diff --git a/jsonnetfile.json b/jsonnetfile.json index 13cd6fa..e3510dd 100644 --- a/jsonnetfile.json +++ b/jsonnetfile.json @@ -4,8 +4,8 @@ "name": "kube-prometheus", "source": { "git": { - "remote": "https://github.com/coreos/prometheus-operator", - "subdir": "contrib/kube-prometheus/jsonnet/kube-prometheus" + "remote": "https://github.com/coreos/kube-prometheus", + "subdir": "jsonnet/kube-prometheus" } }, "version": "master" diff --git a/jsonnetfile.lock.json b/jsonnetfile.lock.json index f892b64..f3fdde5 100644 --- a/jsonnetfile.lock.json +++ b/jsonnetfile.lock.json @@ -4,11 +4,11 @@ "name": "kube-prometheus", "source": { "git": { - "remote": "https://github.com/coreos/prometheus-operator", - "subdir": "contrib/kube-prometheus/jsonnet/kube-prometheus" + "remote": "https://github.com/coreos/kube-prometheus", + "subdir": "jsonnet/kube-prometheus" } }, - "version": "650359b3e627ae97a1f18cbd10d7ed9b2293c240" + "version": "7bd745ef78dce3c14ef9e315506d0c1c32fdcc9e" }, { "name": "ksonnet", @@ -28,7 +28,7 @@ "subdir": "" } }, - "version": "19db38fc449df024446059f21d5a329babaa3927" + "version": "7360753d27aa428758c918434503c1c35afcd3bb" }, { "name": "grafonnet", @@ -48,7 +48,7 @@ "subdir": "grafana-builder" } }, - "version": "e30a6040f3d7270655a980ab04d16142da4b429d" + "version": "ecaeaed2e21c0db29098811e7826a9b923e706c5" }, { "name": "grafana", @@ -78,7 +78,7 @@ "subdir": "Documentation/etcd-mixin" } }, - "version": "7a5acb4a43aa06bd9e32ab59a46271ab88d497e4" + "version": "216808eab50f74e02410e96878bbf2175d2916cb" } ] } diff --git a/manifests/alertmanager-service.yaml b/manifests/alertmanager-service.yaml index 0c75679..df4c9ff 100644 --- a/manifests/alertmanager-service.yaml +++ b/manifests/alertmanager-service.yaml @@ -13,3 +13,4 @@ spec: selector: alertmanager: main app: alertmanager + sessionAffinity: ClientIP diff --git a/manifests/grafana-dashboardDefinitions.yaml b/manifests/grafana-dashboardDefinitions.yaml index e51d002..5f7fd5e 100644 --- a/manifests/grafana-dashboardDefinitions.yaml +++ b/manifests/grafana-dashboardDefinitions.yaml @@ -4071,6 +4071,42 @@ items: "pattern": "Time", "type": "hidden" }, + { + "alias": "Pods", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 0, + "link": true, + "linkTooltip": "Drill down to pods", + "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Workloads", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 0, + "link": true, + "linkTooltip": "Drill down to workloads", + "linkUrl": "/d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, { "alias": "CPU Usage", "colorMode": null, @@ -4082,7 +4118,7 @@ items: "link": false, "linkTooltip": "Drill down", "linkUrl": "", - "pattern": "Value #A", + "pattern": "Value #C", "thresholds": [ ], @@ -4100,7 +4136,7 @@ items: "link": false, "linkTooltip": "Drill down", "linkUrl": "", - "pattern": "Value #B", + "pattern": "Value #D", "thresholds": [ ], @@ -4118,7 +4154,7 @@ items: "link": false, "linkTooltip": "Drill down", "linkUrl": "", - "pattern": "Value #C", + "pattern": "Value #E", "thresholds": [ ], @@ -4136,7 +4172,7 @@ items: "link": false, "linkTooltip": "Drill down", "linkUrl": "", - "pattern": "Value #D", + "pattern": "Value #F", "thresholds": [ ], @@ -4154,7 +4190,7 @@ items: "link": false, "linkTooltip": "Drill down", "linkUrl": "", - "pattern": "Value #E", + "pattern": "Value #G", "thresholds": [ ], @@ -4170,7 +4206,7 @@ items: "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": true, - "linkTooltip": "Drill down", + "linkTooltip": "Drill down to pods", "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", "pattern": "namespace", "thresholds": [ @@ -4197,7 +4233,7 @@ items: ], "targets": [ { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", + "expr": "count(mixin_pod_workload{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4206,7 +4242,7 @@ items: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "count(avg(mixin_pod_workload{cluster=\"$cluster\"}) by (workload, namespace)) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4215,7 +4251,7 @@ items: "step": 10 }, { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4224,7 +4260,7 @@ items: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4233,13 +4269,31 @@ items: "step": 10 }, { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "E", "step": 10 + }, + { + "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + }, + { + "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "G", + "step": 10 } ], "thresholds": [ @@ -4437,43 +4491,43 @@ items: "type": "hidden" }, { - "alias": "Memory Usage", + "alias": "Pods", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", - "decimals": 