diff --git a/hack/cluster-monitoring/deploy b/hack/cluster-monitoring/deploy
index 41e051876498ccf22256c52b731d2a86be297d2d..9e1b488199671eff5c1a8f6f9f5d52bb8f20db3c 100755
--- a/hack/cluster-monitoring/deploy
+++ b/hack/cluster-monitoring/deploy
@@ -1,10 +1,18 @@
#!/usr/bin/env bash
+# exit immediately when a command fails
+set -e
+# only exit with zero if all commands of the pipeline exit successfully
+set -o pipefail
+# error on unset variables
+set -u
+# print each command before executing it
+set -x
manifest_prefix=${1-.}
kubectl create namespace monitoring
-kubectl apply -f ${manifest_prefix}/manifests/prometheus-operator/
+find ${manifest_prefix}/manifests/prometheus-operator/ -type f ! -name prometheus-operator-service-monitor.yaml -exec kubectl apply -f {} \;
# Wait for CRDs to be ready.
printf "Waiting for Operator to register custom resource definitions..."
@@ -16,9 +24,15 @@ until kubectl get prometheuses.monitoring.coreos.com > /dev/null 2>&1; do sleep
until kubectl get alertmanagers.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
echo "done!"
+# need to ensure that ServiceMonitors are registered before we can create the prometheus-operator ServiceMonitor
+kubectl apply -f ${manifest_prefix}/manifests/prometheus-operator/prometheus-operator-service-monitor.yaml
+
kubectl apply -f ${manifest_prefix}/manifests/node-exporter/
kubectl apply -f ${manifest_prefix}/manifests/kube-state-metrics/
-kubectl apply -f ${manifest_prefix}/manifests/grafana/
+find ${manifest_prefix}/manifests/grafana/ -type f ! -name grafana-dashboard-definitions.yaml -exec kubectl apply -f {} \;
+
+# kubectl apply wants to put the previous version in an annotation, which is too large, therefore create instead of apply
+kubectl create -f ${manifest_prefix}/manifests/grafana/grafana-dashboard-definitions.yaml
kubectl apply -f ${manifest_prefix}/manifests/prometheus-k8s/
kubectl apply -f ${manifest_prefix}/manifests/alertmanager-main/
diff --git a/jsonnet/prometheus/prometheus-k8s-role-binding-config.libsonnet b/jsonnet/prometheus/prometheus-k8s-role-binding-config.libsonnet
index 2319aa356018ee946a77043820a931d2a9b053a0..631e5fa566ab5edd1e68316b7a506d817b2cb6fc 100644
--- a/jsonnet/prometheus/prometheus-k8s-role-binding-config.libsonnet
+++ b/jsonnet/prometheus/prometheus-k8s-role-binding-config.libsonnet
@@ -1,5 +1,5 @@
local prometheusNamespaceRoleBinding = import "prometheus-namespace-role-binding.libsonnet";
{
- new(namespace):: prometheusNamespaceRoleBinding.new(namespace, namespace, "prometheus-k8s-config")
+ new(namespace):: prometheusNamespaceRoleBinding.new(namespace, namespace, "prometheus-k8s-config", "prometheus-k8s")
}
diff --git a/jsonnet/prometheus/prometheus-k8s-role-binding-default.libsonnet b/jsonnet/prometheus/prometheus-k8s-role-binding-default.libsonnet
index f5d38ce7c8c10b74021e0775b5fc4671c82b4086..e88ece995daa60c35498e78e9268354c8cc255ae 100644
--- a/jsonnet/prometheus/prometheus-k8s-role-binding-default.libsonnet
+++ b/jsonnet/prometheus/prometheus-k8s-role-binding-default.libsonnet
@@ -1,5 +1,5 @@
local prometheusNamespaceRoleBinding = import "prometheus-namespace-role-binding.libsonnet";
{
- new(namespace):: prometheusNamespaceRoleBinding.new(namespace, "default", "prometheus-k8s")
+ new(namespace):: prometheusNamespaceRoleBinding.new(namespace, "default", "prometheus-k8s", "prometheus-k8s")
}
diff --git a/jsonnet/prometheus/prometheus-k8s-role-binding-kube-system.libsonnet b/jsonnet/prometheus/prometheus-k8s-role-binding-kube-system.libsonnet
index 04c481cacfd36443c86582c7a6537ee115e70905..33967e0a08cd17996274166bc9e7bb39103daf09 100644
--- a/jsonnet/prometheus/prometheus-k8s-role-binding-kube-system.libsonnet
+++ b/jsonnet/prometheus/prometheus-k8s-role-binding-kube-system.libsonnet
@@ -1,5 +1,5 @@
local prometheusNamespaceRoleBinding = import "prometheus-namespace-role-binding.libsonnet";
{
- new(namespace):: prometheusNamespaceRoleBinding.new(namespace, "kube-system", "prometheus-k8s")
+ new(namespace):: prometheusNamespaceRoleBinding.new(namespace, "kube-system", "prometheus-k8s", "prometheus-k8s")
}
diff --git a/jsonnet/prometheus/prometheus-k8s-role-binding-namespace.libsonnet b/jsonnet/prometheus/prometheus-k8s-role-binding-namespace.libsonnet
index 7833f78534e4351653f4365d660a92029d0435e0..d70ed6ac803235eecb2e0f1c9c045cdf0486540f 100644
--- a/jsonnet/prometheus/prometheus-k8s-role-binding-namespace.libsonnet
+++ b/jsonnet/prometheus/prometheus-k8s-role-binding-namespace.libsonnet
@@ -1,5 +1,5 @@
local prometheusNamespaceRoleBinding = import "prometheus-namespace-role-binding.libsonnet";
{
- new(namespace):: prometheusNamespaceRoleBinding.new(namespace, namespace, "prometheus-k8s")
+ new(namespace):: prometheusNamespaceRoleBinding.new(namespace, namespace, "prometheus-k8s", "prometheus-k8s")
}
diff --git a/jsonnet/prometheus/prometheus-k8s-rules.libsonnet b/jsonnet/prometheus/prometheus-k8s-rules.libsonnet
