apiVersion: v1 items: - apiVersion: v1 data: coredns-dashboard.json: |- { "annotations": { "list": [ { "builtIn": 1, "datasource": "-- Grafana --", "enable": true, "hide": true, "iconColor": "rgba(0, 211, 255, 1)", "name": "Annotations & Alerts", "type": "dashboard" } ] }, "description": "A dashboard for the CoreDNS DNS server.", "editable": true, "gnetId": 5926, "graphTooltip": 0, "id": 14, "iteration": 1549319226130, "links": [ ], "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 8, "x": 0, "y": 0 }, "id": 1, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "total", "yaxis": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(coredns_dns_request_count_total{instance=~\"$instance\"}[5m])) by (proto)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{proto}}", "refId": "A", "step": 60 }, { "expr": "sum(rate(coredns_dns_request_count_total{instance=~\"$instance\"}[5m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "total", "refId": "B", "step": 60 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Requests (total)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 8, "x": 8, "y": 0 }, "id": 12, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "total", "yaxis": 2 }, { "alias": "other", "yaxis": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(coredns_dns_request_type_count_total{instance=~\"$instance\"}[5m])) by (type)", "intervalFactor": 2, "legendFormat": "{{type}}", "refId": "A", "step": 60 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Requests (by qtype)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 8, "x": 16, "y": 0 }, "id": 2, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "total", "yaxis": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(coredns_dns_request_count_total{instance=~\"$instance\"}[5m])) by (zone)", "intervalFactor": 2, "legendFormat": "{{zone}}", "refId": "A", "step": 60 }, { "expr": "sum(rate(coredns_dns_request_count_total{instance=~\"$instance\"}[5m]))", "intervalFactor": 2, "legendFormat": "total", "refId": "B", "step": 60 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Requests (by zone)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 0, "y": 7 }, "id": 10, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "total", "yaxis": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(coredns_dns_request_do_count_total{instance=~\"$instance\"}[5m]))", "intervalFactor": 2, "legendFormat": "DO", "refId": "A", "step": 40 }, { "expr": "sum(rate(coredns_dns_request_count_total{instance=~\"$instance\"}[5m]))", "intervalFactor": 2, "legendFormat": "total", "refId": "B", "step": 40 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Requests (DO bit)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "pps", "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 6, "x": 12, "y": 7 }, "id": 9, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "tcp:90", "yaxis": 2 }, { "alias": "tcp:99 ", "yaxis": 2 }, { "alias": "tcp:50", "yaxis": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "histogram_quantile(0.99, sum(rate(coredns_dns_request_size_bytes_bucket{instance=~\"$instance\",proto=\"udp\"}[5m])) by (le,proto))", "intervalFactor": 2, "legendFormat": "{{proto}}:99 ", "refId": "A", "step": 60 }, { "expr": "histogram_quantile(0.90, sum(rate(coredns_dns_request_size_bytes_bucket{instance=~\"$instance\",proto=\"udp\"}[5m])) by (le,proto))", "intervalFactor": 2, "legendFormat": "{{proto}}:90", "refId": "B", "step": 60 }, { "expr": "histogram_quantile(0.50, sum(rate(coredns_dns_request_size_bytes_bucket{instance=~\"$instance\",proto=\"udp\"}[5m])) by (le,proto))", "intervalFactor": 2, "legendFormat": "{{proto}}:50", "refId": "C", "step": 60 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Requests (size, udp)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 6, "x": 18, "y": 7 }, "id": 14, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "tcp:90", "yaxis": 1 }, { "alias": "tcp:99 ", "yaxis": 1 }, { "alias": "tcp:50", "yaxis": 1 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "histogram_quantile(0.99, sum(rate(coredns_dns_request_size_bytes_bucket{instance=~\"$instance\",proto=\"tcp\"}[5m])) by (le,proto))", "intervalFactor": 2, "legendFormat": "{{proto}}:99 ", "refId": "A", "step": 60 }, { "expr": "histogram_quantile(0.90, sum(rate(coredns_dns_request_size_bytes_bucket{instance=~\"$instance\",proto=\"tcp\"}[5m])) by (le,proto))", "intervalFactor": 2, "legendFormat": "{{proto}}:90", "refId": "B", "step": 60 }, { "expr": "histogram_quantile(0.50, sum(rate(coredns_dns_request_size_bytes_bucket{instance=~\"$instance\",proto=\"tcp\"}[5m])) by (le,proto))", "intervalFactor": 2, "legendFormat": "{{proto}}:50", "refId": "C", "step": 60 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Requests (size,tcp)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 0, "y": 14 }, "id": 5, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(coredns_dns_response_rcode_count_total{instance=~\"$instance\"}[5m])) by (rcode)", "intervalFactor": 2, "legendFormat": "{{rcode}}", "refId": "A", "step": 40 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Responses (by rcode)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 12, "y": 14 }, "id": 3, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "histogram_quantile(0.99, sum(rate(coredns_dns_request_duration_milliseconds_bucket{instance=~\"$instance\"}[5m])) by (le, job))", "intervalFactor": 2, "legendFormat": "99%", "refId": "A", "step": 40 }, { "expr": "histogram_quantile(0.90, sum(rate(coredns_dns_request_duration_milliseconds_bucket{instance=~\"$instance\"}[5m])) by (le))", "intervalFactor": 2, "legendFormat": "90%", "refId": "B", "step": 40 }, { "expr": "histogram_quantile(0.50, sum(rate(coredns_dns_request_duration_milliseconds_bucket{instance=~\"$instance\"}[5m])) by (le))", "intervalFactor": 2, "legendFormat": "50%", "refId": "C", "step": 40 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Responses (duration)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "ms", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 0, "y": 21 }, "id": 8, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "udp:50%", "yaxis": 1 }, { "alias": "tcp:50%", "yaxis": 2 }, { "alias": "tcp:90%", "yaxis": 2 }, { "alias": "tcp:99%", "yaxis": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "histogram_quantile(0.99, sum(rate(coredns_dns_response_size_bytes_bucket{instance=~\"$instance\",proto=\"udp\"}[5m])) by (le,proto)) ", "intervalFactor": 2, "legendFormat": "{{proto}}:99%", "refId": "A", "step": 40 }, { "expr": "histogram_quantile(0.90, sum(rate(coredns_dns_response_size_bytes_bucket{instance=\"$instance\",proto=\"udp\"}[5m])) by (le,proto)) ", "intervalFactor": 2, "legendFormat": "{{proto}}:90%", "refId": "B", "step": 40 }, { "expr": "histogram_quantile(0.50, sum(rate(coredns_dns_response_size_bytes_bucket{instance=~\"$instance\",proto=\"udp\"}[5m])) by (le,proto)) ", "intervalFactor": 2, "legendFormat": "{{proto}}:50%", "metric": "", "refId": "C", "step": 40 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Responses (size, udp)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 12, "y": 21 }, "id": 13, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "udp:50%", "yaxis": 1 }, { "alias": "tcp:50%", "yaxis": 1 }, { "alias": "tcp:90%", "yaxis": 1 }, { "alias": "tcp:99%", "yaxis": 1 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "histogram_quantile(0.99, sum(rate(coredns_dns_response_size_bytes_bucket{instance=~\"$instance\",proto=\"tcp\"}[5m])) by (le,proto)) ", "intervalFactor": 2, "legendFormat": "{{proto}}:99%", "refId": "A", "step": 40 }, { "expr": "histogram_quantile(0.90, sum(rate(coredns_dns_response_size_bytes_bucket{instance=~\"$instance\",proto=\"tcp\"}[5m])) by (le,proto)) ", "intervalFactor": 2, "legendFormat": "{{proto}}:90%", "refId": "B", "step": 40 }, { "expr": "histogram_quantile(0.50, sum(rate(coredns_dns_response_size_bytes_bucket{instance=~\"$instance\",proto=\"tcp\"}[5m])) by (le, proto)) ", "intervalFactor": 2, "legendFormat": "{{proto}}:50%", "metric": "", "refId": "C", "step": 40 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Responses (size, tcp)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 0, "y": 28 }, "id": 15, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(coredns_cache_size{instance=~\"$instance\"}) by (type)", "intervalFactor": 2, "legendFormat": "{{type}}", "refId": "A", "step": 40 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Cache (size)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 12, "y": 28 }, "id": 16, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "misses", "yaxis": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(coredns_cache_hits_total{instance=~\"$instance\"}[5m])) by (type)", "intervalFactor": 2, "legendFormat": "hits:{{type}}", "refId": "A", "step": 40 }, { "expr": "sum(rate(coredns_cache_misses_total{instance=~\"$instance\"}[5m])) by (type)", "intervalFactor": 2, "legendFormat": "misses", "refId": "B", "step": 40 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Cache (hitrate)", "tooltip": { "shared": true, "sort": 0, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "pps", "logBase": 1, "max": null, "min": 0, "show": true } ], "yaxis": { "align": false, "alignLevel": null } } ], "schemaVersion": 16, "style": "dark", "tags": [ "dns", "coredns" ], "templating": { "list": [ { "allValue": ".*", "current": { "text": "All", "value": "$__all" }, "datasource": "prometheus", "definition": "", "hide": 0, "includeAll": true, "label": "Instance", "multi": false, "name": "instance", "options": [ ], "query": "up{job=\"coredns\"}", "refresh": 1, "regex": ".*instance=\"(.*?)\".