2, - "link": false, - "linkTooltip": "Drill down", - "linkUrl": "", + "decimals": 0, + "link": true, + "linkTooltip": "Drill down to pods", + "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", "pattern": "Value #A", "thresholds": [ ], "type": "number", - "unit": "bytes" + "unit": "short" }, { - "alias": "Memory Requests", + "alias": "Workloads", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", - "decimals": 2, - "link": false, - "linkTooltip": "Drill down", - "linkUrl": "", + "decimals": 0, + "link": true, + "linkTooltip": "Drill down to workloads", + "linkUrl": "/d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", "pattern": "Value #B", "thresholds": [ ], "type": "number", - "unit": "bytes" + "unit": "short" }, { - "alias": "Memory Requests %", + "alias": "CPU Usage", "colorMode": null, "colors": [ @@ -4488,10 +4542,10 @@ items: ], "type": "number", - "unit": "percentunit" + "unit": "short" }, { - "alias": "Memory Limits", + "alias": "Memory Usage", "colorMode": null, "colors": [ @@ -4509,7 +4563,7 @@ items: "unit": "bytes" }, { - "alias": "Memory Limits %", + "alias": "Memory Requests", "colorMode": null, "colors": [ @@ -4522,6 +4576,60 @@ items: "pattern": "Value #E", "thresholds": [ + ], + "type": "number", + "unit": "bytes" + }, + { + "alias": "Memory Requests %", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "percentunit" + }, + { + "alias": "Memory Limits", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #G", + "thresholds": [ + + ], + "type": "number", + "unit": "bytes" + }, + { + "alias": "Memory Limits %", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #H", + "thresholds": [ + ], "type": "number", "unit": "percentunit" @@ -4535,7 +4643,7 @@ items: "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": true, - "linkTooltip": "Drill down", + "linkTooltip": "Drill down to pods", "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", "pattern": "namespace", "thresholds": [ @@ -4562,7 +4670,7 @@ items: ], "targets": [ { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)", + "expr": "count(mixin_pod_workload{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4571,7 +4679,7 @@ items: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "count(avg(mixin_pod_workload{cluster=\"$cluster\"}) by (workload, namespace)) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4580,7 +4688,7 @@ items: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4589,7 +4697,7 @@ items: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4598,13 +4706,31 @@ items: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "E", "step": 10 + }, + { + "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + }, + { + "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "G", + "step": 10 } ], "thresholds": [ @@ -5686,7 +5812,7 @@ items: ] }, "timezone": "", - "title": "Kubernetes / Compute Resources / Namespace", + "title": "Kubernetes / Compute Resources / Namespace (Pods)", "uid": "85a562078cdf77779eaa1add43ccec1e", "version": 0 } @@ -6684,6 +6810,1882 @@ items: metadata: name: grafana-dashboard-k8s-resources-pod namespace: monitoring +- apiVersion: v1 + data: + k8s-resources-workload.json: |- + { + "annotations": { + "list": [ + + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "links": [ + + ], + "refresh": "10s", + "rows": [ + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 1, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{pod}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Usage", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "CPU Usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$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": true, + "linkTooltip": "Drill down", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "pattern": "pod", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "pattern": "/.*/", + "thresholds": [ + + ], + "type": "string", + "unit": "short" + } + ], + "targets": [ + { + "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Quota", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "transform": "table", + "type": "table", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "CPU Quota", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$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(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{pod}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Usage", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Memory Usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$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": "Pod", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": true, + "linkTooltip": "Drill down", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "pattern": "pod", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "pattern": "/.