sum(code:apiserver_request_total:increase30d{verb="write",code=~"5.."} or vector(0))
)
/
sum(code:apiserver_request_total:increase30d{verb="write"})
@@ -389,60 +434,87 @@ spec:
verb: write
record: apiserver_request:availability30d
- expr: |
- sum by (code, verb) (increase(apiserver_request_total{job="apiserver"}[30d]))
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"2.."}[30d]))
record: code_verb:apiserver_request_total:increase30d
- expr: |
- sum by (code) (code_verb:apiserver_request_total:increase30d{verb=~"LIST|GET"})
- labels:
- verb: read
- record: code:apiserver_request_total:increase30d
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"2.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
- expr: |
- sum by (code) (code_verb:apiserver_request_total:increase30d{verb=~"POST|PUT|PATCH|DELETE"})
- labels:
- verb: write
- record: code:apiserver_request_total:increase30d
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"2.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
- expr: |
- sum by (code,resource) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[5m]))
- labels:
- verb: read
- record: code_resource:apiserver_request_total:rate5m
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"2.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
- expr: |
- sum by (code,resource) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))
- labels:
- verb: write
- record: code_resource:apiserver_request_total:rate5m
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"2.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
- expr: |
- histogram_quantile(0.99, sum by (le, resource) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET"}[5m]))) > 0
- labels:
- quantile: "0.99"
- verb: read
- record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"2.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
- expr: |
- histogram_quantile(0.99, sum by (le, resource) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))) > 0
- labels:
- quantile: "0.99"
- verb: write
- record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"3.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
- expr: |
- sum(rate(apiserver_request_duration_seconds_sum{subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod)
- /
- sum(rate(apiserver_request_duration_seconds_count{subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod)
- record: cluster:apiserver_request_duration_seconds:mean5m
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"3.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
- expr: |
- histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod))
- labels:
- quantile: "0.99"
- record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"3.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
- expr: |
- histogram_quantile(0.9, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod))
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"3.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"3.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"3.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"4.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"4.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"4.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"4.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"4.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"4.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"5.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"5.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"5.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"5.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"5.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"5.."}[30d]))
+ record: code_verb:apiserver_request_total:increase30d
+ - expr: |
+ sum by (code) (code_verb:apiserver_request_total:increase30d{verb=~"LIST|GET"})
labels:
- quantile: "0.9"
- record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ verb: read
+ record: code:apiserver_request_total:increase30d
- expr: |
- histogram_quantile(0.5, sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",subresource!="log",verb!~"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT"}[5m])) without(instance, pod))
+ sum by (code) (code_verb:apiserver_request_total:increase30d{verb=~"POST|PUT|PATCH|DELETE"})
labels:
- quantile: "0.5"
- record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
+ verb: write
+ record: code:apiserver_request_total:increase30d
- name: k8s.rules
rules:
- expr: |
@@ -452,31 +524,31 @@ spec:
sum by (cluster, namespace, pod, container) (
rate(container_cpu_usage_seconds_total{job="kubelet", metrics_path="/metrics/cadvisor", image!="", container!="POD"}[5m])
) * on (cluster, namespace, pod) group_left(node) topk by (cluster, namespace, pod) (
- 1, max by(cluster, namespace, pod, node) (kube_pod_info)
+ 1, max by(cluster, namespace, pod, node) (kube_pod_info{node!=""})
)
record: node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate
- expr: |
container_memory_working_set_bytes{job="kubelet", metrics_path="/metrics/cadvisor", image!=""}
* on (namespace, pod) group_left(node) topk by(namespace, pod) (1,