*", "skipUrlSync": false, "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-3h", "to": "now" }, "timepicker": { "now": true, "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "utc", "title": "CoreDNS", "uid": "q-QViRumz", "version": 1 } kind: ConfigMap metadata: name: grafana-dashboard-coredns-dashboard namespace: monitoring - apiVersion: v1 data: k8s-cluster-rsrc-use.json: |- { "annotations": { "list": [ ] }, "editable": true, "gnetId": null, "graphTooltip": 0, "hideControls": false, "links": [ ], "refresh": "10s", "rows": [ { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 1, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, "targets": [ { "expr": "node:cluster_cpu_utilisation:ratio{cluster=\"$cluster\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": 1, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 2, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, "targets": [ { "expr": "node:node_cpu_saturation_load1:{cluster=\"$cluster\"} / scalar(sum(min(kube_pod_info{cluster=\"$cluster\"}) by (node)))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Saturation (Load1)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": 1, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "CPU", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 3, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, "targets": [ { "expr": "node:cluster_memory_utilisation:ratio{cluster=\"$cluster\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": 1, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 4, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, "targets": [ { "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Saturation (Swap I/O)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "Bps", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Memory", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 5, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, "targets": [ { "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk IO Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": 1, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 6, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, "targets": [ { "expr": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk IO Saturation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": 1, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Disk", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 7, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, "targets": [ { "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Net Utilisation (Transmitted)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "Bps", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 8, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, "targets": [ { "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Net Saturation (Dropped)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "Bps", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Network", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 9, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": true, "steppedLine": false, "targets": [ { "expr": "sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"} - node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)) by (pod,namespace)\n/ scalar(sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)))\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\"}\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{node}}", "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk Capacity", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": 1, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Storage", "titleSize": "h6" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(kube_node_info, cluster)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-1h", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / USE Method / Cluster", "uid": "a6e7d1362e1ddbb79db21d5bb40d7137", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-k8s-cluster-rsrc-use namespace: monitoring - apiVersion: v1 data: k8s-node-rsrc-use.json: |- { "annotations": { "list": [ ] }, "editable": true, "gnetId": null, "graphTooltip": 0, "hideControls": false, "links": [ ], "refresh": "10s", "rows": [ { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 1, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_cpu_utilisation:avg1m{cluster=\"$cluster\", node=\"$node\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Utilisation", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 2, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_cpu_saturation_load1:{cluster=\"$cluster\", node=\"$node\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Saturation", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Saturation (Load1)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "CPU", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 3, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_memory_utilisation:{cluster=\"$cluster\", node=\"$node\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Memory", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 4, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\", node=\"$node\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Swap IO", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Saturation (Swap I/O)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "Bps", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Memory", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 5, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\", node=\"$node\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Utilisation", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk IO Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 6, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\", node=\"$node\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Saturation", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk IO Saturation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Disk", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 7, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\", node=\"$node\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Utilisation", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Net Utilisation (Transmitted)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "Bps", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 8, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\", node=\"$node\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Saturation", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Net Saturation (Dropped)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "Bps", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Net", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 9, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_filesystem_usage:{cluster=\"$cluster\"}\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\", node=\"$node\"}\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{device}}", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Disk", "titleSize": "h6" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(kube_node_info, cluster)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "node", "multi": false, "name": "node", "options": [ ], "query": "label_values(kube_node_info{cluster=\"$cluster\"}, node)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-1h", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / USE Method / Node", "uid": "4ac4f123aae0ff6dbaf4f4f66120033b", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-k8s-node-rsrc-use namespace: monitoring - apiVersion: v1 data: k8s-resources-cluster.json: |- { "annotations": { "list": [ ] }, "editable": true, "gnetId": null, "graphTooltip": 0, "hideControls": false, "links": [ ], "refresh": "10s", "rows": [ { "collapse": false, "height": "100px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "format": "percentunit", "id": 1, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 2, "stack": false, "steppedLine": false, "targets": [ { "expr": "1 - avg(rate(node_cpu_seconds_total{mode=\"idle\", cluster=\"$cluster\"}[1m]))", "format": "time_series", "instant": true, "intervalFactor": 2, "refId": "A" } ], "thresholds": "70,80", "timeFrom": null, "timeShift": null, "title": "CPU Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "singlestat", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "format": "percentunit", "id": 2, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 2, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) / sum(node:node_num_cpu:sum{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, "refId": "A" } ], "thresholds": "70,80", "timeFrom": null, "timeShift": null, "title": "CPU Requests Commitment", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "singlestat", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "format": "percentunit", "id": 3, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 2, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) / sum(node:node_num_cpu:sum{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, "refId": "A" } ], "thresholds": "70,80", "timeFrom": null, "timeShift": null, "title": "CPU Limits Commitment", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "singlestat", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "format": "percentunit", "id": 4, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 2, "stack": false, "steppedLine": false, "targets": [ { "expr": "1 - sum(:node_memory_MemFreeCachedBuffers_bytes:sum{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, "refId": "A" } ], "thresholds": "70,80", "timeFrom": null, "timeShift": null, "title": "Memory Utilisation", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "singlestat", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "format": "percentunit", "id": 5, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 2, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, "refId": "A" } ], "thresholds": "70,80", "timeFrom": null, "timeShift": null, "title": "Memory Requests Commitment", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "singlestat", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "format": "percentunit", "id": 6, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 2, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, "refId": "A" } ], "thresholds": "70,80", "timeFrom": null, "timeShift": null, "title": "Memory Limits Commitment", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "singlestat", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Headlines", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 7, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": true, "steppedLine": false, "targets": [ { "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "CPU", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 8, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "styles": [ { "alias": "Time", "dateFormat": "YYYY-MM-DD HH:mm:ss", "pattern": "Time", "type": "hidden" }, { "alias": "CPU Usage", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #A", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Requests", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #B", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Requests %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #C", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "CPU Limits", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #D", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Limits %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #E", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Namespace", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": true, "linkTooltip": "Drill down", "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", "pattern": "namespace", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "/.