*/", + "thresholds": [ + + ], + "type": "string", + "unit": "short" + } + ], + "targets": [ + { + "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Quota", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "transform": "table", + "type": "table", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Memory Quota", + "titleSize": "h6" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + "text": "prod", + "value": "prod" + }, + "datasource": "$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": "$datasource", + "hide": 0, + "includeAll": false, + "label": "namespace", + "multi": false, + "name": "namespace", + "options": [ + + ], + "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + "text": "prod", + "value": "prod" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": "workload", + "multi": false, + "name": "workload", + "options": [ + + ], + "query": "label_values(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}, workload)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + "text": "prod", + "value": "prod" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": "type", + "multi": false, + "name": "type", + "options": [ + + ], + "query": "label_values(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\"}, workload_type)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "", + "title": "Kubernetes / Compute Resources / Workload", + "uid": "a164a7f0339f99e89cea5cb47e9be617", + "version": 0 + } + kind: ConfigMap + metadata: + name: grafana-dashboard-k8s-resources-workload + namespace: monitoring +- apiVersion: v1 + data: + k8s-resources-workloads-namespace.json: |- + { + "annotations": { + "list": [ + + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "links": [ + + ], + "refresh": "10s", + "rows": [ + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 1, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "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(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{workload}} - {{workload_type}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Usage", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "CPU Usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$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": "Running Pods", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 0, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "CPU Usage", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "CPU Requests", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #C", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "CPU Requests %", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #D", + "thresholds": [ + + ], + "type": "number", + "unit": "percentunit" + }, + { + "alias": "CPU Limits", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #E", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "CPU Limits %", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "percentunit" + }, + { + "alias": "Workload", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": true, + "linkTooltip": "Drill down", + "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2", + "pattern": "workload", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Workload Type", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "workload_type", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "pattern": "/.*/", + "thresholds": [ + + ], + "type": "string", + "unit": "short" + } + ], + "targets": [ + { + "expr": "count(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}) by (workload, workload_type)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Quota", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "transform": "table", + "type": "table", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "CPU Quota", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$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(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{workload}} - {{workload_type}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Usage", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Memory Usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$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": "Running Pods", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 0, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Memory Usage", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "bytes" + }, + { + "alias": "Memory