index abe98fa943dd632ff6fbda18a2ea7dd7af3f0183..d20145694b4cc86856beee3ee435594aa0bfea88 100644
--- a/jsonnet/prometheus/prometheus-k8s-rules.libsonnet
+++ b/jsonnet/prometheus/prometheus-k8s-rules.libsonnet
@@ -4,5 +4,6 @@ local configMap = k.core.v1.configMap;
{
new(namespace, ruleFiles)::
configMap.new("prometheus-k8s-rules", ruleFiles) +
+ configMap.mixin.metadata.withLabels({role: "alert-rules", prometheus: "k8s"}) +
configMap.mixin.metadata.withNamespace(namespace)
}
diff --git a/jsonnet/prometheus/prometheus-namespace-role-binding.libsonnet b/jsonnet/prometheus/prometheus-namespace-role-binding.libsonnet
index 8b255fa0fd54534f908748ca2d0bc5d85937692e..a63bcc9ccde4497b8f5eb37fb77e145444a14dd2 100644
--- a/jsonnet/prometheus/prometheus-namespace-role-binding.libsonnet
+++ b/jsonnet/prometheus/prometheus-namespace-role-binding.libsonnet
@@ -2,12 +2,12 @@ local k = import "ksonnet.beta.3/k.libsonnet";
local roleBinding = k.rbac.v1.roleBinding;
{
- new(serviceAccountNamespace, namespace, name)::
+ new(serviceAccountNamespace, namespace, roleName, serviceAccountName)::
roleBinding.new() +
- roleBinding.mixin.metadata.withName(name) +
+ roleBinding.mixin.metadata.withName(roleName) +
roleBinding.mixin.metadata.withNamespace(namespace) +
roleBinding.mixin.roleRef.withApiGroup("rbac.authorization.k8s.io") +
- roleBinding.mixin.roleRef.withName(name) +
+ roleBinding.mixin.roleRef.withName(roleName) +
roleBinding.mixin.roleRef.mixinInstance({kind: "Role"}) +
- roleBinding.withSubjects([{kind: "ServiceAccount", name: name, namespace: serviceAccountNamespace}])
+ roleBinding.withSubjects([{kind: "ServiceAccount", name: serviceAccountName, namespace: serviceAccountNamespace}])
}
diff --git a/manifests/grafana/grafana-dashboard-definitions.yaml b/manifests/grafana/grafana-dashboard-definitions.yaml
index df4c5203b2537f72ed93f6d438ef9a188e7e2961..573281af2e269348f2a248d33089fd7e37a0a6dc 100644
--- a/manifests/grafana/grafana-dashboard-definitions.yaml
+++ b/manifests/grafana/grafana-dashboard-definitions.yaml
@@ -3502,7 +3502,7 @@ data:
"title": "Kubernetes Cluster Status",
"version": 0
}
- kubernetes-kubelet-dashboard.json: |-
+ kubernetes-control-plane-status-dashboard.json: |-
{
"annotations": {
"list": [
@@ -3517,7 +3517,7 @@ data:
"links": [
],
- "refresh": "",
+ "refresh": "10s",
"rows": [
{
"collapse": false,
@@ -3529,16 +3529,16 @@ data:
"colorBackground": false,
"colorValue": false,
"colors": [
- "#299c46",
+ "rgba(245, 54, 54, 0.9)",
"rgba(237, 129, 40, 0.89)",
- "#d44a3a"
+ "rgba(50, 172, 45, 0.97)"
],
"datasource": "prometheus",
- "format": "none",
+ "format": "percent",
"gauge": {
"maxValue": 100,
"minValue": 0,
- "show": false,
+ "show": true,
"thresholdLabels": false,
"thresholdMarkers": true
},
@@ -3575,151 +3575,130 @@ data:
"to": "null"
}
],
- "span": 2,
+ "span": 3,
"sparkline": {
"fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
"lineColor": "rgb(31, 120, 193)",
- "show": true
+ "show": false
},
"tableColumn": "",
"targets": [
{
- "expr": "sum(kubelet_running_pod_count{instance=\u007e\"$instance\"})",
+ "expr": "(sum(up{job=\"apiserver\"} == 1) / sum(up{job=\"apiserver\"})) * 100",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": ""
}
],
- "thresholds": "",
- "title": "Count",
+ "thresholds": "50, 80",
+ "title": "API Servers UP",
"type": "singlestat",
"valueFontSize": "80%",
"valueMaps": [
{
"op": "=",
- "text": "0",
+ "text": "N/A",
"value": "null"
}
],
"valueName": "current"
},
{
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
"datasource": "prometheus",
- "fill": 1,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
"gridPos": {
},
"id": 3,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "seriesOverrides": [
+ "interval": null,
+ "links": [
],
- "spaceLength": 10,
- "span": 10,
- "stack": true,
- "steppedLine": false,
- "targets": [
+ "mappingType": 1,
+ "mappingTypes": [
{
- "expr": "kubelet_running_pod_count{instance=\u007e\"$instance\"}",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{ instance }}",
- "refId": "A"
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
}
],
- "thresholds": [
-
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
],
- "timeFrom": null,
- "timeShift": null,
- "title": "Count",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": [
-
- ]
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
},
- "yaxes": [
+ "tableColumn": "",
+ "targets": [
{
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": 0,
- "show": true
- },
+ "expr": "(sum(up{job=\"kube-controller-manager\"} == 1) / sum(up{job=\"kube-controller-manager\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "Controller Mangers UP",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
{
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": 0,
- "show": true
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
}
- ]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Pods",
- "titleSize": "h4",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "height": "250px",