*/", "thresholds": [ ], "type": "string", "unit": "short" } ], "targets": [ { "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "A", "step": 10 }, { "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "B", "step": 10 }, { "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "C", "step": 10 }, { "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "D", "step": 10 }, { "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "E", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Quota", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "transform": "table", "type": "table", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "CPU Quota", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 9, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": true, "steppedLine": false, "targets": [ { "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Usage (w/o cache)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Memory", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 10, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "styles": [ { "alias": "Time", "dateFormat": "YYYY-MM-DD HH:mm:ss", "pattern": "Time", "type": "hidden" }, { "alias": "Memory Usage", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #A", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Requests", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #B", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Requests %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #C", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Memory Limits", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #D", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Limits %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #E", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Namespace", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": true, "linkTooltip": "Drill down", "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", "pattern": "namespace", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "/.*/", "thresholds": [ ], "type": "string", "unit": "short" } ], "targets": [ { "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "A", "step": 10 }, { "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "B", "step": 10 }, { "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "C", "step": 10 }, { "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "D", "step": 10 }, { "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "E", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Requests by Namespace", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "transform": "table", "type": "table", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Memory Requests", "titleSize": "h6" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(node_cpu_seconds_total, cluster)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-1h", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / Compute Resources / Cluster", "uid": "efa86fd1d0c121a26444b636a3f509a8", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-k8s-resources-cluster namespace: monitoring - apiVersion: v1 data: k8s-resources-namespace.json: |- { "annotations": { "list": [ ] }, "editable": true, "gnetId": null, "graphTooltip": 0, "hideControls": false, "links": [ ], "refresh": "10s", "rows": [ { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 1, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": true, "steppedLine": false, "targets": [ { "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod_name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod_name}}", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "CPU Usage", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 2, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "styles": [ { "alias": "Time", "dateFormat": "YYYY-MM-DD HH:mm:ss", "pattern": "Time", "type": "hidden" }, { "alias": "CPU Usage", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #A", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Requests", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #B", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Requests %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #C", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "CPU Limits", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #D", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Limits %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #E", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Pod", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": true, "linkTooltip": "Drill down", "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", "pattern": "pod", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "/.*/", "thresholds": [ ], "type": "string", "unit": "short" } ], "targets": [ { "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "A", "step": 10 }, { "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "B", "step": 10 }, { "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "C", "step": 10 }, { "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "D", "step": 10 }, { "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "E", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Quota", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "transform": "table", "type": "table", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "CPU Quota", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 3, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": true, "steppedLine": false, "targets": [ { "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"}) by (pod_name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod_name}}", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Usage (w/o cache)", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Memory Usage", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 4, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "styles": [ { "alias": "Time", "dateFormat": "YYYY-MM-DD HH:mm:ss", "pattern": "Time", "type": "hidden" }, { "alias": "Memory Usage", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #A", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Requests", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #B", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Requests %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #C", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Memory Limits", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #D", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Limits %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #E", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Memory Usage (RSS)", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #F", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Usage (Cache)", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #G", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Usage (Swap", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #H", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Pod", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": true, "linkTooltip": "Drill down", "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", "pattern": "pod", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "/.*/", "thresholds": [ ], "type": "string", "unit": "short" } ], "targets": [ { "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "A", "step": 10 }, { "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "B", "step": 10 }, { "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "C", "step": 10 }, { "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "D", "step": 10 }, { "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "E", "step": 10 }, { "expr": "sum(label_replace(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "F", "step": 10 }, { "expr": "sum(label_replace(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "G", "step": 10 }, { "expr": "sum(label_replace(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "H", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Quota", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "transform": "table", "type": "table", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Memory Quota", "titleSize": "h6" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(kube_pod_info, cluster)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "namespace", "multi": false, "name": "namespace", "options": [ ], "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-1h", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / Compute Resources / Namespace", "uid": "85a562078cdf77779eaa1add43ccec1e", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-k8s-resources-namespace namespace: monitoring - apiVersion: v1 data: k8s-resources-pod.json: |- { "annotations": { "list": [ ] }, "editable": true, "gnetId": null, "graphTooltip": 0, "hideControls": false, "links": [ ], "refresh": "10s", "rows": [ { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 1, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": true, "steppedLine": false, "targets": [ { "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", cluster=\"$cluster\"}) by (container_name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{container_name}}", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "CPU Usage", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 2, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "styles": [ { "alias": "Time", "dateFormat": "YYYY-MM-DD HH:mm:ss", "pattern": "Time", "type": "hidden" }, { "alias": "CPU Usage", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #A", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Requests", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #B", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Requests %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #C", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "CPU Limits", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #D", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "CPU Limits %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #E", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Container", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "container", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "/.