Requests", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #C", + "thresholds": [ + + ], + "type": "number", + "unit": "bytes" + }, + { + "alias": "Memory Requests %", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #D", + "thresholds": [ + + ], + "type": "number", + "unit": "percentunit" + }, + { + "alias": "Memory Limits", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #E", + "thresholds": [ + + ], + "type": "number", + "unit": "bytes" + }, + { + "alias": "Memory Limits %", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "percentunit" + }, + { + "alias": "Workload", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": true, + "linkTooltip": "Drill down", + "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2", + "pattern": "workload", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Workload Type", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "workload_type", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "pattern": "/.*/", + "thresholds": [ + + ], + "type": "string", + "unit": "short" + } + ], + "targets": [ + { + "expr": "count(mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}) by (workload, workload_type)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + }, + { + "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Quota", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "transform": "table", + "type": "table", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Memory Quota", + "titleSize": "h6" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + "text": "prod", + "value": "prod" + }, + "datasource": "$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": "$datasource", + "hide": 0, + "includeAll": false, + "label": "namespace", + "multi": false, + "name": "namespace", + "options": [ + + ], + "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "", + "title": "Kubernetes / Compute Resources / Namespace (Workloads)", + "uid": "a87fb0d919ec0ea5f6543124e16c42a5", + "version": 0 + } + kind: ConfigMap + metadata: + name: grafana-dashboard-k8s-resources-workloads-namespace + namespace: monitoring - apiVersion: v1 data: kubernetes-cluster-dashboard.json: |- @@ -8629,6 +10631,13 @@ items: "intervalFactor": 2, "legendFormat": "load 15m", "refId": "C" + }, + { + "expr": "count(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", mode=\"user\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "logical cores", + "refId": "D" } ], "thresholds": [ @@ -9933,7 +11942,7 @@ items: }, "id": 2, "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": true, "current": true, "max": true, @@ -9958,16 +11967,23 @@ items: ], "spaceLength": 10, - "span": 12, - "stack": false, + "span": 9, + "stack": true, "steppedLine": false, "targets": [ { - "expr": "(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"} - kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"}) / kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"} * 100\n", + "expr": "(\n sum without(instance, node) (kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n", "format": "time_series", "intervalFactor": 1, - "legendFormat": "{{ Usage }}", + "legendFormat": "Used Space", "refId": "A" + }, + { + "expr": "sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Free Space", + "refId": "B" } ], "thresholds": [ @@ -9993,22 +12009,106 @@ items: }, "yaxes": [ { - "format": "percent", + "format": "bytes", "label": null, "logBase": 1, - "max": 100, + "max": null, "min": 0, "show": true }, { - "format": "percent", + "format": "bytes", "label": null, "logBase": 1, - "max": 100, + "max": null, "min": 0, "show": true } ] + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "$datasource", + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + + }, + "id": 3, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 3, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "(\n kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n -\n kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n)\n/\nkubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "80, 90", + "title": "Volume Space Usage", + "tooltip": { + "shared": false + }, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" } ], "repeat": null, @@ -10035,9 +12135,9 @@ items: "gridPos": { }, - "id": 3, + "id": 4, "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": true, "current": true, "max": true, @@ -10062,16 +12162,23 @@ items: ], "spaceLength": 10, - "span": 12, - "stack": false, + "span": 9, + "stack": true, "steppedLine": false, "targets": [ { - "expr": "kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"} / kubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"} * 