- "panels": [
+ ],
+ "valueName": "current"
+ },
{
"cacheTimeout": null,
"colorBackground": false,
"colorValue": false,
"colors": [
- "#299c46",
+ "rgba(245, 54, 54, 0.9)",
"rgba(237, 129, 40, 0.89)",
- "#d44a3a"
+ "rgba(50, 172, 45, 0.97)"
],
"datasource": "prometheus",
- "format": "none",
+ "format": "percent",
"gauge": {
"maxValue": 100,
"minValue": 0,
- "show": false,
+ "show": true,
"thresholdLabels": false,
"thresholdMarkers": true
},
@@ -3756,129 +3735,122 @@ data:
"to": "null"
}
],
- "span": 2,
+ "span": 3,
"sparkline": {
"fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
"lineColor": "rgb(31, 120, 193)",
- "show": true
+ "show": false
},
"tableColumn": "",
"targets": [
{
- "expr": "sum(kubelet_running_container_count{instance=\u007e\"$instance\"})",
+ "expr": "(sum(up{job=\"kube-scheduler\"} == 1) / sum(up{job=\"kube-scheduler\"})) * 100",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": ""
}
],
- "thresholds": "",
- "title": "Count",
+ "thresholds": "50, 80",
+ "title": "Schedulers UP",
"type": "singlestat",
"valueFontSize": "80%",
"valueMaps": [
{
"op": "=",
- "text": "0",
+ "text": "N/A",
"value": "null"
}
],
"valueName": "current"
},
{
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
"datasource": "prometheus",
- "fill": 1,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
"gridPos": {
},
"id": 5,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "seriesOverrides": [
+ "interval": null,
+ "links": [
],
- "spaceLength": 10,
- "span": 10,
- "stack": true,
- "steppedLine": false,
- "targets": [
+ "mappingType": 1,
+ "mappingTypes": [
{
- "expr": "kubelet_running_container_count{instance=\u007e\"$instance\"}",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{ instance }}",
- "refId": "A"
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
}
],
- "thresholds": [
-
- ],
- "timeFrom": null,
- "timeShift": null,
- "title": "Count",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": [
-
- ]
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
},
- "yaxes": [
+ "tableColumn": "",
+ "targets": [
{
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": 0,
- "show": true
- },
+ "expr": "max(sum by(instance) (rate(apiserver_request_count{code=\u007e\"5..\"}[5m])) / sum by(instance) (rate(apiserver_request_count[5m]))) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "API Request Error Rate",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
{
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": 0,
- "show": true
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
}
- ]
+ ],
+ "valueName": "current"
}
],
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "Containers",
- "titleSize": "h4",
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
"type": "row"
},
{
@@ -3894,7 +3866,6 @@ data:
"dashLength": 10,
"dashes": false,
"datasource": "prometheus",
- "description": "Rate of Kubelet Operations in 5min",
"fill": 1,
"gridPos": {
@@ -3928,10 +3899,10 @@ data:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(kubelet_runtime_operations{instance=\u007e\"$instance\"}[5m])) by (instance)",
+ "expr": "sum by(verb) (rate(apiserver_latency_seconds:quantile[5m]) >= 0)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{ instance }}",
+ "legendFormat": "",
"refId": "A"
}
],
@@ -3940,7 +3911,7 @@ data:
],
"timeFrom": null,
"timeShift": null,
- "title": "Operations",
+ "title": "API Request Latency",
"tooltip": {
"shared": true,
"sort": 0,
@@ -3962,7 +3933,7 @@ data:
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
@@ -3970,7 +3941,7 @@ data:
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
}
]
@@ -3979,97 +3950,11 @@ data:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": true,
- "title": "Kubelet",
- "titleSize": "h4",
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
"type": "row"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [
-
- ],
- "templating": {
- "list": [
- {
- "allValue": null,
- "current": {
-
- },
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": true,
- "label": null,
- "multi": false,
- "name": "instance",
- "options": [
-
- ],
- "query": "label_values(kubelet_running_pod_count,instance)",
- "refresh": 2,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubelet",
- "version": 0
- }
- nodes.json: |-
- {
- "annotations": {
- "list": [
-
- ]
- },
- "editable": false,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "id": null,
- "links": [
-
- ],
- "refresh": "",
- "rows": [
+ },
{
"collapse": false,
"collapsed": false,
@@ -4087,7 +3972,7 @@ data:
"gridPos": {
},
- "id": 2,
+ "id": 7,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -4111,16 +3996,23 @@ data:
],
"spaceLength": 10,
- "span": 6,
+ "span": 12,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "100 - (avg by (cpu) (irate(node_cpu{mode=\"idle\", instance=\"$server\"}[5m])) * 100)",
+ "expr": "sum by(instance) (rate(apiserver_request_count{code!\u007e\"2..\"}[5m]))",
"format": "time_series",
- "intervalFactor": 10,