*/", "thresholds": [ ], "type": "string", "unit": "short" } ], "targets": [ { "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "A", "step": 10 }, { "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "B", "step": 10 }, { "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "C", "step": 10 }, { "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "D", "step": 10 }, { "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "E", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Quota", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "transform": "table", "type": "table", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "CPU Quota", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 10, "id": 3, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 0, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": true, "steppedLine": false, "targets": [ { "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{container_name}} (RSS)", "legendLink": null, "step": 10 }, { "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{container_name}} (Cache)", "legendLink": null, "step": 10 }, { "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{container_name}} (Swap)", "legendLink": null, "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Memory Usage", "titleSize": "h6" }, { "collapse": false, "height": "250px", "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "id": 4, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "styles": [ { "alias": "Time", "dateFormat": "YYYY-MM-DD HH:mm:ss", "pattern": "Time", "type": "hidden" }, { "alias": "Memory Usage", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #A", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Requests", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #B", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Requests %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #C", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Memory Limits", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #D", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Limits %", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #E", "thresholds": [ ], "type": "number", "unit": "percentunit" }, { "alias": "Memory Usage (RSS)", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #F", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Usage (Cache)", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #G", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Memory Usage (Swap", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "Value #H", "thresholds": [ ], "type": "number", "unit": "bytes" }, { "alias": "Container", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "link": false, "linkTooltip": "Drill down", "linkUrl": "", "pattern": "container", "thresholds": [ ], "type": "number", "unit": "short" }, { "alias": "", "colorMode": null, "colors": [ ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "/.*/", "thresholds": [ ], "type": "string", "unit": "short" } ], "targets": [ { "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "A", "step": 10 }, { "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "B", "step": 10 }, { "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "C", "step": 10 }, { "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "D", "step": 10 }, { "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "E", "step": 10 }, { "expr": "sum(label_replace(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "F", "step": 10 }, { "expr": "sum(label_replace(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "G", "step": 10 }, { "expr": "sum(label_replace(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", "format": "table", "instant": true, "intervalFactor": 2, "legendFormat": "", "refId": "H", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Quota", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "transform": "table", "type": "table", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": false } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": true, "title": "Memory Quota", "titleSize": "h6" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(kube_pod_info, cluster)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "namespace", "multi": false, "name": "namespace", "options": [ ], "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { "text": "prod", "value": "prod" }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "pod", "multi": false, "name": "pod", "options": [ ], "query": "label_values(kube_pod_info{cluster=\"$cluster\", namespace=\"$namespace\"}, pod)", "refresh": 1, "regex": "", "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-1h", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / Compute Resources / Pod", "uid": "6581e46e4e5c7ba40a07646395ef7b23", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-k8s-resources-pod namespace: monitoring - apiVersion: v1 data: kubernetes-cluster-dashboard.json: |- { "annotations": { "list": [ { "builtIn": 1, "datasource": "-- Grafana --", "enable": true, "hide": true, "iconColor": "rgba(0, 211, 255, 1)", "name": "Annotations & Alerts", "type": "dashboard" } ] }, "description": "Monitor a Kubernetes cluster using Prometheus TSDB. Shows overall cluster CPU / Memory / Disk usage as well as individual pod statistics. ", "editable": true, "gnetId": 162, "graphTooltip": 1, "id": 7, "links": [ ], "panels": [ { "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", "editable": true, "error": false, "format": "percent", "gauge": { "maxValue": 100, "minValue": 0, "show": true, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 7, "w": 8, "x": 0, "y": 0 }, "id": 4, "interval": null, "isNew": true, "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" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "(sum(node_memory_MemTotal_bytes) - sum(node_memory_MemFree_bytes+node_memory_Buffers_bytes+node_memory_Cached_bytes) ) / sum(node_memory_MemTotal_bytes) * 100", "format": "time_series", "interval": "10s", "intervalFactor": 1, "refId": "A", "step": 10 } ], "thresholds": "65, 90", "title": "Cluster memory usage", "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "current" }, { "cacheTimeout": null, "colorBackground": true, "colorValue": false, "colors": [ "rgba(0, 0, 0, 0)", "rgb(210, 1, 1)", "#890f02" ], "datasource": "prometheus", "format": "percentunit", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 2, "w": 8, "x": 8, "y": 0 }, "id": 23, "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" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "avg(up{job=\"kubelet\"}) BY (job)", "format": "time_series", "instant": true, "intervalFactor": 1, "legendFormat": "", "refId": "A" } ], "thresholds": "1.1", "title": "Up Nodes", "type": "singlestat", "valueFontSize": "120%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "avg" }, { "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": 0, "editable": true, "error": false, "format": "percent", "gauge": { "maxValue": 100, "minValue": 0, "show": true, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 7, "w": 8, "x": 16, "y": 0 }, "id": 6, "interval": null, "isNew": true, "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" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "avg(100 - (avg by (instance) (irate(node_cpu_seconds_total{job=\"node-exporter\",mode=\"idle\"}[5m])) * 100))", "format": "time_series", "interval": "10s", "intervalFactor": 1, "refId": "A", "step": 10 } ], "thresholds": "65, 90", "title": "Cluster CPU usage", "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "current" }, { "columns": [ ], "datasource": "prometheus", "fontSize": "90%", "gridPos": { "h": 5, "w": 8, "x": 8, "y": 2 }, "id": 25, "links": [ ], "pageSize": null, "scroll": true, "showHeader": true, "sort": { "col": 2, "desc": false }, "styles": [ { "alias": "Time", "dateFormat": "YYYY-MM-DD HH:mm:ss", "link": false, "pattern": "Time", "type": "date" }, { "alias": "Uptime", "colorMode": null, "colors": [ "rgba(50, 172, 45, 0.97)", "rgba(237, 129, 40, 0.89)", "rgba(245, 54, 54, 0.9)" ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "Value", "thresholds": [ ], "type": "number", "unit": "s" }, { "alias": "", "colorMode": null, "colors": [ "rgba(245, 54, 54, 0.9)", "rgba(237, 129, 40, 0.89)", "rgba(50, 172, 45, 0.97)" ], "decimals": 2, "pattern": "/endpoint|job|namespace|pod|service/", "thresholds": [ ], "type": "hidden", "unit": "short" }, { "alias": "", "colorMode": null, "colors": [ "rgba(245, 54, 54, 0.9)", "rgba(237, 129, 40, 0.89)", "rgba(50, 172, 45, 0.97)" ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "instance", "preserveFormat": false, "sanitize": false, "thresholds": [ ], "type": "string", "unit": "short" } ], "targets": [ { "expr": "(time() - node_boot_time_seconds)", "format": "table", "instant": true, "intervalFactor": 1, "refId": "A" } ], "title": "Node Uptime", "transform": "table", "transparent": true, "type": "table" }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 7 }, "id": 15, "panels": [ ], "title": "Nodes", "type": "row" }, { "alert": { "conditions": [ { "evaluator": { "params": [ 0.84999999999999998 ], "type": "gt" }, "operator": { "type": "and" }, "query": { "params": [ "B", "5m", "now" ] }, "reducer": { "params": [ ], "type": "max" }, "type": "query" } ], "executionErrorState": "alerting", "for": "0m", "frequency": "60s", "handler": 1, "name": "Memory Usage alert", "noDataState": "no_data", "notifications": [ ] }, "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 9, "w": 12, "x": 0, "y": 8 }, "id": 10, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "node_memory_MemTotal_bytes - (node_memory_MemFree_bytes+node_memory_Buffers_bytes+node_memory_Cached_bytes)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ instance }}", "refId": "A" }, { "expr": "(node_memory_MemTotal_bytes - (node_memory_MemFree_bytes+node_memory_Buffers_bytes+node_memory_Cached_bytes))/node_memory_MemTotal_bytes", "format": "time_series", "hide": true, "intervalFactor": 1, "refId": "B" } ], "thresholds": [ { "colorMode": "critical", "fill": true, "line": true, "op": "gt", "value": 0.84999999999999998 } ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Memory Usage", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "decbytes", "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 } }, { "alert": { "conditions": [ { "evaluator": { "params": [ 90 ], "type": "gt" }, "operator": { "type": "and" }, "query": { "params": [ "A", "15m", "now" ] }, "reducer": { "params": [ ], "type": "max" }, "type": "query" } ], "executionErrorState": "alerting", "frequency": "60s", "handler": 1, "name": "CPU Usage alert", "noDataState": "no_data", "notifications": [ ] }, "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 9, "w": 12, "x": 12, "y": 8 }, "id": 11, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "100 - (avg by (instance) (irate(node_cpu_seconds_total{job=\"node-exporter\",mode=\"idle\"}[5m])) * 100)", "format": "time_series", "intervalFactor": 3, "legendFormat": "{{instance}}", "refId": "A" } ], "thresholds": [ { "colorMode": "critical", "fill": true, "line": true, "op": "gt", "value": 90 } ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "CPU Usage", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "decimals": null, "format": "percent", "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 } }, { "columns": [ ], "datasource": "prometheus", "fontSize": "100%", "gridPos": { "h": 9, "w": 12, "x": 0, "y": 17 }, "id": 31, "links": [ ], "pageSize": null, "scroll": true, "showHeader": true, "sort": { "col": 0, "desc": true }, "styles": [ { "alias": "Time", "dateFormat": "YYYY-MM-DD HH:mm:ss", "link": false, "pattern": "Time", "type": "date" }, { "alias": "", "colorMode": null, "colors": [ "rgba(245, 54, 54, 0.9)", "rgba(237, 129, 40, 0.89)", "rgba(50, 172, 45, 0.97)" ], "dateFormat": "YYYY-MM-DD HH:mm:ss", "decimals": 2, "pattern": "condition|container|daemonset|endpoint|namespace|node", "thresholds": [ ], "type": "hidden", "unit": "short" }, { "alias": "", "colorMode": null, "colors": [ "rgba(245, 54, 54, 0.9)", "rgba(237, 129, 40, 0.89)", "rgba(50, 172, 45, 0.97)" ], "decimals": 2, "pattern": "/.