diff --git a/grafana/grafana-prometheus.yaml b/grafana/grafana-prometheus.yaml index c4a2208..345762b 100644 --- a/grafana/grafana-prometheus.yaml +++ b/grafana/grafana-prometheus.yaml @@ -23,6 +23,2158 @@ objects: labels: template: grafana-prometheus data: + openshift-cluster.json: |- + { + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "description": "Monitors Kubernetes cluster using Prometheus. Shows overall cluster CPU / Memory / Filesystem usage as well as individual pod, containers, systemd services statistics. Uses cAdvisor metrics only.", + "editable": true, + "gnetId": 315, + "graphTooltip": 0, + "hideControls": false, + "id": 2, + "links": [], + "refresh": "10s", + "rows": [ + { + "collapse": false, + "height": "200px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "fill": 1, + "grid": {}, + "height": "200px", + "id": 32, + "legend": { + "alignAsTable": false, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": false, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum (rate (container_network_receive_bytes_total{id=\"/\"}[5m]))", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "Received", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{id=\"/\"}[5m]))", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "Sent", + "metric": "network", + "refId": "B", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Network I/O pressure", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "transparent": false, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Network I/O pressure", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": true, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "prometheus", + "editable": true, + "error": false, + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "180px", + "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)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{id=\"/\",kubernetes_io_hostname=~\"^$Node$\"}) / sum (machine_memory_bytes{kubernetes_io_hostname=~\"^$Node$\"}) * 100", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "65, 90", + "title": "Cluster memory usage", + "transparent": false, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": true, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "180px", + "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": 4, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{id=\"/\"}[5m])) / sum (machine_cpu_cores{}) * 100", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "65, 90", + "title": "Cluster CPU usage (5m avg)", + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": true, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "180px", + "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": 4, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (container_fs_usage_bytes{device=~\"^/dev/([sv]d[a-z][1-9]|mapper/[-a-zA-Z0-9]+)$\",id=\"/\"}) / sum (container_fs_limit_bytes{device=~\"^/dev/([sv]d[a-z][1-9]|mapper/[-a-zA-Z0-9]+)$\",id=\"/\"}) * 100", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "", + "metric": "", + "refId": "A", + "step": 10 + } + ], + "thresholds": "65, 90", + "title": "Cluster filesystem usage", + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "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", + "decimals": 2, + "editable": true, + "error": false, + "format": "bytes", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 9, + "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": "20%", + "prefix": "", + "prefixFontSize": "20%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{id=\"/\",kubernetes_io_hostname=~\"^$Node$\"})", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Used", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "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", + "decimals": 2, + "editable": true, + "error": false, + "format": "bytes", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 10, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (machine_memory_bytes{kubernetes_io_hostname=~\"^$Node$\"})", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Total", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "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", + "decimals": 2, + "editable": true, + "error": false, + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 11, + "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": " cores", + "postfixFontSize": "30%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{id=\"/\"}[5m]))", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Used", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "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", + "decimals": 2, + "editable": true, + "error": false, + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 12, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": " cores", + "postfixFontSize": "30%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (machine_cpu_cores{kubernetes_io_hostname=~\"^$Node$\"})", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Total", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "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", + "decimals": 2, + "editable": true, + "error": false, + "format": "bytes", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 13, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (container_fs_usage_bytes{device=~\"^/dev/([sv]d[a-z][1-9]|mapper/[-a-zA-Z0-9]+)$\",id=\"/\"})", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Used", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "format": "bytes", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 14, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum (container_fs_limit_bytes{device=~\"^/dev/([sv]d[a-z][1-9]|mapper/[-a-zA-Z0-9]+)$\",id=\"/\"})", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Total", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Total usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 3, + "editable": true, + "error": false, + "fill": 0, + "grid": {}, + "height": "", + "id": 17, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name=~\"^k8s_.*\"}[5m])) by (pod_name)", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ pod_name }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Pods CPU usage (1m avg)", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "transparent": false, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "none", + "label": "cores", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Pods CPU usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 3, + "editable": true, + "error": false, + "fill": 0, + "grid": {}, + "height": "", + "id": 23, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{id=~\"^/system.slice.*\"}[5m])) by (systemd_service_name)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ systemd_service_name }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "System services CPU usage (5m avg)", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "none", + "label": "cores", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "System services CPU usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 3, + "editable": true, + "error": false, + "fill": 0, + "grid": {}, + "height": "", + "id": 24, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": false, + "hideZero": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name=~\"^k8s_.*\",container_name!=\"POD\"}[5m])) by (container_name, pod_name)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "pod: {{ pod_name }} | {{ container_name }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + }, + { + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name!~\"^k8s_.*\"}[5m])) by (kubernetes_io_hostname, name, image)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})", + "metric": "container_cpu", + "refId": "B", + "step": 10 + }, + { + "expr": "sum (rate (container_cpu_usage_seconds_total{rkt_container_name!