- max by(namespace, pod, node) (kube_pod_info)
+ max by(namespace, pod, node) (kube_pod_info{node!=""})
)
record: node_namespace_pod_container:container_memory_working_set_bytes
- expr: |
container_memory_rss{job="kubelet", metrics_path="/metrics/cadvisor", image!=""}
* on (namespace, pod) group_left(node) topk by(namespace, pod) (1,
- max by(namespace, pod, node) (kube_pod_info)
+ max by(namespace, pod, node) (kube_pod_info{node!=""})
)
record: node_namespace_pod_container:container_memory_rss
- expr: |
container_memory_cache{job="kubelet", metrics_path="/metrics/cadvisor", image!=""}
* on (namespace, pod) group_left(node) topk by(namespace, pod) (1,
- max by(namespace, pod, node) (kube_pod_info)
+ max by(namespace, pod, node) (kube_pod_info{node!=""})
)
record: node_namespace_pod_container:container_memory_cache
- expr: |
container_memory_swap{job="kubelet", metrics_path="/metrics/cadvisor", image!=""}
* on (namespace, pod) group_left(node) topk by(namespace, pod) (1,
- max by(namespace, pod, node) (kube_pod_info)
+ max by(namespace, pod, node) (kube_pod_info{node!=""})
)
record: node_namespace_pod_container:container_memory_swap
- expr: |
@@ -591,12 +663,12 @@ spec:
- name: node.rules
rules:
- expr: |
- sum(min(kube_pod_info) by (cluster, node))
+ sum(min(kube_pod_info{node!=""}) by (cluster, node))
record: ':kube_pod_info_node_count:'
- expr: |
topk by(namespace, pod) (1,
max by (node, namespace, pod) (
- label_replace(kube_pod_info{job="kube-state-metrics"}, "pod", "$1", "pod", "(.*)")
+ label_replace(kube_pod_info{job="kube-state-metrics",node!=""}, "pod", "$1", "pod", "(.*)")
))
record: 'node_namespace_pod:kube_pod_info:'
- expr: |
@@ -849,13 +921,22 @@ spec:
severity: warning
- alert: NodeHighNumberConntrackEntriesUsed
annotations:
- description: '{{ $value | humanizePercentage }} of conntrack entries are used'
+ description: '{{ $value | humanizePercentage }} of conntrack entries are used.'
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-nodehighnumberconntrackentriesused
- summary: Number of conntrack are getting close to the limit
+ summary: Number of conntrack are getting close to the limit.
expr: |
(node_nf_conntrack_entries / node_nf_conntrack_entries_limit) > 0.75
labels:
severity: warning
+ - alert: NodeTextFileCollectorScrapeError
+ annotations:
+ description: Node Exporter text file collector failed to scrape.
+ runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-nodetextfilecollectorscrapeerror
+ summary: Node Exporter text file collector failed to scrape.
+ expr: |
+ node_textfile_scrape_error{job="node-exporter"} == 1
+ labels:
+ severity: warning
- alert: NodeClockSkewDetected
annotations:
message: Clock on {{ $labels.instance }} is out of sync by more than 300s.
@@ -896,20 +977,26 @@ spec:
}}) is restarting {{ printf "%.2f" $value }} times / 5 minutes.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubepodcrashlooping
expr: |
- rate(kube_pod_container_status_restarts_total{job="kube-state-metrics"}[15m]) * 60 * 5 > 0
+ rate(kube_pod_container_status_restarts_total{job="kube-state-metrics"}[5m]) * 60 * 5 > 0
for: 15m
labels:
- severity: critical
+ severity: warning
- alert: KubePodNotReady
annotations:
message: Pod {{ $labels.namespace }}/{{ $labels.pod }} has been in a non-ready
state for longer than 15 minutes.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubepodnotready
expr: |
- sum by (namespace, pod) (max by(namespace, pod) (kube_pod_status_phase{job="kube-state-metrics", phase=~"Pending|Unknown"}) * on(namespace, pod) group_left(owner_kind) max by(namespace, pod, owner_kind) (kube_pod_owner{owner_kind!="Job"})) > 0
+ sum by (namespace, pod) (
+ max by(namespace, pod) (
+ kube_pod_status_phase{job="kube-state-metrics", phase=~"Pending|Unknown"}
+ ) * on(namespace, pod) group_left(owner_kind) topk by(namespace, pod) (
+ 1, max by(namespace, pod, owner_kind) (kube_pod_owner{owner_kind!="Job"})
+ )
+ ) > 0
for: 15m
labels:
- severity: critical
+ severity: warning
- alert: KubeDeploymentGenerationMismatch
annotations:
message: Deployment generation for {{ $labels.namespace }}/{{ $labels.deployment
@@ -922,7 +1009,7 @@ spec:
kube_deployment_metadata_generation{job="kube-state-metrics"}
for: 15m
labels:
- severity: critical
+ severity: warning
- alert: KubeDeploymentReplicasMismatch
annotations:
message: Deployment {{ $labels.namespace }}/{{ $labels.deployment }} has not
@@ -940,7 +1027,7 @@ spec:
)
for: 15m
labels:
- severity: critical
+ severity: warning
- alert: KubeStatefulSetReplicasMismatch
annotations:
message: StatefulSet {{ $labels.namespace }}/{{ $labels.statefulset }} has
@@ -958,7 +1045,7 @@ spec:
)
for: 15m
labels:
- severity: critical
+ severity: warning
- alert: KubeStatefulSetGenerationMismatch
annotations:
message: StatefulSet generation for {{ $labels.namespace }}/{{ $labels.statefulset
@@ -971,27 +1058,33 @@ spec:
kube_statefulset_metadata_generation{job="kube-state-metrics"}
for: 15m
labels:
- severity: critical
+ severity: warning
- alert: KubeStatefulSetUpdateNotRolledOut
annotations:
message: StatefulSet {{ $labels.namespace }}/{{ $labels.statefulset }} update
has not been rolled out.