*/", "thresholds": [ ], "type": "number", "unit": "short" } ], "targets": [ { "expr": "ALERTS{alertstate=\"firing\"}", "format": "table", "instant": true, "intervalFactor": 1, "refId": "A" }, { "expr": "ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"}", "format": "table", "hide": true, "intervalFactor": 1, "refId": "B" } ], "title": "Active Alerts", "transform": "table", "type": "table" }, { "dashboardFilter": "", "dashboardTags": [ ], "folderId": null, "gridPos": { "h": 9, "w": 5, "x": 12, "y": 17 }, "id": 27, "limit": 10, "links": [ ], "nameFilter": "", "onlyAlertsOnDashboard": false, "show": "current", "sortOrder": 1, "stateFilter": [ ], "title": "Alarms", "type": "alertlist" }, { "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": null, "editable": true, "error": false, "format": "percent", "gauge": { "maxValue": 100, "minValue": 0, "show": true, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 5, "w": 7, "x": 17, "y": 17 }, "id": 7, "interval": null, "isNew": true, "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" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "(sum(node_filesystem_size_bytes{device=~\"/dev/.*\"}) - sum(node_filesystem_free_bytes{device=~\"/dev/.*\"}) ) / sum(node_filesystem_size_bytes{device=~\"/dev/.*\"}) * 100", "format": "time_series", "interval": "10s", "intervalFactor": 1, "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" }, { "alert": { "conditions": [ { "evaluator": { "params": [ 1 ], "type": "lt" }, "operator": { "type": "and" }, "query": { "params": [ "C", "5m", "now" ] }, "reducer": { "params": [ ], "type": "avg" }, "type": "query" } ], "executionErrorState": "alerting", "frequency": "60s", "handler": 1, "name": "Node Down", "noDataState": "alerting", "notifications": [ ] }, "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 4, "w": 7, "x": 17, "y": 22 }, "id": 29, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(up{job=\"kubelet\"}) BY (job)", "format": "time_series", "hide": true, "instant": false, "intervalFactor": 1, "legendFormat": "Up Nodes", "refId": "A" }, { "expr": "count(up{job=\"kubelet\"})", "format": "time_series", "hide": true, "instant": false, "intervalFactor": 1, "legendFormat": "Total Nodes", "refId": "B" }, { "expr": "avg(up{job=\"kubelet\"}) BY (job)", "format": "time_series", "hide": false, "intervalFactor": 1, "refId": "C" } ], "thresholds": [ { "colorMode": "critical", "fill": true, "line": true, "op": "lt", "value": 1 } ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Up Nodes", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "alert": { "conditions": [ { "evaluator": { "params": [ 85 ], "type": "gt" }, "operator": { "type": "and" }, "query": { "params": [ "A", "1m", "now" ] }, "reducer": { "params": [ ], "type": "avg" }, "type": "query" } ], "executionErrorState": "alerting", "frequency": "60s", "handler": 1, "name": "CPU Temperature alert", "noDataState": "no_data", "notifications": [ ] }, "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 6, "w": 24, "x": 0, "y": 26 }, "id": 13, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "rpi_cpu_temperature_celsius", "format": "time_series", "intervalFactor": 5, "legendFormat": "{{instance}}", "refId": "A" } ], "thresholds": [ { "colorMode": "critical", "fill": true, "line": true, "op": "gt", "value": 85 } ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "CPU Temperature", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "celsius", "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 } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 32 }, "id": 17, "panels": [ ], "title": "Pods", "type": "row" }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "decimals": 0, "editable": true, "error": false, "fill": 0, "grid": { }, "gridPos": { "h": 7, "w": 24, "x": 0, "y": 33 }, "id": 3, "isNew": true, "legend": { "alignAsTable": true, "avg": true, "current": true, "hideEmpty": true, "hideZero": true, "max": false, "min": false, "rightSide": true, "show": true, "sideWidth": 270, "sort": "current", "sortDesc": true, "total": false, "values": true }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "null as zero", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "topk(10,sum by (container_name)( rate(container_cpu_usage_seconds_total{image!=\"\",container_name!=\"POD\"}[1m] ) ))", "format": "time_series", "instant": false, "interval": "", "intervalFactor": 1, "legendFormat": "{{ container_name}}", "metric": "container_cpu", "refId": "A", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Pod CPU usage", "tooltip": { "msResolution": true, "shared": true, "sort": 2, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "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 } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "decimals": 2, "editable": true, "error": false, "fill": 0, "grid": { }, "gridPos": { "h": 7, "w": 24, "x": 0, "y": 40 }, "id": 2, "isNew": true, "legend": { "alignAsTable": true, "avg": true, "current": true, "max": false, "min": false, "rightSide": true, "show": true, "sideWidth": 250, "sort": "avg", "sortDesc": true, "total": false, "values": true }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sort_desc(sum(container_memory_usage_bytes{image!=\"\"}) by (container_name, image))", "format": "time_series", "hide": true, "interval": "10s", "intervalFactor": 1, "legendFormat": "{{ container_name }}", "metric": "container_memory_usage:sort_desc", "refId": "A", "step": 10 }, { "expr": "topk(10,sum(container_memory_rss{name=~\".+\"}) by (container_name))", "format": "time_series", "intervalFactor": 1, "legendFormat": "{{ container_name }}", "refId": "B" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Pod 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": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 0, "y": 47 }, "id": 19, "legend": { "alignAsTable": true, "avg": true, "current": false, "hideEmpty": true, "hideZero": true, "max": false, "min": false, "rightSide": false, "show": true, "sideWidth": 550, "sort": "avg", "sortDesc": true, "total": false, "values": true }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "topk(10,sum(rate(container_network_transmit_bytes_total{pod_name=~\".+\"}[5m])) by (pod_name))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ pod_name }}", "refId": "A", "step": 240 }, { "expr": "rate(container_network_transmit_bytes_total{id=\"/\"}[$interval])", "format": "time_series", "hide": true, "intervalFactor": 2, "legendFormat": "", "refId": "B", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Sent Network Traffic per Container", "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": "Bps", "label": "", "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": "", "logBase": 10, "max": 8, "min": 0, "show": false } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 12, "y": 47 }, "id": 21, "legend": { "alignAsTable": true, "avg": true, "current": false, "hideEmpty": true, "hideZero": true, "max": false, "min": false, "rightSide": false, "show": true, "sideWidth": 150, "sort": "avg", "sortDesc": true, "total": false, "values": true }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "topk(10,sum(rate(container_network_receive_bytes_total{pod_name=~\".+\"}[5m])) by (pod_name))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod_name}}", "refId": "A", "step": 240 }, { "expr": "- rate(container_network_transmit_bytes_total{pod_name=~\".+\"}[$interval])", "format": "time_series", "hide": true, "intervalFactor": 2, "legendFormat": "{{pod_name}}", "refId": "B", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Received Network Traffic per Container", "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": "Bps", "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 } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "decimals": 2, "editable": true, "error": false, "fill": 0, "grid": { }, "gridPos": { "h": 7, "w": 24, "x": 0, "y": 54 }, "id": 8, "isNew": true, "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, "stack": false, "steppedLine": false, "targets": [ { "expr": "sort_desc(sum by (kubernetes_pod_name) (rate (container_network_receive_bytes_total{name!=\"\", kubernetes_pod_name=~\".*\"}[1m]) ))", "format": "time_series", "interval": "10s", "intervalFactor": 1, "legendFormat": "Receive Traffic", "metric": "network", "refId": "A", "step": 10 }, { "expr": "sort_desc(sum by (kubernetes_pod_name) (rate (container_network_transmit_bytes_total{name!=\"\", kubernetes_pod_name=~\".*\"}[1m]) ))", "format": "time_series", "interval": "10s", "intervalFactor": 1, "legendFormat": "Transmit Traffic", "metric": "network", "refId": "B", "step": 10 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Pod Network i/o", "tooltip": { "msResolution": false, "shared": true, "sort": 0, "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": true } ], "yaxis": { "align": false, "alignLevel": null } } ], "refresh": "10s", "schemaVersion": 16, "style": "dark", "tags": [ "custom" ], "templating": { "list": [ ] }, "time": { "from": "now-3h", "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)", "uid": "82pBZCmRkasd", "version": 1 } kind: ConfigMap metadata: name: grafana-dashboard-kubernetes-cluster-dashboard namespace: monitoring - apiVersion: v1 data: nodes.json: |- { "__inputs": [ ], "__requires": [ ], "annotations": { "list": [ ] }, "editable": false, "gnetId": null, "graphTooltip": 0, "hideControls": false, "id": null, "links": [ ], "refresh": "", "rows": [ { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 2, "legend": { "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, "rightSide": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(node_load1{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "load 1m", "refId": "A" }, { "expr": "max(node_load5{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "load 5m", "refId": "B" }, { "expr": "max(node_load15{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "load 15m", "refId": "C" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "System load", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 3, "legend": { "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, "rightSide": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[5m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{cpu}}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Usage Per Core", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 4, "legend": { "alignAsTable": "true", "avg": "true", "current": "true", "max": "false", "min": "false", "rightSide": "true", "show": "true", "total": "false", "values": "true" }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 9, "stack": false, "steppedLine": false, "targets": [ { "expr": "max (sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m])) ) * 100\n", "format": "time_series", "intervalFactor": 10, "legendFormat": "{{ cpu }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Utilization", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percent", "label": null, "logBase": 