=\"\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}", + "metric": "container_cpu", + "refId": "C", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Containers CPU usage (5m avg)", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "none", + "label": "cores", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Containers CPU usage", + "titleSize": "h6" + }, + { + "collapse": true, + "height": "500px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "datasource": "prometheus", + "decimals": 3, + "editable": true, + "error": false, + "fill": 0, + "grid": {}, + "id": 20, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{id!=\"/\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (id)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ id }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "All processes CPU usage (1m avg)", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "none", + "label": "cores", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "All processes CPU usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "fill": 0, + "grid": {}, + "id": 25, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{image!=\"\",name=~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}) by (pod_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ pod_name }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Pods memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "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": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Pods memory usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "fill": 0, + "grid": {}, + "id": 26, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{id=~\"^/system.slice/.*\"}) by (id)", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ id }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "System services memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "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": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "System services memory usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "fill": 0, + "grid": {}, + "id": 27, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{image!=\"\",name=~\"^k8s_.*\",container_name!=\"POD\",kubernetes_io_hostname=~\"^$Node$\"}) by (container_name, pod_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "pod: {{ pod_name }} | {{ container_name }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + }, + { + "expr": "sum (container_memory_working_set_bytes{image!=\"\",name!~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}) by (kubernetes_io_hostname, name, image)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})", + "metric": "container_memory_usage:sort_desc", + "refId": "B", + "step": 10 + }, + { + "expr": "sum (container_memory_working_set_bytes{rkt_container_name!=\"\",kubernetes_io_hostname=~\"^$Node$\"}) by (kubernetes_io_hostname, rkt_container_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}", + "metric": "container_memory_usage:sort_desc", + "refId": "C", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Containers memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "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": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Containers memory usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "500px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "fill": 0, + "grid": {}, + "id": 28, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{id!=\"/\",kubernetes_io_hostname=~\"^$Node$\"}) by (id)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ id }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "All processes memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "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": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "All processes memory usage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "fill": 1, + "grid": {}, + "id": 16, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name=~\"^k8s_.*\"}[5m])) by (pod_name)", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> {{ pod_name }}", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name=~\"^k8s_.*\"}[5m])) by (pod_name)", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- {{ pod_name }}", + "metric": "network", + "refId": "B", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Pods network I/O (5m avg)", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Pods network I/O", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "fill": 1, + "grid": {}, + "id": 30, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name=~\"^k8s_.*\"}[5m])) by (container_name, pod_name)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> pod: {{ pod_name }} | {{ container_name }}", + "metric": "network", + "refId": "B", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name=~\"^k8s_.*\"}[5m])) by (container_name, pod_name)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- pod: {{ pod_name }} | {{ container_name }}", + "metric": "network", + "refId": "D", + "step": 10 + }, + { + "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name!~\"^k8s_.*\"}[5m])) by (kubernetes_io_hostname, name, image)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name!~\"^k8s_.*\"}[5m])) by (kubernetes_io_hostname, name, image)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})", + "metric": "network", + "refId": "C", + "step": 10 + }, + { + "expr": "sum (rate (container_network_transmit_bytes_total{rkt_container_name!=\"\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}", + "metric": "network", + "refId": "E", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{rkt_container_name!=\"\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}", + "metric": "network", + "refId": "F", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Containers network I/O (5m avg)", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Containers network I/O", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "500px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "decimals": 2, + "editable": true, + "error": false, + "fill": 1, + "grid": {}, + "id": 29, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum (rate (container_network_receive_bytes_total{id!=\"/\"}[5m])) by (id)", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> {{ id }}", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{id!=\"/\"}[5m])) by (id)", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- {{ id }}", + "metric": "network", + "refId": "B", + "step": 10 + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "All processes network I/O (1m avg)", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "All processes network I/O", + "titleSize": "h6" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + "kubernetes" + ], + "templating": { + "list": [ + { + "allValue": ".*", + "current": { + "text": "All", + "value": "$__all" + }, + "datasource": "prometheus", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "Node", + "options": [], + "query": "label_values(kubernetes_io_hostname)", + "refresh": 1, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-5m", + "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 cluster monitoring (via Prometheus)", + "version": 2 + } - kind: ConfigMap apiVersion: v1 @@ -107,6 +2259,10 @@ objects: port: 443 protocol: TCP targetPort: 8443 + - name: prometheus + port: 9090 + protocol: TCP + targetPort: 9090 selector: app: prometheus sessionAffinity: None @@ -121,7 +2277,7 @@ objects: name: alerts namespace: ${NAMESPACE} annotations: - service.alpha.openshift.io/serving-cert-secret-name: alerts-tls + service.alpha.openshift.io/serving-cert-secret-name: prometheus-alerts-tls spec: ports: - name: oauth-proxy @@ -335,19 +2491,19 @@ objects: labels: template: grafana-prometheus data: - prometheus.additional.rules: |- - # Customize me ! prometheus.rules: |- groups: - name: example-rules interval: 30s # defaults to global interval rules: - prometheus.yml: |- - rule_files: - - 'prometheus.rules' - - 'prometheus.additional.rules' + prometheus.yml: |- + global: + scrape_interval: 30s + evaluation_interval: 30s + rule_files: + - '*.rules' # A scrape configuration for running Prometheus on a Kubernetes cluster. # This uses separate scrape configs for cluster components (i.e. API server, node)