- "legendFormat": "{{cpu}}",
+ "intervalFactor": 2,
+ "legendFormat": "Error Rate",
"refId": "A"
+ },
+ {
+ "expr": "sum by(instance) (rate(apiserver_request_count[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Request Rate",
+ "refId": "B"
}
],
"thresholds": [
@@ -4128,7 +4020,7 @@ data:
],
"timeFrom": null,
"timeShift": null,
- "title": "Idle CPU",
+ "title": "API Request Rate",
"tooltip": {
"shared": true,
"sort": 0,
@@ -4146,23 +4038,37 @@ data:
},
"yaxes": [
{
- "format": "percent",
+ "format": "short",
"label": null,
"logBase": 1,
- "max": 100,
- "min": 0,
+ "max": null,
+ "min": null,
"show": true
},
{
- "format": "percent",
+ "format": "short",
"label": null,
"logBase": 1,
- "max": 100,
- "min": 0,
+ "max": null,
+ "min": null,
"show": true
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
{
"aliasColors": {
@@ -4175,7 +4081,7 @@ data:
"gridPos": {
},
- "id": 3,
+ "id": 8,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -4199,30 +4105,16 @@ data:
],
"spaceLength": 10,
- "span": 6,
+ "span": 12,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "node_load1{instance=\"$server\"} * 100",
+ "expr": "cluster:scheduler_e2e_scheduling_latency_seconds:quantile",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "load 1m",
+ "legendFormat": "",
"refId": "A"
- },
- {
- "expr": "node_load5{instance=\"$server\"} * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "load 5m",
- "refId": "B"
- },
- {
- "expr": "node_load15{instance=\"$server\"} * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "load 15m",
- "refId": "C"
}
],
"thresholds": [
@@ -4230,7 +4122,7 @@ data:
],
"timeFrom": null,
"timeShift": null,
- "title": "System load",
+ "title": "End to End Scheduling Latency",
"tooltip": {
"shared": true,
"sort": 0,
@@ -4248,7 +4140,7 @@ data:
},
"yaxes": [
{
- "format": "percent",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -4256,7 +4148,7 @@ data:
"show": true
},
{
- "format": "percent",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -4273,17 +4165,157 @@ data:
"title": "Dashboard Row",
"titleSize": "h6",
"type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {
-
- },
- "bars": false,
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+
+ ]
+ },
+ "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": "browser",
+ "title": "Kubernetes Control Plane Status",
+ "version": 0
+ }
+ kubernetes-kubelet-dashboard.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kubelet_running_pod_count{instance=\u007e\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Count",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
"dashLength": 10,
"dashes": false,
"datasource": "prometheus",
@@ -4291,7 +4323,7 @@ data:
"gridPos": {
},
- "id": 4,
+ "id": 3,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -4315,37 +4347,16 @@ data:
],
"spaceLength": 10,
- "span": 9,
- "stack": false,
+ "span": 10,
+ "stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}",
+ "expr": "kubelet_running_pod_count{instance=\u007e\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "memory used",
+ "legendFormat": "{{ instance }}",
"refId": "A"
- },
- {
- "expr": "node_memory_Buffers{instance=\"$server\"}",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "memory buffers",
- "refId": "B"
- },
- {
- "expr": "node_memory_Cached{instance=\"$server\"}",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "memory cached",
- "refId": "C"
- },
- {
- "expr": "node_memory_MemFree{instance=\"$server\"}",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "memory free",
- "refId": "D"
}
],
"thresholds": [
@@ -4353,7 +4364,7 @@ data:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage",
+ "title": "Count",
"tooltip": {
"shared": true,
"sort": 0,
@@ -4371,45 +4382,59 @@ data:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
}
]
- },
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Pods",
+ "titleSize": "h4",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
{
"cacheTimeout": null,
"colorBackground": false,
"colorValue": false,
"colors": [
- "rgba(50, 172, 45, 0.97)",
+ "#299c46",
"rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
+ "#d44a3a"
],
"datasource": "prometheus",
- "format": "percent",
+ "format": "none",
"gauge": {
"maxValue": 100,
"minValue": 0,
- "show": true,
+ "show": false,
"thresholdLabels": false,
"thresholdMarkers": true
},
"gridPos": {
},
- "id": 5,
+ "id": 4,
"interval": null,
"links": [
@@ -4439,49 +4464,34 @@ data:
"to": "null"
}
],
- "span": 3,
+ "span": 2,
"sparkline": {
"fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
"lineColor": "rgb(31, 120, 193)",
- "show": false
+ "show": true
},
"tableColumn": "",
"targets": [
{
- "expr": "((node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}) / node_memory_MemTotal{instance=\"$server\"}) * 100",
+ "expr": "sum(kubelet_running_container_count{instance=\u007e\"$instance\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": ""
}
],
- "thresholds": "80, 90",