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubestatefulsetupdatenotrolledout
expr: |
- max without (revision) (
- kube_statefulset_status_current_revision{job="kube-state-metrics"}
- unless
- kube_statefulset_status_update_revision{job="kube-state-metrics"}
- )
- *
(
- kube_statefulset_replicas{job="kube-state-metrics"}
- !=
- kube_statefulset_status_replicas_updated{job="kube-state-metrics"}
+ max without (revision) (
+ kube_statefulset_status_current_revision{job="kube-state-metrics"}
+ unless
+ kube_statefulset_status_update_revision{job="kube-state-metrics"}
+ )
+ *
+ (
+ kube_statefulset_replicas{job="kube-state-metrics"}
+ !=
+ kube_statefulset_status_replicas_updated{job="kube-state-metrics"}
+ )
+ ) and (
+ changes(kube_statefulset_status_replicas_updated{job="kube-state-metrics"}[5m])
+ ==
+ 0
)
for: 15m
labels:
- severity: critical
+ severity: warning
- alert: KubeDaemonSetRolloutStuck
annotations:
message: Only {{ $value | humanizePercentage }} of the desired Pods of DaemonSet
@@ -1003,7 +1096,7 @@ spec:
kube_daemonset_status_desired_number_scheduled{job="kube-state-metrics"} < 1.00
for: 15m
labels:
- severity: critical
+ severity: warning
- alert: KubeContainerWaiting
annotations:
message: Pod {{ $labels.namespace }}/{{ $labels.pod }} container {{ $labels.container}}
@@ -1254,7 +1347,9 @@ spec:
sum(apiserver_request:burnrate5m) > (14.40 * 0.01000)
for: 2m
labels:
+ long: 1h
severity: critical
+ short: 5m
- alert: KubeAPIErrorBudgetBurn
annotations:
message: The API server is burning too much error budget
@@ -1265,7 +1360,9 @@ spec:
sum(apiserver_request:burnrate30m) > (6.00 * 0.01000)
for: 15m
labels:
+ long: 6h
severity: critical
+ short: 30m
- alert: KubeAPIErrorBudgetBurn
annotations:
message: The API server is burning too much error budget
@@ -1276,7 +1373,9 @@ spec:
sum(apiserver_request:burnrate2h) > (3.00 * 0.01000)
for: 1h
labels:
+ long: 1d
severity: warning
+ short: 2h
- alert: KubeAPIErrorBudgetBurn
annotations:
message: The API server is burning too much error budget
@@ -1287,7 +1386,9 @@ spec:
sum(apiserver_request:burnrate6h) > (1.00 * 0.01000)
for: 3h
labels:
+ long: 3d
severity: warning
+ short: 6h
- name: kubernetes-system-apiserver
rules:
- alert: KubeAPILatencyHigh
@@ -1296,6 +1397,10 @@ spec:
{{ $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_duration_seconds:histogram_quantile{job="apiserver",quantile="0.99"}
+ >
+ 1
+ and on (verb,resource)
(
cluster:apiserver_request_duration_seconds:mean5m{job="apiserver"}
>
@@ -1307,10 +1412,6 @@ spec:
)
) > on (verb) group_left()
1.2 * avg by (verb) (cluster:apiserver_request_duration_seconds:mean5m{job="apiserver"} >= 0)
- and on (verb,resource)
- cluster_quantile:apiserver_request_duration_seconds:histogram_quantile{job="apiserver",quantile="0.99"}
- >
- 1
for: 5m
labels:
severity: warning
@@ -1391,8 +1492,7 @@ spec:
message: '{{ $labels.node }} is unreachable and some workloads may be rescheduled.'
runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubenodeunreachable
expr: |
- kube_node_spec_taint{job="kube-state-metrics",key="node.kubernetes.io/unreachable",effect="NoSchedule"} == 1
- for: 2m
+ (kube_node_spec_taint{job="kube-state-metrics",key="node.kubernetes.io/unreachable",effect="NoSchedule"} unless ignoring(key,value) kube_node_spec_taint{job="kube-state-metrics",key="ToBeDeletedByClusterAutoscaler"}) == 1
labels:
severity: warning
- alert: KubeletTooManyPods
diff --git a/manifests/setup/prometheus-operator-0prometheusCustomResourceDefinition.yaml b/manifests/setup/prometheus-operator-0prometheusCustomResourceDefinition.yaml
index d391b6b4e2f763d6edb1b7b1d9f198f71b38ec10..3c2126c834b4ec8d49c80ea197ebe65ef53af392 100644
--- a/manifests/setup/prometheus-operator-0prometheusCustomResourceDefinition.yaml
+++ b/manifests/setup/prometheus-operator-0prometheusCustomResourceDefinition.yaml
@@ -2177,6 +2177,15 @@ spec:
of origin for each alert and metric that is user created. The label
value will always be the namespace of the object that is being created.
type: string
+ enforcedSampleLimit:
+ description: EnforcedSampleLimit defines global limit on number of
+ scraped samples that will be accepted. This overrides any SampleLimit
+ set per ServiceMonitor or/and PodMonitor. It is meant to be used
+ by admins to enforce the SampleLimit to keep overall number of samples/series
+ under the desired limit. Note that if SampleLimit is lower that
+ value will be taken instead.
+ format: int64
+ type: integer
evaluationInterval:
description: Interval between consecutive evaluations.
type: string
@@ -3428,6 +3437,27 @@ spec:
instance name. Defaults to the value of `prometheus`. External label
will _not_ be added when value is set to empty string (`""`).
type: string
+ prometheusRulesExcludedFromEnforce:
+ description: PrometheusRulesExcludedFromEnforce - list of prometheus
+ rules to be excluded from enforcing of adding namespace labels.
+ Works only if enforcedNamespaceLabel set to true. Make sure both
+ ruleNamespace and ruleName are set for each pair
+ items:
+ description: PrometheusRuleExcludeConfig enables users to configure
+ excluded PrometheusRule names and their namespaces to be ignored
+ while enforcing namespace label for alerts and metrics.
+ properties:
+ ruleName:
+ description: RuleNamespace - name of excluded rule
+ type: string
+ ruleNamespace:
+ description: RuleNamespace - namespace of excluded rule
+ type: string
+ required:
+ - ruleName
+ - ruleNamespace
+ type: object
+ type: array
query:
description: QuerySpec defines the query command line flags when starting
Prometheus.
@@ -4114,6 +4144,10 @@ spec:
scrapeInterval:
description: Interval between consecutive scrapes.
type: string
+ scrapeTimeout:
+ description: Number of seconds to wait for target to respond before
+ erroring.
+ type: string
secrets:
description: Secrets is a list of Secrets in the same namespace as
the Prometheus object, which shall be mounted into the Prometheus
@@ -4762,6 +4796,12 @@ spec:
logLevel:
description: LogLevel for Thanos sidecar to be configured with.
type: string
+ minTime:
+ description: MinTime for Thanos sidecar to be configured with.
+ Option can be a constant time in RFC3339 format or time duration
+ relative to current time, such as -1d or 2h45m. Valid duration
+ units are ms, s, m, h, d, w, y.
+ type: string
objectStorageConfig:
description: ObjectStorageConfig configures object storage in
Thanos.
diff --git a/manifests/setup/prometheus-operator-0thanosrulerCustomResourceDefinition.yaml b/manifests/setup/prometheus-operator-0thanosrulerCustomResourceDefinition.yaml
index 3396c98885617a43fd3477dd1f1c67c4d537a32b..74db7b3a8e4206672cf367e21b494b836a47ee66 100644
--- a/manifests/setup/prometheus-operator-0thanosrulerCustomResourceDefinition.yaml
+++ b/manifests/setup/prometheus-operator-0thanosrulerCustomResourceDefinition.yaml
@@ -2998,6 +2998,27 @@ spec:
priorityClassName:
description: Priority class assigned to the Pods
type: string
+ prometheusRulesExcludedFromEnforce:
+ description: PrometheusRulesExcludedFromEnforce - list of Prometheus
+ rules to be excluded from enforcing of adding namespace labels.
+ Works only if enforcedNamespaceLabel set to true. Make sure both
+ ruleNamespace and ruleName are set for each pair
+ items:
+ description: PrometheusRuleExcludeConfig enables users to configure
+ excluded PrometheusRule names and their namespaces to be ignored
+ while enforcing namespace label for alerts and metrics.
+ properties:
+ ruleName:
+ description: RuleNamespace - name of excluded rule
+ type: string
+ ruleNamespace:
+ description: RuleNamespace - namespace of excluded rule
+ type: string
+ required:
+ - ruleName
+ - ruleNamespace
+ type: object
+ type: array
queryConfig:
description: Define configuration for connecting to thanos query instances.
If this is defined, the QueryEndpoints field will be ignored. Maps
diff --git a/manifests/setup/prometheus-operator-clusterRole.yaml b/manifests/setup/prometheus-operator-clusterRole.yaml
index 6afaa774935ff4ded6bf773ddd4cd83d9961c9d4..b3d490a797638d31def47bd34f7d653b7575315d 100644
--- a/manifests/setup/prometheus-operator-clusterRole.yaml
+++ b/manifests/setup/prometheus-operator-clusterRole.yaml
@@ -4,7 +4,7 @@ metadata:
labels:
app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator
- app.kubernetes.io/version: v0.39.0
+ app.kubernetes.io/version: v0.40.0
name: prometheus-operator
rules:
- apiGroups:
diff --git a/manifests/setup/prometheus-operator-clusterRoleBinding.yaml b/manifests/setup/prometheus-operator-clusterRoleBinding.yaml
index c493a746c20718ed5629c5033289d8f01cc1aeaf..5ac1066ff657e96d015e1e8e008627b72812cf3d 100644
--- a/manifests/setup/prometheus-operator-clusterRoleBinding.yaml
+++ b/manifests/setup/prometheus-operator-clusterRoleBinding.yaml
@@ -4,7 +4,7 @@ metadata:
labels:
app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator
- app.kubernetes.io/version: v0.39.0
+ app.kubernetes.io/version: v0.40.0
name: prometheus-operator
roleRef:
apiGroup: rbac.authorization.k8s.io
diff --git a/manifests/setup/prometheus-operator-deployment.yaml b/manifests/setup/prometheus-operator-deployment.yaml
index 1bfb05c4c56961f5f844cd60577ed53856882f58..c00043183a8667004a69d390e299e601b9e7e7a2 100644
--- a/manifests/setup/prometheus-operator-deployment.yaml
+++ b/manifests/setup/prometheus-operator-deployment.yaml
@@ -4,7 +4,7 @@ metadata:
labels:
app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator
- app.kubernetes.io/version: v0.39.0
+ app.kubernetes.io/version: v0.40.0
name: prometheus-operator
namespace: monitoring
spec:
@@ -18,15 +18,15 @@ spec:
labels:
app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator
- app.kubernetes.io/version: v0.39.0
+ app.kubernetes.io/version: v0.40.0
spec:
containers:
- args:
- --kubelet-service=kube-system/kubelet
- --logtostderr=true
- --config-reloader-image=jimmidyson/configmap-reload:v0.3.0
- - --prometheus-config-reloader=quay.io/coreos/prometheus-config-reloader:v0.39.0
- image: quay.io/coreos/prometheus-operator:v0.39.0
+ - --prometheus-config-reloader=quay.io/coreos/prometheus-config-reloader:v0.40.0
+ image: quay.io/coreos/prometheus-operator:v0.40.0
name: prometheus-operator
ports:
- containerPort: 8080
diff --git a/manifests/setup/prometheus-operator-service.yaml b/manifests/setup/prometheus-operator-service.yaml
index c3c3debb2cfbf7aa4b123931fd2d5c59116cf0d7..3f8dd96e04332ed17f61331d1d27c87b0a96c345 100644
--- a/manifests/setup/prometheus-operator-service.yaml
+++ b/manifests/setup/prometheus-operator-service.yaml
@@ -4,7 +4,7 @@ metadata:
labels:
app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator
- app.kubernetes.io/version: v0.39.0
+ app.kubernetes.io/version: v0.40.0
name: prometheus-operator
namespace: monitoring
spec:
diff --git a/manifests/setup/prometheus-operator-serviceAccount.yaml b/manifests/setup/prometheus-operator-serviceAccount.yaml
index 39d14d95e3e285a320654ab69d25199767178b9a..c0bffc96cd258c3857773b0e6bc695966dfe9b71 100644
--- a/manifests/setup/prometheus-operator-serviceAccount.yaml
+++ b/manifests/setup/prometheus-operator-serviceAccount.yaml
@@ -4,6 +4,6 @@ metadata:
labels:
app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator
- app.kubernetes.io/version: v0.39.0
+ app.kubernetes.io/version: v0.40.0
name: prometheus-operator
namespace: monitoring