1, "max": 100, "min": 0, "show": true }, { "format": "percent", "label": null, "logBase": 1, "max": 100, "min": 0, "show": true } ] }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "rgba(50, 172, 45, 0.97)", "rgba(237, 129, 40, 0.89)", "rgba(245, 54, 54, 0.9)" ], "datasource": "$datasource", "format": "percent", "gauge": { "maxValue": 100, "minValue": 0, "show": true, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 5, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 3, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "avg(sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m]))) * 100\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "80, 90", "title": "CPU Usage", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "current" } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 6, "legend": { "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, "rightSide": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 9, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "memory used", "refId": "A" }, { "expr": "max(node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "memory buffers", "refId": "B" }, { "expr": "max(node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "memory cached", "refId": "C" }, { "expr": "max(node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "memory free", "refId": "D" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "rgba(50, 172, 45, 0.97)", "rgba(237, 129, 40, 0.89)", "rgba(245, 54, 54, 0.9)" ], "datasource": "$datasource", "format": "percent", "gauge": { "maxValue": 100, "minValue": 0, "show": true, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 7, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 3, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "max(\n (\n (\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "80, 90", "title": "Memory Usage", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "current" } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 8, "legend": { "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, "rightSide": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ { "alias": "read", "yaxis": 1 }, { "alias": "io time", "yaxis": 2 } ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(rate(node_disk_read_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "read", "refId": "A" }, { "expr": "max(rate(node_disk_written_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "written", "refId": "B" }, { "expr": "max(rate(node_disk_io_time_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "io time", "refId": "C" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk I/O", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "ms", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 9, "legend": { "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, "rightSide": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "node:node_filesystem_usage:{cluster=\"$cluster\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{device}}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk Space Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "percentunit", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 10, "legend": { "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, "rightSide": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(rate(node_network_receive_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{device}}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Network Received", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 11, "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", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(rate(node_network_transmit_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{device}}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Network Transmitted", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 12, "legend": { "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, "rightSide": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 9, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "inodes used", "refId": "A" }, { "expr": "max(node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "inodes free", "refId": "B" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Inodes Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "rgba(50, 172, 45, 0.97)", "rgba(237, 129, 40, 0.89)", "rgba(245, 54, 54, 0.9)" ], "datasource": "$datasource", "format": "percent", "gauge": { "maxValue": 100, "minValue": 0, "show": true, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "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": 3, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "max(\n (\n (\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "80, 90", "title": "Inodes Usage", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "current" } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(kube_pod_info, cluster)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": null, "multi": false, "name": "instance", "options": [ ], "query": "label_values(node_boot_time_seconds{cluster=\"$cluster\", job=\"node-exporter\"}, instance)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-1h", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / Nodes", "uid": "fa49a4706d07a042595b664c87fb33ea", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-nodes namespace: monitoring - apiVersion: v1 data: persistentvolumesusage.json: |- { "__inputs": [ ], "__requires": [ ], "annotations": { "list": [ ] }, "editable": false, "gnetId": null, "graphTooltip": 0, "hideControls": false, "id": null, "links": [ ], "refresh": "", "rows": [ { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 2, "legend": { "alignAsTable": false, "avg": true, "current": true, "max": true, "min": true, "rightSide": false, "show": true, "total": false, "values": true }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "targets": [ { "expr": "(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"} - kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"}) / kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"} * 100\n", "format": "time_series", "intervalFactor": 1, "legendFormat": "{{ Usage }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Volume Space Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percent", "label": null, "logBase": 1, "max": 100, "min": 0, "show": true }, { "format": "percent", "label": null, "logBase": 1, "max": 100, "min": 0, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 3, "legend": { "alignAsTable": false, "avg": true, "current": true, "max": true, "min": true, "rightSide": false, "show": true, "total": false, "values": true }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "targets": [ { "expr": "kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"} / kubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", persistentvolumeclaim=\"$volume\"} * 100\n", "format": "time_series", "intervalFactor": 1, "legendFormat": "{{ Usage }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Volume inodes Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "percent", "label": null, "logBase": 1, "max": 100, "min": 0, "show": true }, { "format": "percent", "label": null, "logBase": 1, "max": 100, "min": 0, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(kubelet_volume_stats_capacity_bytes, cluster)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "Namespace", "multi": false, "name": "namespace", "options": [ ], "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\"}, exported_namespace)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "PersistentVolumeClaim", "multi": false, "name": "volume", "options": [ ], "query": "label_values(kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", exported_namespace=\"$namespace\"}, persistentvolumeclaim)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-7d", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / Persistent Volumes", "uid": "919b92a8e8041bd567af9edab12c840c", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-persistentvolumesusage namespace: monitoring - apiVersion: v1 data: pods.json: |- { "__inputs": [ ], "__requires": [ ], "annotations": { "list": [ ] }, "editable": false, "gnetId": null, "graphTooltip": 0, "hideControls": false, "id": null, "links": [ ], "refresh": "", "rows": [ { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 2, "legend": { "alignAsTable": true, "avg": true, "current": true, "max": false, "min": false, "rightSide": true, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum by(container_name) (container_memory_usage_bytes{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Current: {{ container_name }}", "refId": "A" }, { "expr": "sum by(container) (kube_pod_container_resource_requests_memory_bytes{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Requested: {{ container }}", "refId": "B" }, { "expr": "sum by(container) (kube_pod_container_resource_limits_memory_bytes{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Limit: {{ container }}", "refId": "C" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": 0, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 3, "legend": { "alignAsTable": true, "avg": true, "current": true, "max": false, "min": false, "rightSide": true, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum by (container_name) (rate(container_cpu_usage_seconds_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", image!