100\n", + "expr": "sum without(instance, node) (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n", "format": "time_series", "intervalFactor": 1, - "legendFormat": "{{ Usage }}", + "legendFormat": "Used inodes", "refId": "A" + }, + { + "expr": "(\n sum without(instance, node) (kubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": " Free inodes", + "refId": "B" } ], "thresholds": [ @@ -10097,22 +12204,106 @@ items: }, "yaxes": [ { - "format": "percent", + "format": "none", "label": null, "logBase": 1, - "max": 100, + "max": null, "min": 0, "show": true }, { - "format": "percent", + "format": "none", "label": null, "logBase": 1, - "max": 100, + "max": null, "min": 0, "show": true } ] + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "$datasource", + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + + }, + "id": 5, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 3, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n/\nkubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "80, 90", + "title": "Volume inodes Usage", + "tooltip": { + "shared": false + }, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" } ], "repeat": null, @@ -10187,7 +12378,7 @@ items: "options": [ ], - "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\"}, exported_namespace)", + "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\"}, namespace)", "refresh": 2, "regex": "", "sort": 0, @@ -10213,7 +12404,7 @@ items: "options": [ ], - "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", exported_namespace=\"$namespace\"}, persistentvolumeclaim)", + "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\"}, persistentvolumeclaim)", "refresh": 2, "regex": "", "sort": 0, @@ -10277,7 +12468,20 @@ items: ], "annotations": { "list": [ - + { + "builtIn": 1, + "datasource": "$datasource", + "enable": true, + "expr": "time() == BOOL timestamp(rate(kube_pod_container_status_restarts_total{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[2m]) > 0)", + "hide": false, + "iconColor": "rgba(215, 44, 44, 1)", + "name": "Restarts", + "showIn": 0, + "tags": [ + "restart" + ], + "type": "rows" + } ] }, "editable": false, @@ -10345,18 +12549,25 @@ items: "refId": "A" }, { - "expr": "sum by(container) (kube_pod_container_resource_requests_memory_bytes{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", + "expr": "sum by(container) (kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\", pod=\"$pod\", container=~\"$container\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Requested: {{ container }}", "refId": "B" }, { - "expr": "sum by(container) (kube_pod_container_resource_limits_memory_bytes{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", + "expr": "sum by(container) (kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\", pod=\"$pod\", container=~\"$container\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Limit: {{ container }}", "refId": "C" + }, + { + "expr": "sum by(container_name) (container_memory_cache{job=\"kubelet\", namespace=\"$namespace\", pod_name=~\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Cache: {{ container_name }}", + "refId": "D" } ], "thresholds": [ @@ -10463,14 +12674,14 @@ items: "refId": "A" }, { - "expr": "sum by(container) (kube_pod_container_resource_requests_cpu_cores{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", + "expr": "sum by(container) (kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\", pod=\"$pod\", container=~\"$container\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Requested: {{ container }}", "refId": "B" }, { - "expr": "sum by(container) (kube_pod_container_resource_limits_cpu_cores{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", + "expr": "sum by(container) (kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\", pod=\"$pod\", container=~\"$container\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Limit: {{ container }}", @@ -10577,8 +12788,15 @@ items: "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}[1m])))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{ pod_name }}", + "legendFormat": "RX: {{ pod_name }}", "refId": "A" + }, + { + "expr": "sort_desc(sum by (pod_name) (rate(container_network_transmit_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}[1m])))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "TX: {{ pod_name }}", + "refId": "B" } ], "thresholds": [ @@ -10629,6 +12847,110 @@ items: "title": "Dashboard Row", "titleSize": "h6", "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 5, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "max by (container) (kube_pod_container_status_restarts_total{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Restarts: {{ container }}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Total Restarts Per Container", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" } ], "schemaVersion": 14, diff --git a/manifests/grafana-deployment.yaml b/manifests/grafana-deployment.yaml index cbf0a4c..8ad3aa9 100644 --- a/manifests/grafana-deployment.yaml +++ b/manifests/grafana-deployment.yaml @@ -60,6 +60,12 @@ spec: - mountPath: /grafana-dashboard-definitions/0/k8s-resources-pod name: grafana-dashboard-k8s-resources-pod readOnly: false + - mountPath: /grafana-dashboard-definitions/0/k8s-resources-workload + name: grafana-dashboard-k8s-resources-workload + readOnly: false + - mountPath: /grafana-dashboard-definitions/0/k8s-resources-workloads-namespace + name: grafana-dashboard-k8s-resources-workloads-namespace + readOnly: false - mountPath: /grafana-dashboard-definitions/0/kubernetes-cluster-dashboard name: grafana-dashboard-kubernetes-cluster-dashboard readOnly: false @@ -114,6 +120,12 @@ spec: - configMap: name: grafana-dashboard-k8s-resources-pod name: grafana-dashboard-k8s-resources-pod + - configMap: + name: grafana-dashboard-k8s-resources-workload + name: grafana-dashboard-k8s-resources-workload + - configMap: + name: grafana-dashboard-k8s-resources-workloads-namespace + name: grafana-dashboard-k8s-resources-workloads-namespace - configMap: name: grafana-dashboard-kubernetes-cluster-dashboard name: grafana-dashboard-kubernetes-cluster-dashboard diff --git a/manifests/node-exporter-daemonset.yaml b/manifests/node-exporter-daemonset.yaml index 3d2a434..d04f659 100644 --- a/manifests/node-exporter-daemonset.yaml +++ b/manifests/node-exporter-daemonset.yaml @@ -22,6 +22,7 @@ spec: - --path.rootfs=/host/root - --collector.filesystem.ignored-mount-points=^/(dev|proc|sys|var/lib/docker/.+)($|/) - --collector.filesystem.ignored-fs-types=^(autofs|binfmt_misc|cgroup|configfs|debugfs|devpts|devtmpfs|fusectl|hugetlbfs|mqueue|overlay|proc|procfs|pstore|rpc_pipefs|securityfs|sysfs|tracefs)$ + - --collector.ntp image: carlosedp/node_exporter:v0.17.0 name: node-exporter resources: diff --git a/manifests/prometheus-adapter-configMap.yaml b/manifests/prometheus-adapter-configMap.yaml index d6ebd78..15434fc 100644 --- a/manifests/prometheus-adapter-configMap.yaml +++ b/manifests/prometheus-adapter-configMap.yaml @@ -3,7 +3,7 @@ data: config.yaml: | resourceRules: cpu: - containerQuery: sum(rate(container_cpu_usage_seconds_total{<<.LabelMatchers>>}[1m])) by (<<.GroupBy>>) + containerQuery: sum(rate(container_cpu_usage_seconds_total{<<.LabelMatchers>>,container_name!="POD",container_name!="",pod_name!=""}[1m])) by (<<.GroupBy>>) nodeQuery: sum(1 - rate(node_cpu_seconds_total{mode="idle"}[1m]) * on(namespace, pod) group_left(node) node_namespace_pod:kube_pod_info:{<<.LabelMatchers>>}) by (<<.GroupBy>>) resources: overrides: @@ -15,7 +15,7 @@ data: resource: pod containerLabel: container_name memory: - containerQuery: sum(container_memory_working_set_bytes{<<.LabelMatchers>>}) by (<<.GroupBy>>) + containerQuery: sum(container_memory_working_set_bytes{<<.LabelMatchers>>,container_name!="POD",container_name!="",pod_name!=""}) by (<<.GroupBy>>) nodeQuery: sum(node:node_memory_bytes_total:sum{<<.LabelMatchers>>} - node:node_memory_bytes_available:sum{<<.LabelMatchers>>}) by (<<.GroupBy>>) resources: overrides: diff --git a/manifests/prometheus-rules.yaml b/manifests/prometheus-rules.yaml index d63f739..3db51cc 100644 --- a/manifests/prometheus-rules.yaml +++ b/manifests/prometheus-rules.yaml @@ -44,11 +44,44 @@ spec: record: namespace_name:kube_pod_container_resource_requests_memory_bytes:sum - expr: | sum by (namespace, label_name) ( - sum(kube_pod_container_resource_requests_cpu_cores{job="kube-state-metrics"} and on(pod) kube_pod_status_scheduled{condition="true"}) by (namespace, pod) + sum(kube_pod_container_resource_requests_cpu_cores{job="kube-state-metrics"} * on (endpoint, instance, job, namespace, pod, service) group_left(phase) (kube_pod_status_phase{phase=~"^(Pending|Running)$"} == 1)) by (namespace, pod) * on (namespace, pod) group_left(label_name) label_replace(kube_pod_labels{job="kube-state-metrics"}, "pod_name", "$1", "pod", "(.*)") ) record: namespace_name:kube_pod_container_resource_requests_cpu_cores:sum + - expr: | + sum( + label_replace( + label_replace( + kube_pod_owner{job="kube-state-metrics", owner_kind="ReplicaSet"}, + "replicaset", "$1", "owner_name", "(.*)" + ) * on(replicaset, namespace) group_left(owner_name) kube_replicaset_owner{job="kube-state-metrics"}, + "workload", "$1", "owner_name", "(.