diff --git a/operations/observability/mixins/workspace/dashboards/components/ws-daemon.json b/operations/observability/mixins/workspace/dashboards/components/ws-daemon.json index 85d9825e40e218..04f016afa675f8 100644 --- a/operations/observability/mixins/workspace/dashboards/components/ws-daemon.json +++ b/operations/observability/mixins/workspace/dashboards/components/ws-daemon.json @@ -1,69 +1,730 @@ { - "__inputs": [], - "__requires": [ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "editable": true, + "fiscalYearStartMonth": 0, + "graphTooltip": 1, + "id": null, + "links": [], + "liveNow": false, + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 65, + "panels": [], + "title": "Controller Manager", + "type": "row" + }, { - "type": "grafana", - "id": "grafana", - "name": "Grafana", - "version": "8.1.2" + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 8, + "x": 0, + "y": 1 + }, + "hiddenSeries": false, + "id": 67, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.3.6", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "sum(rate(workqueue_adds_total{job=\"ws-daemon\", pod=~\"$pod\"}[5m])) by (cluster, pod, name)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}} - {{pod}} - {{name}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Work Queue Add Rate", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:123", + "format": "ops", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:124", + "format": "ops", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } }, { - "type": "panel", - "id": "graph", - "name": "Graph (old)", - "version": "" + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 8, + "x": 8, + "y": 1 + }, + "hiddenSeries": false, + "id": 69, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.3.6", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "sum(workqueue_depth{job=\"ws-daemon\", cluster=~\"$cluster\"}) by (cluster, pod, name)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}} - {{pod}} - {{name}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Work Queue Depth", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:123", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:124", + "format": "ops", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } }, { - "type": "datasource", - "id": "prometheus", - "name": "Prometheus", - "version": "1.0.0" + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 8, + "x": 16, + "y": 1 + }, + "hiddenSeries": false, + "id": 71, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.3.6", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{job=\"ws-daemon\", pod=~\"$pod\"}[5m])) by (cluster, pod, name, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}} - {{pod}} - {{name}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Work Queue Latency", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:463", + "format": "s", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:464", + "format": "s", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 8, + "x": 0, + "y": 10 + }, + "hiddenSeries": false, + "id": 73, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.3.6", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "sum(rate(rest_client_requests_total{job=\"ws-daemon\", pod=~\"$pod\",code=~\"2..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "2xx", + "range": true, + "refId": "A" + }, + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "sum(rate(rest_client_requests_total{job=\"ws-daemon\", pod=~\"$pod\",code=~\"3..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "3xx", + "range": true, + "refId": "B" + }, + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "sum(rate(rest_client_requests_total{job=\"ws-daemon\", pod=~\"$pod\",code=~\"4..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "4xx", + "range": true, + "refId": "C" + }, + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "sum(rate(rest_client_requests_total{job=\"ws-daemon\", pod=~\"$pod\",code=~\"5..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "5xx", + "range": true, + "refId": "D" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Kube API Request Rate", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:705", + "format": "ops", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:706", + "format": "ops", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": ".*error" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": ".*requeue.*" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "blue", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": ".*success" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "green", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 8, + "y": 10 + }, + "id": 75, + "links": [], + "options": { + "legend": { + "calcs": [ + "lastNotNull" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "9.3.6", + "targets": [ + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "sum(rate(controller_runtime_reconcile_total{job=\"ws-daemon\", pod=~\"$pod\"}[5m])) by (cluster, pod, controller, result)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}} - {{pod}} - {{controller}} - {{result}}", + "range": true, + "refId": "A" + } + ], + "title": "Reconciliations", + "type": "timeseries" }, { - "type": "panel", - "id": "timeseries", - "name": "Time series", - "version": "" - } - ], - "annotations": { - "list": [ - { - "builtIn": 1, - "datasource": "-- Grafana --", - "enable": true, - "hide": true, - "iconColor": "rgba(0, 211, 255, 1)", - "name": "Annotations & Alerts", - "target": { - "limit": 100, - "matchAny": false, - "tags": [], - "type": "dashboard" + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, + "fieldConfig": { + "defaults": { + "links": [] }, - "type": "dashboard" + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 8, + "x": 16, + "y": 10 + }, + "hiddenSeries": false, + "id": 77, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.3.6", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "histogram_quantile(0.99, sum(rate(workqueue_work_duration_seconds_bucket{job=\"ws-daemon\", pod=~\"$pod\"}[5m])) by (cluster, pod, name, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}} - {{pod}} - {{name}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Reconcile Work Duration", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:463", + "format": "s", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:464", + "format": "s", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false } - ] - }, - "editable": true, - "gnetId": null, - "graphTooltip": 1, - "id": null, - "iteration": 1630955787054, - "links": [], - "panels": [ + }, { "collapsed": true, - "datasource": null, - "fieldConfig": { - "defaults": {}, - "overrides": [] + "datasource": { + "type": "datasource", + "uid": "grafana" }, "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 0 + "y": 19 }, "id": 52, "panels": [ @@ -72,7 +733,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "description": "", "fieldConfig": { "defaults": { @@ -87,7 +750,7 @@ "h": 9, "w": 12, "x": 0, - "y": 1 + "y": 57 }, "hiddenSeries": false, "id": 54, @@ -120,6 +783,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum(rate(grpc_server_handled_total{job=\"ws-daemon\", cluster=~\"$cluster\", grpc_method=~\"$grpc_method\"}[1m])) by (grpc_code, grpc_method, cluster)", "interval": "", "legendFormat": "{{cluster}} - {{grpc_method}} - {{grpc_code}}", @@ -128,9 +794,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "gRPC calls handled (Server-side)", "tooltip": { "shared": true, @@ -139,9 +803,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -149,24 +811,18 @@ { "decimals": 2, "format": "reqps", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "reqps", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -174,7 +830,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "description": "", "fieldConfig": { "defaults": { @@ -189,7 +847,7 @@ "h": 9, "w": 12, "x": 12, - "y": 1 + "y": 57 }, "hiddenSeries": false, "id": 56, @@ -228,6 +886,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum(rate(grpc_server_started_total{job=\"ws-daemon\", cluster=~\"$cluster\", grpc_method=~\"$grpc_method\"}[1m])) by (grpc_code, grpc_method, cluster)", "interval": "", "legendFormat": "{{cluster}} - {{grpc_method}} - Started", @@ -235,6 +896,9 @@ "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "sum(rate(grpc_server_handled_total{job=\"ws-daemon\", cluster=~\"$cluster\", grpc_method=~\"$grpc_method\"}[1m])) by (grpc_method, cluster)", "hide": false, "interval": "", @@ -244,9 +908,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "gRPC start and finish rate", "tooltip": { "shared": true, @@ -255,9 +917,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -265,24 +925,18 @@ { "decimals": 2, "format": "reqps", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "reqps", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -290,7 +944,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "description": "", "fieldConfig": { "defaults": { @@ -305,7 +961,7 @@ "h": 9, "w": 12, "x": 0, - "y": 10 + "y": 66 }, "hiddenSeries": false, "id": 58, @@ -340,6 +996,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "histogram_quantile(0.99, \n sum(\n rate(grpc_server_handling_seconds_bucket{job=\"ws-daemon\", cluster=~\"$cluster\", grpc_method=~\"$grpc_method\"}[5m])\n ) by (cluster, grpc_method, le)\n)", "interval": "", "legendFormat": "{{cluster}} - {{grpc_method}} - 99th Percentile", @@ -347,6 +1006,9 @@ "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "histogram_quantile(0.95, \n sum(\n rate(grpc_server_handling_seconds_bucket{job=\"ws-daemon\", cluster=~\"$cluster\", grpc_method=~\"$grpc_method\"}[5m])\n ) by (cluster, grpc_method, le)\n)", "interval": "", "legendFormat": "{{cluster}} - {{grpc_method}} - 95th Percentile", @@ -354,6 +1016,9 @@ "refId": "B" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "histogram_quantile(0.50, \n sum(\n rate(grpc_server_handling_seconds_bucket{job=\"ws-daemon\", cluster=~\"$cluster\", grpc_method=~\"$grpc_method\"}[5m])\n ) by (cluster, grpc_method, le)\n)", "interval": "", "legendFormat": "{{cluster}} - {{grpc_method}} - 50th Percentile", @@ -361,6 +1026,9 @@ "refId": "C" }, { + "datasource": { + "uid": "$datasource" + }, "expr": " sum(\n rate(grpc_server_handling_seconds_sum{job=\"ws-daemon\", cluster=~\"$cluster\", grpc_method=\"$grpc_method\"}[5m])\n ) by (cluster, grpc_method)\n /\n sum(\n rate(grpc_server_handling_seconds_count{job=\"ws-daemon\", cluster=~\"$cluster\", grpc_method=\"$grpc_method\"}[5m])\n ) by (cluster, grpc_method)", "hide": false, "interval": "", @@ -370,9 +1038,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "$grpc_method: Time handling gRPC calls", "tooltip": { "shared": true, @@ -381,65 +1047,78 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, "yaxes": [ { "format": "s", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "reqps", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } } ], + "targets": [ + { + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "refId": "A" + } + ], "title": "gRPC Metrics", "type": "row" }, { "collapsed": false, - "datasource": null, - "fieldConfig": { - "defaults": {}, - "overrides": [] + "datasource": { + "type": "datasource", + "uid": "grafana" }, "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 1 + "y": 20 }, "id": 16, "panels": [], + "targets": [ + { + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "refId": "A" + } + ], "title": "Pod Metrics", "type": "row" }, { - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fieldConfig": { "defaults": { "color": { "mode": "palette-classic" }, "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, @@ -488,21 +1167,26 @@ "h": 7, "w": 10, "x": 0, - "y": 2 + "y": 21 }, "id": 63, "options": { "legend": { "calcs": [], "displayMode": "table", - "placement": "right" + "placement": "right", + "showLegend": true }, "tooltip": { - "mode": "single" + "mode": "single", + "sort": "none" } }, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "exemplar": true, "expr": "kube_pod_container_info{cluster=~\"$cluster\", pod=~\"$pod\", image=~\".+\", container=\"ws-daemon\"}", "interval": "", @@ -519,7 +1203,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "decimals": 2, "fill": 1, "fillGradient": 0, @@ -527,7 +1213,7 @@ "h": 7, "w": 7, "x": 10, - "y": 2 + "y": 21 }, "hiddenSeries": false, "id": 2, @@ -550,7 +1236,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -560,6 +1246,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum(\n rate(container_cpu_usage_seconds_total{container!=\"POD\", pod!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}} - Cores being used", @@ -568,9 +1257,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "CPU Utilization", "tooltip": { "shared": true, @@ -579,9 +1266,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -589,24 +1274,18 @@ { "decimals": 2, "format": "short", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -614,7 +1293,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "description": "Saturation > 100% means that the container is requesting more than its limits.\n\nKubernetes will start to throttle CPU when that happens. That's a sign of degraded performance.\n\n'No Data' indicates that the pod has no CPU limits.", "fill": 1, "fillGradient": 0, @@ -622,7 +1303,7 @@ "h": 7, "w": 7, "x": 17, - "y": 2 + "y": 21 }, "hiddenSeries": false, "id": 4, @@ -642,7 +1323,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -657,6 +1338,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum(\n rate(container_cpu_usage_seconds_total{container!=\"POD\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)\n/\nsum(\n kube_pod_container_resource_limits_cpu_cores{container!=\"POD\", cluster=\"$cluster\", pod=~\"$pod\"}\n) by (pod, cluster, node)\n", "interval": "", "legendFormat": "{{cluster}} - {{node} - {{pod}} - CPU Saturation", @@ -664,6 +1348,9 @@ "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "sum(\nrate(container_cpu_cfs_throttled_seconds_total{container!=\"POD\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod)", "interval": "", "legendFormat": "{{pod}} - CPU Throttles", @@ -672,9 +1359,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "CPU Saturation", "tooltip": { "shared": true, @@ -683,9 +1368,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -693,24 +1376,18 @@ { "decimals": 2, "format": "percentunit", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "s", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -718,14 +1395,16 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fill": 1, "fillGradient": 0, "gridPos": { "h": 7, "w": 12, "x": 0, - "y": 9 + "y": 28 }, "hiddenSeries": false, "id": 6, @@ -746,7 +1425,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -756,6 +1435,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum(container_memory_working_set_bytes{container!=\"POD\", container!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}) by (pod, cluster, node)", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}}", @@ -764,9 +1446,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Memory Utilization", "tooltip": { "shared": true, @@ -775,33 +1455,25 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, "yaxes": [ { "format": "bytes", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -809,7 +1481,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "decimals": 4, "description": "Memory can't be throttled. When a container reaches 100% of its memory limits, Kubernetes will kill the container and restart it.\n\n'No Data' indicates that the pod doesn't have Memory limits.", "fill": 1, @@ -818,7 +1492,7 @@ "h": 7, "w": 12, "x": 12, - "y": 9 + "y": 28 }, "hiddenSeries": false, "id": 8, @@ -839,7 +1513,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -849,6 +1523,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum(\nrate(container_memory_working_set_bytes{container!=\"POD\", container!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)\n/\nsum(\n kube_pod_container_resource_limits{container!=\"POD\", cluster=\"$cluster\", pod=~\"$pod\", resource=\"memory\"}\n) by (pod, cluster, node)\n", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}} - Memory Saturation", @@ -857,9 +1534,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Memory Saturation", "tooltip": { "shared": true, @@ -868,9 +1543,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -878,24 +1551,18 @@ { "decimals": 2, "format": "percentunit", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -903,14 +1570,16 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fill": 1, "fillGradient": 0, "gridPos": { "h": 7, "w": 8, "x": 0, - "y": 16 + "y": 35 }, "hiddenSeries": false, "id": 10, @@ -931,7 +1600,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -941,6 +1610,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum (\n rate(container_network_receive_bytes_total{container!=\"POD\", pod!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}} - Received", @@ -948,6 +1620,9 @@ "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "sum (\n rate(container_network_transmit_bytes_total{container!=\"POD\", pod!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}} - Transmitted", @@ -956,9 +1631,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Network Utilization", "tooltip": { "shared": true, @@ -967,33 +1640,25 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, "yaxes": [ { "format": "binBps", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1001,14 +1666,16 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fill": 1, "fillGradient": 0, "gridPos": { "h": 7, "w": 8, "x": 8, - "y": 16 + "y": 35 }, "hiddenSeries": false, "id": 44, @@ -1030,7 +1697,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -1040,6 +1707,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum (\n rate(container_network_receive_packets_dropped_total{container!=\"POD\", pod!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}} - Receive", @@ -1047,6 +1717,9 @@ "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "sum (\n rate(container_network_transmit_packets_dropped_total{container!=\"POD\", pod!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}} - Transmit", @@ -1055,9 +1728,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Network Saturation (Packets Dropped)", "tooltip": { "shared": true, @@ -1066,9 +1737,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -1076,24 +1745,18 @@ { "decimals": 2, "format": "pps", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1101,14 +1764,16 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fill": 1, "fillGradient": 0, "gridPos": { "h": 7, "w": 8, "x": 16, - "y": 16 + "y": 35 }, "hiddenSeries": false, "id": 46, @@ -1129,7 +1794,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -1139,6 +1804,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "sum (\n rate(container_network_receive_errors_total{container!=\"POD\", pod!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}} - Received", @@ -1146,6 +1814,9 @@ "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "sum (\n rate(container_network_transmit_errors_total{container!=\"POD\", pod!=\"\", cluster=~\"$cluster\", node=~\"$node\", pod=~\"$pod\"}[1m])\n) by (pod, cluster, node)", "interval": "", "legendFormat": "{{cluster}} - {{node}} - {{pod}} - Transmitted", @@ -1154,9 +1825,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Network Errors", "tooltip": { "shared": true, @@ -1165,9 +1834,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -1175,24 +1842,18 @@ { "decimals": 2, "format": "Errors/s", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1200,7 +1861,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "decimals": 4, "description": "", "fill": 1, @@ -1209,7 +1872,7 @@ "h": 9, "w": 8, "x": 0, - "y": 23 + "y": 42 }, "hiddenSeries": false, "id": 36, @@ -1230,7 +1893,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -1240,6 +1903,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "rate(kube_pod_container_status_restarts_total{cluster=~\"$cluster\", pod=~\"$pod\"}[1m])", "interval": "", "legendFormat": "{{cluster}} - {{kubernetes_pod_node_name}} - {{pod}} ", @@ -1248,9 +1914,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Pod Restarts", "tooltip": { "shared": true, @@ -1259,9 +1923,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -1269,24 +1931,18 @@ { "decimals": 2, "format": "short", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1294,7 +1950,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "decimals": 0, "description": "", "fill": 1, @@ -1303,7 +1961,7 @@ "h": 9, "w": 8, "x": 8, - "y": 23 + "y": 42 }, "hiddenSeries": false, "id": 42, @@ -1324,7 +1982,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -1334,6 +1992,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "kube_pod_container_status_running{cluster=~\"$cluster\", pod=~\"$pod\"} == 1 ", "interval": "", "legendFormat": "{{pod}} - RUNNING", @@ -1341,6 +2002,9 @@ "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "(\n sum by (pod) (kube_pod_container_status_terminated{cluster=~\"$cluster\", pod=~\"$pod\"}) == 1\n) * on(pod) group_left(reason) (\n sum by (pod, reason) (kube_pod_container_status_terminated_reason{cluster=~\"$cluster\", pod=~\"$pod\"}) == 1\n)", "interval": "", "legendFormat": "{{pod}} - TERMINATED -> {{reason}}", @@ -1348,6 +2012,9 @@ "refId": "B" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "(\n sum by (pod) (kube_pod_container_status_waiting{cluster=~\"$cluster\", pod=~\"$pod\"}) == 1\n) * on(pod) group_left(reason) (\n sum by (pod, reason) (kube_pod_container_status_waiting_reason{cluster=~\"$cluster\", pod=~\"$pod\"}) == 1\n)", "interval": "", "legendFormat": "{{pod}} - WAITING -> {{reason}}", @@ -1356,9 +2023,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Pod Status", "tooltip": { "shared": true, @@ -1367,9 +2032,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -1377,24 +2040,17 @@ { "decimals": 0, "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1402,7 +2058,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "decimals": 0, "fill": 1, "fillGradient": 0, @@ -1410,7 +2068,7 @@ "h": 9, "w": 8, "x": 16, - "y": 23 + "y": 42 }, "hiddenSeries": false, "id": 40, @@ -1431,7 +2089,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.1.2", + "pluginVersion": "9.3.6", "pointradius": 2, "points": false, "renderer": "flot", @@ -1441,6 +2099,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "kube_daemonset_status_desired_number_scheduled{cluster=~\"$cluster\", daemonset=\"ws-daemon\"}", "interval": "", "legendFormat": "{{cluster}} - {{daemonset}} - Desired", @@ -1448,6 +2109,9 @@ "refId": "C" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "kube_daemonset_status_number_available{cluster=~\"$cluster\", daemonset=\"ws-daemon\"}", "interval": "", "legendFormat": "{{cluster}} - {{daemonset}} - Available replicas", @@ -1455,6 +2119,9 @@ "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "kube_daemonset_status_number_unavailable{cluster=~\"$cluster\", daemonset=\"ws-daemon\"}", "interval": "", "legendFormat": "{{cluster}} - {{daemonset}} - Unvailable replicas", @@ -1463,9 +2130,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Replicas availability", "tooltip": { "shared": true, @@ -1474,9 +2139,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, @@ -1484,38 +2147,30 @@ { "decimals": 0, "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { "collapsed": true, - "datasource": null, - "fieldConfig": { - "defaults": {}, - "overrides": [] + "datasource": { + "type": "datasource", + "uid": "grafana" }, "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 32 + "y": 51 }, "id": 22, "panels": [ @@ -1524,7 +2179,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fieldConfig": { "defaults": { "custom": {} @@ -1537,7 +2194,7 @@ "h": 9, "w": 12, "x": 0, - "y": 3 + "y": 59 }, "hiddenSeries": false, "id": 32, @@ -1568,6 +2225,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "process_resident_memory_bytes{cluster=~\"$cluster\", job=\"ws-daemon\", pod=~\"$pod\"}", "interval": "", "legendFormat": "{{cluster}} - {{pod}}", @@ -1576,9 +2236,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Memory Usage (as seen by the runtime process)", "tooltip": { "shared": true, @@ -1587,33 +2245,25 @@ }, "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": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1621,7 +2271,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "decimals": 2, "editable": true, "error": false, @@ -1637,7 +2289,7 @@ "h": 9, "w": 12, "x": 12, - "y": 3 + "y": 59 }, "hiddenSeries": false, "id": 34, @@ -1671,6 +2323,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "rate(process_cpu_seconds_total{cluster=~\"$cluster\", job=\"ws-daemon\", pod=~\"$pod\"}[5m])", "interval": "", "intervalFactor": 2, @@ -1681,9 +2336,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "CPU Usage (as seen by the runtime process)", "tooltip": { "msResolution": false, @@ -1693,9 +2346,7 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [ "avg" @@ -1705,24 +2356,18 @@ { "decimals": 2, "format": "none", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1730,7 +2375,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fieldConfig": { "defaults": { "custom": {} @@ -1743,7 +2390,7 @@ "h": 9, "w": 12, "x": 0, - "y": 12 + "y": 68 }, "hiddenSeries": false, "id": 26, @@ -1774,6 +2421,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "go_memstats_heap_sys_bytes{cluster=~\"$cluster\", job=\"ws-daemon\", pod=~\"$pod\"}", "interval": "", "legendFormat": "{{cluster}} - {{pod}}", @@ -1782,9 +2432,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Heap Usage", "tooltip": { "shared": true, @@ -1793,33 +2441,25 @@ }, "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": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1827,7 +2467,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fieldConfig": { "defaults": { "custom": {}, @@ -1841,7 +2483,7 @@ "h": 9, "w": 12, "x": 12, - "y": 12 + "y": 68 }, "hiddenSeries": false, "id": 48, @@ -1866,7 +2508,6 @@ "pointradius": 2, "points": false, "renderer": "flot", - "repeat": null, "repeatDirection": "h", "seriesOverrides": [ { @@ -1879,6 +2520,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "rate(go_memstats_alloc_bytes_total{cluster=~\"$cluster\", job=\"ws-daemon\", pod=~\"$pod\"}[5m])", "interval": "", "legendFormat": "{{cluster}} - {{pod}}", @@ -1886,9 +2530,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Memory Allocation rate", "tooltip": { "shared": true, @@ -1897,33 +2539,25 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, "yaxes": [ { "format": "binBps", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -1931,7 +2565,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fieldConfig": { "defaults": { "custom": {} @@ -1944,7 +2580,7 @@ "h": 9, "w": 12, "x": 0, - "y": 21 + "y": 77 }, "hiddenSeries": false, "id": 24, @@ -1975,6 +2611,9 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "go_goroutines{cluster=~\"$cluster\", job=\"ws-daemon\", pod=~\"$pod\"}", "interval": "", "legendFormat": "{{cluster}} - {{pod}}", @@ -1983,9 +2622,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Go Routines", "tooltip": { "shared": true, @@ -1994,33 +2631,25 @@ }, "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": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } }, { @@ -2028,7 +2657,9 @@ "bars": false, "dashLength": 10, "dashes": false, - "datasource": "$datasource", + "datasource": { + "uid": "$datasource" + }, "fieldConfig": { "defaults": { "custom": {} @@ -2041,7 +2672,7 @@ "h": 9, "w": 12, "x": 12, - "y": 21 + "y": 77 }, "hiddenSeries": false, "id": 50, @@ -2072,18 +2703,27 @@ "steppedLine": false, "targets": [ { + "datasource": { + "uid": "$datasource" + }, "expr": "go_gc_duration_seconds{cluster=~\"$cluster\", job=\"ws-daemon\", pod=~\"$pod\", quantile=\"0.5\"}", "interval": "", "legendFormat": "{{cluster}} - {{pod}} - GC Duration - 50th percentile", "refId": "A" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "go_gc_duration_seconds{cluster=~\"$cluster\", job=\"ws-daemon\", pod=~\"$pod\", quantile=\"0.75\"}", "interval": "", "legendFormat": "{{cluster}} - {{pod}} - GC Duration - 75th percentile", "refId": "B" }, { + "datasource": { + "uid": "$datasource" + }, "expr": "go_gc_duration_seconds{cluster=~\"$cluster\", job=\"ws-daemon\", pod=~\"$pod\", quantile=\"1\"}", "interval": "", "legendFormat": "{{cluster}} - {{pod}} - GC Duration - 100th percentile", @@ -2091,9 +2731,7 @@ } ], "thresholds": [], - "timeFrom": null, "timeRegions": [], - "timeShift": null, "title": "Garbage collection time", "tooltip": { "shared": true, @@ -2102,42 +2740,43 @@ }, "type": "graph", "xaxis": { - "buckets": null, "mode": "time", - "name": null, "show": true, "values": [] }, "yaxes": [ { "format": "s", - "label": null, "logBase": 1, - "max": null, "min": "0", "show": true }, { "format": "short", - "label": null, "logBase": 1, - "max": null, - "min": null, "show": true } ], "yaxis": { - "align": false, - "alignLevel": null + "align": false } } ], + "targets": [ + { + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "refId": "A" + } + ], "title": "Go Runtime Metrics", "type": "row" } ], "refresh": "30s", - "schemaVersion": 30, + "schemaVersion": 37, "style": "dark", "tags": [ "gitpod-mixin" @@ -2145,12 +2784,16 @@ "templating": { "list": [ { - "allValue": null, - "current": {}, - "datasource": "$datasource", + "current": { + "selected": false, + "text": "All", + "value": "$__all" + }, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, "definition": "label_values(up{job=\"ws-daemon\"}, cluster)", - "description": null, - "error": null, "hide": 0, "includeAll": true, "label": "Cluster", @@ -2171,12 +2814,16 @@ "useTags": false }, { - "allValue": null, - "current": {}, - "datasource": "$datasource", + "current": { + "selected": false, + "text": "All", + "value": "$__all" + }, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, "definition": "label_values(container_cpu_usage_seconds_total{cluster=~\"$cluster\", pod=~\"ws-daemon.*\"}, node)", - "description": null, - "error": null, "hide": 0, "includeAll": true, "label": "Node", @@ -2197,12 +2844,16 @@ "useTags": false }, { - "allValue": null, - "current": {}, - "datasource": "$datasource", + "current": { + "selected": false, + "text": "All", + "value": "$__all" + }, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, "definition": "label_values(container_cpu_usage_seconds_total{cluster=~\"$cluster\", node=~\"$node\", pod=~\"ws-daemon.*\"}, pod)", - "description": null, - "error": null, "hide": 0, "includeAll": true, "label": "Pod", @@ -2223,12 +2874,16 @@ "useTags": false }, { - "allValue": null, - "current": {}, - "datasource": "$datasource", + "current": { + "selected": false, + "text": "All", + "value": "$__all" + }, + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, "definition": "label_values(grpc_server_handled_total{job=\"ws-daemon\", cluster=~\"$cluster\"}, grpc_method)", - "description": null, - "error": null, "hide": 0, "includeAll": true, "label": "gRPC Method", @@ -2254,11 +2909,8 @@ "text": "VictoriaMetrics", "value": "VictoriaMetrics" }, - "description": null, - "error": null, "hide": 2, "includeAll": false, - "label": null, "multi": false, "name": "datasource", "options": [], @@ -2279,4 +2931,4 @@ "title": "Gitpod / Component / ws-daemon", "uid": "ws-daemon", "version": 1 -} \ No newline at end of file +} diff --git a/operations/observability/mixins/workspace/dashboards/components/ws-manager-mk2.json b/operations/observability/mixins/workspace/dashboards/components/ws-manager-mk2.json index bf1675ab128b6e..ae0fb0cfbca322 100644 --- a/operations/observability/mixins/workspace/dashboards/components/ws-manager-mk2.json +++ b/operations/observability/mixins/workspace/dashboards/components/ws-manager-mk2.json @@ -177,59 +177,108 @@ } }, { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, "datasource": { "type": "prometheus", "uid": "$datasource" }, - "decimals": 2, "fieldConfig": { "defaults": { - "links": [] + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "stepAfter", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "ops" }, - "overrides": [] + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "Failures: .*" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + } + ] + } + ] }, - "fill": 1, - "fillGradient": 0, "gridPos": { "h": 11, "w": 8, "x": 8, "y": 1 }, - "hiddenSeries": false, "id": 48, - "legend": { - "alignAsTable": true, - "avg": true, - "current": true, - "max": true, - "min": false, - "rightSide": false, - "show": true, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, "links": [], - "nullPointMode": "null", "options": { - "alertThreshold": true + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } }, - "percentage": false, "pluginVersion": "9.3.6", - "pointradius": 2, - "points": false, - "renderer": "flot", "repeatDirection": "h", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": true, "targets": [ { "datasource": { @@ -242,41 +291,22 @@ "legendFormat": "{{cluster}} - {{type}}", "range": true, "refId": "B" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Workspace Starts", - "tooltip": { - "shared": true, - "sort": 2, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "$$hashKey": "object:198", - "decimals": 2, - "format": "ops", - "logBase": 1, - "min": "0", - "show": true }, { - "$$hashKey": "object:199", - "format": "reqps", - "logBase": 1, - "show": true + "datasource": { + "type": "prometheus", + "uid": "$datasource" + }, + "editorMode": "code", + "expr": "sum(rate(gitpod_ws_manager_mk2_workspace_starts_failure_total{cluster=~\"$cluster\"}[1m])) by (cluster, type)", + "hide": false, + "legendFormat": "Failures: {{cluster}} - {{type}}", + "range": true, + "refId": "A" } ], - "yaxis": { - "align": false - } + "title": "Workspace Starts", + "type": "timeseries" }, { "aliasColors": {}, @@ -1741,7 +1771,7 @@ "h": 1, "w": 24, "x": 0, - "y": 104 + "y": 77 }, "id": 16, "panels": [], @@ -1802,7 +1832,8 @@ "mode": "absolute", "steps": [ { - "color": "green" + "color": "green", + "value": null }, { "color": "red", @@ -1817,7 +1848,7 @@ "h": 7, "w": 10, "x": 0, - "y": 105 + "y": 78 }, "id": 76, "options": { @@ -1865,7 +1896,7 @@ "h": 7, "w": 7, "x": 10, - "y": 105 + "y": 78 }, "hiddenSeries": false, "id": 2, @@ -1957,7 +1988,7 @@ "h": 7, "w": 7, "x": 17, - "y": 105 + "y": 78 }, "hiddenSeries": false, "id": 4, @@ -2061,7 +2092,7 @@ "h": 7, "w": 12, "x": 0, - "y": 112 + "y": 85 }, "hiddenSeries": false, "id": 6, @@ -2151,7 +2182,7 @@ "h": 7, "w": 12, "x": 12, - "y": 112 + "y": 85 }, "hiddenSeries": false, "id": 8, @@ -2240,7 +2271,7 @@ "h": 7, "w": 8, "x": 0, - "y": 119 + "y": 92 }, "hiddenSeries": false, "id": 10, @@ -2338,7 +2369,7 @@ "h": 7, "w": 8, "x": 8, - "y": 119 + "y": 92 }, "hiddenSeries": false, "id": 57, @@ -2436,7 +2467,7 @@ "h": 7, "w": 8, "x": 16, - "y": 119 + "y": 92 }, "hiddenSeries": false, "id": 59, @@ -2535,7 +2566,7 @@ "h": 9, "w": 8, "x": 0, - "y": 126 + "y": 99 }, "hiddenSeries": false, "id": 36, @@ -2626,7 +2657,7 @@ "h": 9, "w": 8, "x": 8, - "y": 126 + "y": 99 }, "hiddenSeries": false, "id": 55, @@ -2736,7 +2767,7 @@ "h": 9, "w": 8, "x": 16, - "y": 126 + "y": 99 }, "hiddenSeries": false, "id": 40, @@ -2846,7 +2877,7 @@ "h": 1, "w": 24, "x": 0, - "y": 135 + "y": 108 }, "id": 22, "panels": [], @@ -2877,7 +2908,7 @@ "h": 9, "w": 12, "x": 0, - "y": 136 + "y": 109 }, "hiddenSeries": false, "id": 32, @@ -2971,7 +3002,7 @@ "h": 9, "w": 12, "x": 12, - "y": 136 + "y": 109 }, "hiddenSeries": false, "id": 34, @@ -3071,7 +3102,7 @@ "h": 9, "w": 12, "x": 0, - "y": 145 + "y": 118 }, "hiddenSeries": false, "id": 26, @@ -3168,7 +3199,7 @@ "h": 9, "w": 12, "x": 12, - "y": 145 + "y": 118 }, "hiddenSeries": false, "id": 61, @@ -3265,7 +3296,7 @@ "h": 9, "w": 12, "x": 0, - "y": 154 + "y": 127 }, "hiddenSeries": false, "id": 24, @@ -3356,7 +3387,7 @@ "h": 9, "w": 12, "x": 12, - "y": 154 + "y": 127 }, "hiddenSeries": false, "id": 63, @@ -3454,7 +3485,7 @@ } } ], - "refresh": "30s", + "refresh": "5s", "schemaVersion": 37, "style": "dark", "tags": [ @@ -3464,9 +3495,13 @@ "list": [ { "current": { - "selected": false, - "text": "All", - "value": "$__all" + "selected": true, + "text": [ + "ephemeral-wv" + ], + "value": [ + "ephemeral-wv" + ] }, "datasource": { "type": "prometheus", @@ -3494,9 +3529,13 @@ }, { "current": { - "selected": false, - "text": "All", - "value": "$__all" + "selected": true, + "text": [ + "All" + ], + "value": [ + "$__all" + ] }, "datasource": { "type": "prometheus", @@ -3524,9 +3563,13 @@ }, { "current": { - "selected": false, - "text": "All", - "value": "$__all" + "selected": true, + "text": [ + "All" + ], + "value": [ + "$__all" + ] }, "datasource": { "type": "prometheus", @@ -3554,9 +3597,13 @@ }, { "current": { - "selected": false, - "text": "All", - "value": "$__all" + "selected": true, + "text": [ + "All" + ], + "value": [ + "$__all" + ] }, "datasource": { "type": "prometheus",