=\"\", pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"}[1m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "Current: {{ container_name }}", "refId": "A" }, { "expr": "sum by(container) (kube_pod_container_resource_requests_cpu_cores{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Requested: {{ container }}", "refId": "B" }, { "expr": "sum by(container) (kube_pod_container_resource_limits_cpu_cores{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "Limit: {{ container }}", "refId": "C" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU Usage", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": 0, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 4, "legend": { "alignAsTable": true, "avg": true, "current": true, "max": false, "min": false, "rightSide": true, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, "targets": [ { "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}[1m])))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ pod_name }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Network I/O", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": 0, "show": true }, { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": 0, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(kube_pod_info, cluster)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "Namespace", "multi": false, "name": "namespace", "options": [ ], "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "Pod", "multi": false, "name": "pod", "options": [ ], "query": "label_values(kube_pod_info{cluster=\"$cluster\", namespace=~\"$namespace\"}, pod)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 0, "includeAll": true, "label": "Container", "multi": false, "name": "container", "options": [ ], "query": "label_values(kube_pod_container_info{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}, container)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-1h", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / Pods", "uid": "ab4f13a9892a76a4d21ce8c2445bf4ea", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-pods namespace: monitoring - apiVersion: v1 data: prometheus-dashboard.json: |- { "annotations": { "list": [ { "builtIn": 1, "datasource": "-- Grafana --", "enable": true, "hide": true, "iconColor": "rgba(0, 211, 255, 1)", "name": "Annotations & Alerts", "type": "dashboard" }, { "datasource": "$datasource", "enable": true, "expr": "count(sum(up{instance=\"$instance\"}) by (instance) < 1)", "hide": false, "iconColor": "rgb(250, 44, 18)", "limit": 100, "name": "downage", "showIn": 0, "step": "30s", "tagKeys": "instance", "textFormat": "prometheus down", "titleFormat": "Downage", "type": "alert" }, { "datasource": "$datasource", "enable": true, "expr": "sum(changes(prometheus_config_last_reload_success_timestamp_seconds[10m])) by (instance)", "hide": false, "iconColor": "#fceaca", "limit": 100, "name": "Reload", "showIn": 0, "step": "5m", "tagKeys": "instance", "tags": [ ], "titleFormat": "Reload", "type": "tags" } ] }, "description": "Dashboard for monitoring of Prometheus v2.x.x", "editable": true, "gnetId": 3681, "graphTooltip": 1, "id": 13, "iteration": 1549118131383, "links": [ { "icon": "info", "tags": [ ], "targetBlank": true, "title": "Dashboard's Github ", "tooltip": "Github repo of this dashboard", "type": "link", "url": "https://github.com/FUSAKLA/Prometheus2-grafana-dashboard" }, { "icon": "doc", "tags": [ ], "targetBlank": true, "title": "Prometheus Docs", "tooltip": "", "type": "link", "url": "http://prometheus.io/docs/introduction/overview/" } ], "panels": [ { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 0 }, "id": 55, "panels": [ ], "repeat": null, "title": "Header instance info", "type": "row" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#bf1b00" ], "datasource": "prometheus", "decimals": 1, "format": "s", "gauge": { "maxValue": 1000000, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 5, "w": 4, "x": 0, "y": 1 }, "id": 41, "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" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "min(time() - process_start_time_seconds{instance=\"$instance\"})", "format": "time_series", "instant": false, "intervalFactor": 2, "refId": "A" } ], "thresholds": "", "title": "Uptime", "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "current" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": true, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#bf1b00" ], "datasource": "prometheus", "format": "short", "gauge": { "maxValue": 1000000, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 5, "w": 8, "x": 4, "y": 1 }, "id": 42, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": true }, "tableColumn": "", "targets": [ { "expr": "prometheus_tsdb_head_series{instance=\"$instance\"}", "format": "time_series", "instant": false, "intervalFactor": 2, "refId": "A" } ], "thresholds": "500000,800000,1000000", "title": "Total count of time series", "type": "singlestat", "valueFontSize": "150%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "current" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "prometheus", "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 5, "w": 3, "x": 12, "y": 1 }, "id": 48, "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" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "version", "targets": [ { "expr": "prometheus_build_info{instance=\"$instance\"}", "format": "table", "instant": true, "intervalFactor": 2, "refId": "A" } ], "thresholds": "", "title": "Version", "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "avg" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "prometheus", "decimals": 2, "format": "ms", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 5, "w": 4, "x": 15, "y": 1 }, "id": 49, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "prometheus_tsdb_head_max_time{instance=\"$instance\"} - prometheus_tsdb_head_min_time{instance=\"$instance\"}", "format": "time_series", "instant": true, "intervalFactor": 2, "refId": "A" } ], "thresholds": "", "title": "Actual head block length", "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "current" }, { "content": "", "gridPos": { "h": 5, "w": 2, "x": 19, "y": 1 }, "height": "", "id": 50, "links": [ ], "mode": "html", "title": "", "transparent": true, "type": "text" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": true, "colors": [ "#e6522c", "rgba(237, 129, 40, 0.89)", "#299c46" ], "datasource": "prometheus", "decimals": 1, "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { "h": 5, "w": 3, "x": 21, "y": 1 }, "id": 52, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "2", "format": "time_series", "intervalFactor": 2, "refId": "A" } ], "thresholds": "10,20", "title": "", "transparent": true, "type": "singlestat", "valueFontSize": "200%", "valueMaps": [ { "op": "=", "text": "N/A", "value": "null" } ], "valueName": "avg" }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 6 }, "id": 56, "panels": [ ], "repeat": null, "title": "Main info", "type": "row" }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 0, "y": 7 }, "id": 15, "legend": { "avg": true, "current": false, "max": false, "min": false, "show": false, "total": false, "values": true }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": true, "steppedLine": false, "targets": [ { "expr": "max(prometheus_engine_query_duration_seconds{instance=\"$instance\"}) by (instance, slice)", "format": "time_series", "intervalFactor": 1, "legendFormat": "max duration for {{slice}}", "metric": "prometheus_local_storage_rushed_mode", "refId": "A", "step": 900 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Query elapsed time", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "s", "label": "", "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 8, "y": 7 }, "id": 17, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(prometheus_tsdb_head_series_created_total{instance=\"$instance\"}[$aggregation_interval])) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "created on {{ instance }}", "metric": "prometheus_local_storage_maintain_series_duration_seconds_count", "refId": "A", "step": 1800 }, { "expr": "sum(increase(prometheus_tsdb_head_series_removed_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) * -1", "format": "time_series", "intervalFactor": 2, "legendFormat": "removed on {{ instance }}", "refId": "B" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Head series created/deleted", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 16, "y": 7 }, "id": 13, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(prometheus_target_scrapes_exceeded_sample_limit_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "exceeded_sample_limit on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "A", "step": 1800 }, { "expr": "sum(increase(prometheus_target_scrapes_sample_duplicate_timestamp_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "duplicate_timestamp on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "B", "step": 1800 }, { "expr": "sum(increase(prometheus_target_scrapes_sample_out_of_bounds_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "out_of_bounds on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "C", "step": 1800 }, { "expr": "sum(increase(prometheus_target_scrapes_sample_out_of_order_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "out_of_order on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "D", "step": 1800 }, { "expr": "sum(increase(prometheus_rule_evaluation_failures_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "rule_evaluation_failure on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "G", "step": 1800 }, { "expr": "sum(increase(prometheus_tsdb_compactions_failed_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "tsdb_compactions_failed on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "K", "step": 1800 }, { "expr": "sum(increase(prometheus_tsdb_reloads_failures_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "tsdb_reloads_failures on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "L", "step": 1800 }, { "expr": "sum(increase(prometheus_tsdb_head_series_not_found{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "head_series_not_found on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "N", "step": 1800 }, { "expr": "sum(increase(prometheus_evaluator_iterations_missed_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "evaluator_iterations_missed on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "O", "step": 1800 }, { "expr": "sum(increase(prometheus_evaluator_iterations_skipped_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "evaluator_iterations_skipped on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "P", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Prometheus errors", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 14 }, "id": 57, "panels": [ ], "repeat": null, "title": "Scrape & rule duration", "type": "row" }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "description": "", "editable": true, "error": false, "fill": 1, "grid": { }, "gridPos": { "h": 7, "w": 12, "x": 0, "y": 15 }, "id": 25, "legend": { "alignAsTable": true, "avg": true, "current": true, "max": true, "min": false, "show": false, "sort": "max", "sortDesc": true, "total": false, "values": true }, "lines": true, "linewidth": 2, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "prometheus_target_interval_length_seconds{instance=\"$instance\",quantile=\"0.99\"} - $scrape_interval", "format": "time_series", "interval": "2m", "intervalFactor": 1, "legendFormat": "{{instance}}", "metric": "", "refId": "A", "step": 300 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Scrape delay (counts with 1m scrape interval)", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "cumulative" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "s", "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 12, "x": 12, "y": 15 }, "id": 14, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "Queue length", "yaxis": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(prometheus_evaluator_duration_seconds{instance=\"$instance\"}) by (instance, quantile)", "format": "time_series", "intervalFactor": 2, "legendFormat": "Queue length", "metric": "prometheus_local_storage_indexing_queue_length", "refId": "B", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Rule evaulation duration", "tooltip": { "msResolution": false, "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "s", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 