- "title": "Memory Usage",
+ "thresholds": "",
+ "title": "Count",
"type": "singlestat",
"valueFontSize": "80%",
"valueMaps": [
{
"op": "=",
- "text": "N/A",
+ "text": "0",
"value": "null"
}
],
"valueName": "current"
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
- "type": "row"
- },
- {
- "collapse": false,
- "collapsed": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -4494,7 +4504,7 @@ data:
"gridPos": {
},
- "id": 6,
+ "id": 5,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -4515,40 +4525,19 @@ data:
"renderer": "flot",
"repeat": null,
"seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 10,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
{
- "alias": "read",
- "yaxis": 1
- },
- {
- "alias": "io time",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by (instance) (rate(node_disk_bytes_read{instance=\"$server\"}[2m]))",
+ "expr": "kubelet_running_container_count{instance=\u007e\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "read",
+ "legendFormat": "{{ instance }}",
"refId": "A"
- },
- {
- "expr": "sum by (instance) (rate(node_disk_bytes_written{instance=\"$server\"}[2m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "written",
- "refId": "B"
- },
- {
- "expr": "sum by (instance) (rate(node_disk_io_time_ms{instance=\"$server\"}[2m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "io time",
- "refId": "C"
}
],
"thresholds": [
@@ -4556,7 +4545,7 @@ data:
],
"timeFrom": null,
"timeShift": null,
- "title": "Disk I/O",
+ "title": "Count",
"tooltip": {
"shared": true,
"sort": 0,
@@ -4574,110 +4563,30 @@ data:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
- "format": "ms",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
}
]
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
-
- },
- "id": 7,
- "interval": null,
- "links": [
-
- ],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "(sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) - sum(node_filesystem_free{device!=\"rootfs\",instance=\"$server\"})) / sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": ""
- }
- ],
- "thresholds": "80, 90",
- "title": "Disk Space Usage",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
}
],
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
+ "showTitle": true,
+ "title": "Containers",
+ "titleSize": "h4",
"type": "row"
},
{
@@ -4693,99 +4602,12 @@ data:
"dashLength": 10,
"dashes": false,
"datasource": "prometheus",
+ "description": "Rate of Kubelet Operations in 5min",
"fill": 1,
"gridPos": {
},
- "id": 8,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "seriesOverrides": [
-
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(node_network_receive_bytes{instance=\"$server\",device!\u007e\"lo\"}[5m])",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{device}}",
- "refId": "A"
- }
- ],
- "thresholds": [
-
- ],
- "timeFrom": null,
- "timeShift": null,
- "title": "Network Received",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": [
-
- ]
- },
- "yaxes": [
- {
- "format": "bytes",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "bytes",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {
-
- },
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "fill": 1,
- "gridPos": {
-
- },
- "id": 9,
+ "id": 6,
"legend": {
"alignAsTable": false,
"avg": false,
@@ -4809,15 +4631,15 @@ data:
],
"spaceLength": 10,
- "span": 6,
+ "span": 12,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "rate(node_network_transmit_bytes{instance=\"$server\",device!\u007e\"lo\"}[5m])",
+ "expr": "sum(rate(kubelet_runtime_operations{instance=\u007e\"$instance\"}[5m])) by (instance)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{device}}",
+ "legendFormat": "{{ instance }}",
"refId": "A"
}
],
@@ -4826,7 +4648,7 @@ data:
],
"timeFrom": null,
"timeShift": null,
- "title": "Network Transmitted",
+ "title": "Operations",
"tooltip": {
"shared": true,
"sort": 0,
@@ -4844,19 +4666,19 @@ data:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
},
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": null,
+ "min": 0,
"show": true
}
]
@@ -4865,9 +4687,9 @@ data:
"repeat": null,
"repeatIteration": null,
"repeatRowId": null,
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6",
+ "showTitle": true,
+ "title": "Kubelet",
+ "titleSize": "h4",
"type": "row"
}
],
@@ -4885,14 +4707,14 @@ data:
},
"datasource": "prometheus",
"hide": 0,
- "includeAll": false,
+ "includeAll": true,
"label": null,
"multi": false,
- "name": "server",
+ "name": "instance",
"options": [
],
- "query": "label_values(node_boot_time, instance)",
+ "query": "label_values(kubelet_running_pod_count,instance)",
"refresh": 2,
"regex": "",
"sort": 0,
@@ -4936,10 +4758,10 @@ data:
]
},
"timezone": "browser",
- "title": "Nodes",
+ "title": "Kubelet",
"version": 0
}
- pods-dashboard.json: |-
+ kubernetes-resource-requests-dashboard.json: |-
{
"annotations": {
"list": [
@@ -4954,7 +4776,7 @@ data:
"links": [
],
- "refresh": "",
+ "refresh": "10s",
"rows": [
{
"collapse": false,