*)" + ) + ) by (namespace, workload, pod) + labels: + workload_type: deployment + record: mixin_pod_workload + - expr: | + sum( + label_replace( + kube_pod_owner{job="kube-state-metrics", owner_kind="DaemonSet"}, + "workload", "$1", "owner_name", "(.*)" + ) + ) by (namespace, workload, pod) + labels: + workload_type: daemonset + record: mixin_pod_workload + - expr: | + sum( + label_replace( + kube_pod_owner{job="kube-state-metrics", owner_kind="StatefulSet"}, + "workload", "$1", "owner_name", "(.*)" + ) + ) by (namespace, workload, pod) + labels: + workload_type: statefulset + record: mixin_pod_workload - name: kube-scheduler.rules rules: - expr: | @@ -235,11 +268,11 @@ spec: ) record: node:node_disk_utilisation:avg_irate - expr: | - avg(irate(node_disk_io_time_weighted_seconds_total{job="node-exporter",device=~"nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+"}[1m]) / 1e3) + avg(irate(node_disk_io_time_weighted_seconds_total{job="node-exporter",device=~"nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+"}[1m])) record: :node_disk_saturation:avg_irate - expr: | avg by (node) ( - irate(node_disk_io_time_weighted_seconds_total{job="node-exporter",device=~"nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+"}[1m]) / 1e3 + irate(node_disk_io_time_weighted_seconds_total{job="node-exporter",device=~"nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+"}[1m]) * on (namespace, pod) group_left(node) node_namespace_pod:kube_pod_info: ) @@ -813,7 +846,7 @@ spec: - alert: KubeClientCertificateExpiration annotations: message: A client certificate used to authenticate to the apiserver is expiring - in less than 7 days. + in less than 7.0 days. runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeclientcertificateexpiration expr: | apiserver_client_certificate_expiration_seconds_count{job="apiserver"} > 0 and histogram_quantile(0.01, sum by (job, le) (rate(apiserver_client_certificate_expiration_seconds_bucket{job="apiserver"}[5m]))) < 604800 @@ -822,7 +855,7 @@ spec: - alert: KubeClientCertificateExpiration annotations: message: A client certificate used to authenticate to the apiserver is expiring - in less than 24 hours. + in less than 24.0 hours. runbook_url: https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeclientcertificateexpiration expr: | apiserver_client_certificate_expiration_seconds_count{job="apiserver"} > 0 and histogram_quantile(0.01, sum by (job, le) (rate(apiserver_client_certificate_expiration_seconds_bucket{job="apiserver"}[5m]))) < 86400 @@ -898,6 +931,46 @@ spec: for: 10m labels: severity: critical + - name: node-time + rules: + - alert: ClockSkewDetected + annotations: + message: Clock skew detected on node-exporter {{ $labels.namespace }}/{{ $labels.pod + }}. Ensure NTP is configured correctly on this host. + expr: | + node_ntp_offset_seconds{job="node-exporter"} < -0.03 or node_ntp_offset_seconds{job="node-exporter"} > 0.03 + for: 2m + labels: + severity: warning + - name: node-network + rules: + - alert: NetworkReceiveErrors + annotations: + message: Network interface "{{ $labels.device }}" showing receive errors on + node-exporter {{ $labels.namespace }}/{{ $labels.pod }}" + expr: | + rate(node_network_receive_errs_total{job="node-exporter",device!~"veth.+"}[2m]) > 0 + for: 2m + labels: + severity: warning + - alert: NetworkTransmitErrors + annotations: + message: Network interface "{{ $labels.device }}" showing transmit errors + on node-exporter {{ $labels.namespace }}/{{ $labels.pod }}" + expr: | + rate(node_network_transmit_errs_total{job="node-exporter",device!~"veth.+"}[2m]) > 0 + for: 2m + labels: + severity: warning + - alert: NodeNetworkInterfaceFlapping + annotations: + message: Network interface "{{ $labels.device }}" changing it's up status + often on node-exporter {{ $labels.namespace }}/{{ $labels.pod }}" + expr: | + changes(node_network_up{job="node-exporter",device!~"veth.+"}[2m]) > 2 + for: 2m + labels: + severity: warning - name: prometheus.rules rules: - alert: PrometheusConfigReloadFailed diff --git a/manifests/prometheus-serviceMonitorKubelet.yaml b/manifests/prometheus-serviceMonitorKubelet.yaml index 590a5cd..91da377 100644 --- a/manifests/prometheus-serviceMonitorKubelet.yaml +++ b/manifests/prometheus-serviceMonitorKubelet.yaml @@ -18,11 +18,6 @@ spec: honorLabels: true interval: 30s metricRelabelings: - - action: drop - regex: container_([a-z_]+); - sourceLabels: - - __name__ - - image - action: drop regex: container_(network_tcp_usage_total|network_udp_usage_total|tasks_state|cpu_load_average_10s) sourceLabels: