From ace8b4d6d3c72602e463af99c0bcff9eec61a49f Mon Sep 17 00:00:00 2001
From: liuxu <liuxu623@gmail.com>
Date: Wed, 16 Oct 2019 16:49:33 +0800
Subject: [PATCH] add k8s-resources-node dashboard
---
jsonnetfile.lock.json | 4 +-
manifests/grafana-dashboardDefinitions.yaml | 987 +++++++++++++++++++-
manifests/grafana-deployment.yaml | 6 +
manifests/prometheus-rules.yaml | 4 +-
4 files changed, 977 insertions(+), 24 deletions(-)
diff --git a/jsonnetfile.lock.json b/jsonnetfile.lock.json
index 1e3eff81..d5433547 100644
--- a/jsonnetfile.lock.json
+++ b/jsonnetfile.lock.json
@@ -27,7 +27,7 @@
"subdir": ""
}
},
- "version": "3da19e5a40fbb9f7a621958040472db918c4de9c"
+ "version": "15ddfa20a6921ffbd43172eb54f6bdc1bcf8d3d3"
},
{
"name": "grafonnet",
@@ -107,7 +107,7 @@
"subdir": "lib/promgrafonnet"
}
},
- "version": "3da19e5a40fbb9f7a621958040472db918c4de9c"
+ "version": "15ddfa20a6921ffbd43172eb54f6bdc1bcf8d3d3"
}
]
}
diff --git a/manifests/grafana-dashboardDefinitions.yaml b/manifests/grafana-dashboardDefinitions.yaml
index 799c493c..1065be04 100644
--- a/manifests/grafana-dashboardDefinitions.yaml
+++ b/manifests/grafana-dashboardDefinitions.yaml
@@ -2980,7 +2980,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{namespace}}",
@@ -3263,7 +3263,7 @@ items:
"step": 10
},
{
- "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -3281,7 +3281,7 @@ items:
"step": 10
},
{
- "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)",
+ "expr": "sum(node_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,
@@ -3299,7 +3299,7 @@ items:
"step": 10
},
{
- "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)",
+ "expr": "sum(node_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,
@@ -3927,7 +3927,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -4156,7 +4156,7 @@ items:
],
"targets": [
{
- "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -4174,7 +4174,7 @@ items:
"step": 10
},
{
- "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)",
+ "expr": "sum(node_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,
@@ -4192,7 +4192,7 @@ items:
"step": 10
},
{
- "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)",
+ "expr": "sum(node_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,
@@ -4814,6 +4814,953 @@ items:
metadata:
name: grafana-dashboard-k8s-resources-namespace
namespace: monitoring
+- apiVersion: v1
+ data:
+ k8s-resources-node.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": "$datasource",
+ "fill": 10,
+ "id": 1,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [
+
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{pod}}",
+ "legendLink": null,
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Usage",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "CPU Usage",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "id": 2,
+ "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": "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": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "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(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "B",
+ "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)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "C",
+ "step": 10
+ },
+ {
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "D",
+ "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)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "E",
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Quota",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "transform": "table",
+ "type": "table",
+ "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 Quota",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 10,
+ "id": 3,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [
+
+ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\", container!=\"\"}) by (pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{pod}}",
+ "legendLink": null,
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Usage (w/o cache)",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Memory Usage",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fill": 1,
+ "id": 4,
+ "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": "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": "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"
+ },
+ {
+ "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"
+ },
+ {
+ "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"
+ },
+ {
+ "alias": "Pod",
+ "colorMode": null,
+ "colors": [
+
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "link": false,
+ "linkTooltip": "Drill down",
+ "linkUrl": "",
+ "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(container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{node=\"$node\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "C",
+ "step": 10
+ },
+ {
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", node=\"$node\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "D",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{node=\"$node\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "E",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "F",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_cache{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "G",
+ "step": 10
+ },
+ {
+ "expr": "sum(container_memory_swap{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "H",
+ "step": 10
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Quota",
+ "tooltip": {
+ "shared": false,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "transform": "table",
+ "type": "table",
+ "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": "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": [
+
+ ],
+ "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": "node",
+ "multi": false,
+ "name": "node",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, node)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 2,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Kubernetes / Compute Resources / Node (Pods)",
+ "uid": "200ac8fdbfbb74b39aff88118e4d1c2c",
+ "version": 0
+ }
+ kind: ConfigMap
+ metadata:
+ name: grafana-dashboard-k8s-resources-node
+ namespace: monitoring
- apiVersion: v1
data:
k8s-resources-pod.json: |-
@@ -4874,7 +5821,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", cluster=\"$cluster\"}) by (container)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", cluster=\"$cluster\"}) by (container)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{container}}",
@@ -5103,7 +6050,7 @@ items:
],
"targets": [
{
- "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\"}) by (container)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5121,7 +6068,7 @@ items:
"step": 10
},
{
- "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5139,7 +6086,7 @@ items:
"step": 10
},
{
- "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5864,7 +6811,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -6093,7 +7040,7 @@ items:
],
"targets": [
{
- "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6111,7 +7058,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$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=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$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=\"$type\"}\n) by (pod)\n",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6129,7 +7076,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$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=\"$type\"}\n) by (pod)\n",
+ "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$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=\"$type\"}\n) by (pod)\n",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6784,7 +7731,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{workload}} - {{workload_type}}",
@@ -7058,7 +8005,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -7076,7 +8023,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$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=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$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=\"$namespace\"}\n) by (workload, workload_type)\n",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -7094,7 +8041,7 @@ items:
"step": 10
},
{
- "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$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=\"$namespace\"}\n) by (workload, workload_type)\n",
+ "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$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,
diff --git a/manifests/grafana-deployment.yaml b/manifests/grafana-deployment.yaml
index fd01d32d..4485b165 100644
--- a/manifests/grafana-deployment.yaml
+++ b/manifests/grafana-deployment.yaml
@@ -54,6 +54,9 @@ spec:
- mountPath: /grafana-dashboard-definitions/0/k8s-resources-namespace
name: grafana-dashboard-k8s-resources-namespace
readOnly: false
+ - mountPath: /grafana-dashboard-definitions/0/k8s-resources-node
+ name: grafana-dashboard-k8s-resources-node
+ readOnly: false
- mountPath: /grafana-dashboard-definitions/0/k8s-resources-pod
name: grafana-dashboard-k8s-resources-pod
readOnly: false
@@ -123,6 +126,9 @@ spec:
- configMap:
name: grafana-dashboard-k8s-resources-namespace
name: grafana-dashboard-k8s-resources-namespace
+ - configMap:
+ name: grafana-dashboard-k8s-resources-node
+ name: grafana-dashboard-k8s-resources-node
- configMap:
name: grafana-dashboard-k8s-resources-pod
name: grafana-dashboard-k8s-resources-pod
diff --git a/manifests/prometheus-rules.yaml b/manifests/prometheus-rules.yaml
index eeb7af5a..007b3548 100644
--- a/manifests/prometheus-rules.yaml
+++ b/manifests/prometheus-rules.yaml
@@ -77,8 +77,8 @@ spec:
- expr: |
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
+ ) * on (namespace, pod) group_left(node) max by(namespace, pod, node) (kube_pod_info)
+ record: node_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
--
GitLab