@@ -4975,12 +4797,12 @@ data:
},
"id": 2,
"legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": true,
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -4997,30 +4819,23 @@ data:
],
"spaceLength": 10,
- "span": 12,
+ "span": 9,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "sum by(container_name) (container_memory_usage_bytes{namespace=\"$namespace\", pod_name=\"$pod\", container_name=\u007e\"$container\", container_name!=\"POD\"})",
+ "expr": "min(sum(kube_node_status_allocatable_cpu_cores) by (instance))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Current: {{ container_name }}",
+ "legendFormat": "Allocatable CPU Cores",
"refId": "A"
},
{
- "expr": "sum by(container) (kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\", pod=\"$pod\", container=\u007e\"$container\", container!=\"POD\"})",
+ "expr": "max(sum(kube_pod_container_resource_requests_cpu_cores) by (instance))",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "Requested: {{ container }}",
+ "legendFormat": "Requested CPU Cores",
"refId": "B"
- },
- {
- "expr": "sum by(container) (kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\", pod=\"$pod\", container=\u007e\"$container\", container!=\"POD\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Limit: {{ container }}",
- "refId": "C"
}
],
"thresholds": [
@@ -5028,7 +4843,7 @@ data:
],
"timeFrom": null,
"timeShift": null,
- "title": "Memory Usage",
+ "title": "CPU Cores",
"tooltip": {
"shared": true,
"sort": 0,
@@ -5046,7 +4861,7 @@ data:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -5054,7 +4869,7 @@ data:
"show": true
},
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
@@ -5062,6 +4877,86 @@ data:
"show": true
}
]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 3,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(sum(kube_pod_container_resource_requests_cpu_cores) by (instance)) / min(sum(kube_node_status_allocatable_cpu_cores) by (instance)) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "CPU Cores",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
}
],
"repeat": null,
@@ -5089,81 +4984,2158 @@ data:
"gridPos": {
},
- "id": 3,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "seriesOverrides": [
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "min(sum(kube_node_status_allocatable_memory_bytes) by (instance))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Allocatable Memory",
+ "refId": "A"
+ },
+ {
+ "expr": "max(sum(kube_pod_container_resource_requests_memory_bytes) by (instance))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Requested Memory",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(sum(kube_pod_container_resource_requests_memory_bytes) by (instance)) / min(sum(kube_node_status_allocatable_memory_bytes) by (instance)) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+
+ ]
+ },
+ "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": "browser",
+ "title": "Kubernetes Resource Requests",
+ "version": 0
+ }
+ nodes.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "100 - (avg by (cpu) (irate(node_cpu{mode=\"idle\", instance=\"$server\"}[5m])) * 100)",
+ "format": "time_series",
+ "intervalFactor": 10,
+ "legendFormat": "{{cpu}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Idle CPU",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": 100,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": 100,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_load1{instance=\"$server\"} * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 1m",
+ "refId": "A"
+ },
+ {
+ "expr": "node_load5{instance=\"$server\"} * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 5m",
+ "refId": "B"
+ },
+ {
+ "expr": "node_load15{instance=\"$server\"} * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 15m",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "System load",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "percent",
+ "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,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory used",
+ "refId": "A"
+ },
+ {
+ "expr": "node_memory_Buffers{instance=\"$server\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory buffers",
+ "refId": "B"
+ },
+ {
+ "expr": "node_memory_Cached{instance=\"$server\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory cached",
+ "refId": "C"
+ },
+ {
+ "expr": "node_memory_MemFree{instance=\"$server\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory free",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "((node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}) / node_memory_MemTotal{instance=\"$server\"}) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory Usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "read",
+ "yaxis": 1
+ },
+ {
+ "alias": "io time",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (instance) (rate(node_disk_bytes_read{instance=\"$server\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "read",
+ "refId": "A"
+ },
+ {
+ "expr": "sum by (instance) (rate(node_disk_bytes_written{instance=\"$server\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "written",
+ "refId": "B"
+ },
+ {
+ "expr": "sum by (instance) (rate(node_disk_io_time_ms{instance=\"$server\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "io time",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "ms",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) - sum(node_filesystem_free{device!=\"rootfs\",instance=\"$server\"})) / sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Disk Space Usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 8,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_receive_bytes{instance=\"$server\",device!\u007e\"lo\"}[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Received",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_transmit_bytes{instance=\"$server\",device!\u007e\"lo\"}[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Transmitted",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": null,
+ "multi": false,
+ "name": "server",
+ "options": [
+
+ ],
+ "query": "label_values(node_boot_time, instance)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Nodes",
+ "version": 0
+ }
+ pods-dashboard.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by(container_name) (container_memory_usage_bytes{namespace=\"$namespace\", pod_name=\"$pod\", container_name=\u007e\"$container\", container_name!=\"POD\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Current: {{ container_name }}",
+ "refId": "A"
+ },
+ {
+ "expr": "sum by(container) (kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\", pod=\"$pod\", container=\u007e\"$container\", container!=\"POD\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Requested: {{ container }}",
+ "refId": "B"
+ },
+ {
+ "expr": "sum by(container) (kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\", pod=\"$pod\", container=\u007e\"$container\", container!=\"POD\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Limit: {{ container }}",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (container_name) (rate(container_cpu_usage_seconds_total{image!=\"\",container_name!=\"POD\",pod_name=\"$pod\"}[1m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{ container_name }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{pod_name=\"$pod\"}[1m])))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{ pod_name }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Namespace",
+ "multi": false,
+ "name": "namespace",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, namespace)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Pod",
+ "multi": false,
+ "name": "pod",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{namespace=\u007e\"$namespace\"}, pod)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": true,
+ "label": "Container",
+ "multi": false,
+ "name": "container",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_container_info{namespace=\"$namespace\", pod=\"$pod\"}, container)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Pods",
+ "version": 0
+ }
+ statefulset-dashboard.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=\"$statefulset_namespace\",pod_name=\u007e\"$statefulset_name.*\"}[3m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "CPU",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 3,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(container_memory_usage_bytes{namespace=\"$statefulset_namespace\",pod_name=\u007e\"$statefulset_name.*\"}) / 1024^3",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Memory",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 4,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(rate(container_network_transmit_bytes_total{namespace=\"$statefulset_namespace\",pod_name=\u007e\"$statefulset_name.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{namespace=\"$statefulset_namespace\",pod_name=\u007e\"$statefulset_name.*\"}[3m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Network",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "100px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(kube_statefulset_replicas{namespace=\"$statefulset_namespace\",statefulset=\"$statefulset_name\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Desired Replicas",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 6,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "min(kube_statefulset_status_replicas_current{namespace=\"$statefulset_namespace\",statefulset=\"$statefulset_name\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Replicas of current version",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
"targets": [
{
- "expr": "sum by (container_name) (rate(container_cpu_usage_seconds_total{image!=\"\",container_name!=\"POD\",pod_name=\"$pod\"}[1m]))",
+ "expr": "max(kube_statefulset_status_observed_generation{namespace=\"$statefulset_namespace\",statefulset=\"$statefulset_name\"}) without (instance, pod)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{ container_name }}",
- "refId": "A"
+ "legendFormat": ""
}
],
- "thresholds": [
-
+ "thresholds": "",
+ "title": "Observed Generation",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
],
- "timeFrom": null,
- "timeShift": null,
- "title": "CPU Usage",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
},
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": [
+ "gridPos": {
- ]
},
- "yaxes": [
+ "id": 8,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
{
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": 0,
- "show": true
+ "name": "value to text",
+ "value": 1
},
{
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": 0,
- "show": true
+ "name": "range to text",
+ "value": 2
}
- ]
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(kube_statefulset_metadata_generation{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Metadata Generation",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
}
],
"repeat": null,
@@ -5191,14 +7163,14 @@ data:
"gridPos": {
},
- "id": 4,
+ "id": 9,
"legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
"max": false,
"min": false,
- "rightSide": true,
+ "rightSide": false,
"show": true,
"total": false,
"values": false
@@ -5220,11 +7192,39 @@ data:
"steppedLine": false,
"targets": [
{
- "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{pod_name=\"$pod\"}[1m])))",
+ "expr": "max(kube_statefulset_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
"format": "time_series",
"intervalFactor": 2,
- "legendFormat": "{{ pod_name }}",
+ "legendFormat": "replicas specified",
"refId": "A"
+ },
+ {
+ "expr": "max(kube_statefulset_status_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "replicas created",
+ "refId": "B"
+ },
+ {
+ "expr": "min(kube_statefulset_status_replicas_ready{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "ready",
+ "refId": "C"
+ },
+ {
+ "expr": "min(kube_statefulset_status_replicas_current{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "replicas of current version",
+ "refId": "D"
+ },
+ {
+ "expr": "min(kube_statefulset_status_replicas_updated{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "updated",
+ "refId": "E"
}
],
"thresholds": [
@@ -5232,7 +7232,7 @@ data:
],
"timeFrom": null,
"timeShift": null,
- "title": "Network I/O",
+ "title": "Replicas",
"tooltip": {
"shared": true,
"sort": 0,
@@ -5250,19 +7250,19 @@ data:
},
"yaxes": [
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
},
{
- "format": "bytes",
+ "format": "short",
"label": null,
"logBase": 1,
"max": null,
- "min": 0,
+ "min": null,
"show": true
}
]
@@ -5294,11 +7294,11 @@ data:
"includeAll": false,
"label": "Namespace",
"multi": false,
- "name": "namespace",
+ "name": "statefulset_namespace",
"options": [
],
- "query": "label_values(kube_pod_info, namespace)",
+ "query": "label_values(kube_statefulset_metadata_generation, namespace)",
"refresh": 2,
"regex": "",
"sort": 0,
@@ -5318,39 +7318,13 @@ data:
"datasource": "prometheus",
"hide": 0,
"includeAll": false,
- "label": "Pod",
- "multi": false,
- "name": "pod",
- "options": [
-
- ],
- "query": "label_values(kube_pod_info{namespace=\u007e\"$namespace\"}, pod)",
- "refresh": 2,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [
-
- ],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {
-
- },
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": true,
- "label": "Container",
+ "label": "Name",
"multi": false,
- "name": "container",
+ "name": "statefulset_name",
"options": [
],
- "query": "label_values(kube_pod_container_info{namespace=\"$namespace\", pod=\"$pod\"}, container)",
+ "query": "label_values(kube_statefulset_metadata_generation{namespace=\"$statefulset_namespace\"}, statefulset)",
"refresh": 2,
"regex": "",
"sort": 0,
@@ -5394,7 +7368,7 @@ data:
]
},
"timezone": "browser",
- "title": "Pods",
+ "title": "StatefulSets",
"version": 0
}
kind: ConfigMap
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-binding-config.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-binding-config.yaml
index f21ff3c5978a18636dd43473d13eb3e62a697f8b..ec0129db5bf7f474b8aa124b63be72d50e97d667 100644
--- a/manifests/prometheus-k8s/prometheus-k8s-role-binding-config.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-binding-config.yaml
@@ -9,5 +9,5 @@ roleRef:
name: prometheus-k8s-config
subjects:
- kind: ServiceAccount
- name: prometheus-k8s-config
+ name: prometheus-k8s
namespace: monitoring
diff --git a/manifests/prometheus-k8s/prometheus-k8s-rules.yaml b/manifests/prometheus-k8s/prometheus-k8s-rules.yaml
index 90cb3f3e165c6a07d0206fa98682b00814bc02ae..0c03de56dab52307ffac3ec4be317e5e8af61f9a 100644
--- a/manifests/prometheus-k8s/prometheus-k8s-rules.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-rules.yaml
@@ -583,5 +583,8 @@ data:
rejected\n"
kind: ConfigMap
metadata:
+ labels:
+ prometheus: k8s
+ role: alert-rules
name: prometheus-k8s-rules
namespace: monitoring