22 }, "id": 58, "panels": [ ], "repeat": null, "title": "Requests & queries", "type": "row" }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 0, "y": 23 }, "id": 18, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(http_requests_total{instance=\"$instance\"}[$aggregation_interval])) by (instance, handler) > 0", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ handler }} on {{ instance }}", "metric": "", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Request count", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "none", "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 } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 6, "y": 23 }, "id": 16, "legend": { "avg": false, "current": false, "hideEmpty": true, "hideZero": true, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(sum(http_request_duration_microseconds{instance=\"$instance\"}) by (instance, handler, quantile)) by (instance, handler) > 0", "format": "time_series", "hide": false, "intervalFactor": 2, "legendFormat": "{{ handler }} on {{ instance }}", "refId": "B" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Request duration per handler", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "µs", "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 } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 12, "y": 23 }, "id": 19, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(http_request_size_bytes{instance=\"$instance\", quantile=\"0.99\"}[$aggregation_interval])) by (instance, handler) > 0", "format": "time_series", "hide": false, "intervalFactor": 2, "legendFormat": "{{ handler }} in {{ instance }}", "refId": "B" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Request size by handler", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Allocated bytes": "#F9BA8F", "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max count collector": "#bf1b00", "Max count harvester": "#bf1b00", "Max to persist": "#3F6833", "RSS": "#890F02" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 18, "y": 23 }, "id": 8, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "/Max.*/", "fill": 0, "linewidth": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(prometheus_engine_queries{instance=\"$instance\"}) by (instance, handler)", "format": "time_series", "intervalFactor": 2, "legendFormat": "Current count ", "metric": "last", "refId": "A", "step": 1800 }, { "expr": "sum(prometheus_engine_queries_concurrent_max{instance=\"$instance\"}) by (instance, handler)", "format": "time_series", "intervalFactor": 2, "legendFormat": "Max count", "metric": "last", "refId": "B", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Cont of concurent queries", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 30 }, "id": 59, "panels": [ ], "repeat": null, "title": "Alerting", "type": "row" }, { "aliasColors": { "Alert queue capacity on o collector": "#bf1b00", "Alert queue capacity on o harvester": "#bf1b00", "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 0, "y": 31 }, "id": 20, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "/.*capacity.*/", "fill": 0, "linewidth": 2 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(prometheus_notifications_queue_capacity{instance=\"$instance\"})by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "Alert queue capacity ", "metric": "prometheus_local_storage_checkpoint_last_size_bytes", "refId": "A", "step": 1800 }, { "expr": "sum(prometheus_notifications_queue_length{instance=\"$instance\"})by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "Alert queue size on ", "metric": "prometheus_local_storage_checkpoint_last_size_bytes", "refId": "B", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Alert queue size", "tooltip": { "msResolution": false, "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 8, "y": 31 }, "id": 21, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(prometheus_notifications_alertmanagers_discovered{instance=\"$instance\"}) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "Checkpoint chunks written/s", "metric": "prometheus_local_storage_checkpoint_series_chunks_written_sum", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Count of discovered alertmanagers", "tooltip": { "msResolution": false, "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "none", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 16, "y": 31 }, "id": 39, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(prometheus_notifications_dropped_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "notifications_dropped on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "F", "step": 1800 }, { "expr": "sum(increase(prometheus_rule_evaluation_failures_total{rule_type=\"alerting\",instance=\"$instance\"}[$aggregation_interval])) by (rule_type,instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "rule_evaluation_failures on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Alerting errors", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 38 }, "id": 60, "panels": [ ], "repeat": null, "title": "Service discovery", "type": "row" }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 0, "y": 39 }, "id": 43, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "increase(prometheus_target_sync_length_seconds_count{scrape_job=\"consul\", instance=\"$instance\"}[$aggregation_interval])", "format": "time_series", "intervalFactor": 2, "legendFormat": "Consul target sync count", "refId": "A", "step": 240 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Consul SD sync count", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 6, "y": 39 }, "id": 44, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "increase(prometheus_target_sync_length_seconds_count{scrape_job=\"marathon\", instance=\"$instance\"}[$aggregation_interval])", "format": "time_series", "intervalFactor": 2, "legendFormat": "Marathon target sync count", "refId": "A", "step": 240 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Marathon SD sync count", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 12, "y": 39 }, "id": 45, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "increase(prometheus_target_sync_length_seconds_count{scrape_job=\"kubernetes\"}[$aggregation_interval])", "format": "time_series", "intervalFactor": 2, "legendFormat": "Count of target synces", "refId": "A", "step": 240 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Kubernetes SD sync count", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 18, "y": 39 }, "id": 46, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(prometheus_target_scrapes_exceeded_sample_limit_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "exceeded_sample_limit on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "A", "step": 1800 }, { "expr": "sum(increase(prometheus_sd_file_read_errors_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "sd_file_read_error on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "E", "step": 1800 }, { "expr": "sum(increase(prometheus_sd_consul_rpc_failures_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "sd_consul_rpc_failure on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "H", "step": 1800 }, { "expr": "sum(increase(prometheus_sd_marathon_refresh_failures_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "sd_marathon_refresh_failure on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "I", "step": 1800 }, { "expr": "sum(increase(prometheus_sd_openstack_refresh_failures_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "interval": "", "intervalFactor": 2, "legendFormat": "sd_openstack_refresh_failure on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "J", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Service discovery errors", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 46 }, "id": 61, "panels": [ ], "repeat": null, "title": "TSDB stats", "type": "row" }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 0, "y": 47 }, "id": 36, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(prometheus_tsdb_reloads_total{instance=\"$instance\"}[30m])) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ instance }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Reloaded block from disk", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 6, "y": 47 }, "id": 5, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(prometheus_tsdb_blocks_loaded{instance=\"$instance\"}) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "Loaded data blocks", "metric": "prometheus_local_storage_memory_chunkdescs", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Loaded data blocks", "tooltip": { "msResolution": false, "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 12, "y": 47 }, "id": 3, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "prometheus_tsdb_head_series{instance=\"$instance\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "Time series count", "metric": "prometheus_local_storage_memory_series", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Time series total count", "tooltip": { "msResolution": false, "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 18, "y": 47 }, "id": 1, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(prometheus_tsdb_head_samples_appended_total{instance=\"$instance\"}[$aggregation_interval])) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "samples/s {{instance}}", "metric": "prometheus_local_storage_ingested_samples_total", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Samples Appended per second", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": "", "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 } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 54 }, "id": 62, "panels": [ ], "repeat": null, "title": "Head block stats", "type": "row" }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833", "To persist": "#9AC48A" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 0, "y": 55 }, "id": 2, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "/Max.*/", "fill": 0 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(prometheus_tsdb_head_chunks{instance=\"$instance\"}) by (instance)", "format": "time_series", "hide": false, "intervalFactor": 2, "legendFormat": "Head chunk count", "metric": "prometheus_local_storage_memory_chunks", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Head chunks count", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 8, "y": 55 }, "id": 35, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(prometheus_tsdb_head_max_time{instance=\"$instance\"}) by (instance) - min(prometheus_tsdb_head_min_time{instance=\"$instance\"}) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ instance }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Length of head block", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "ms", "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 } }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 16, "y": 55 }, "id": 4, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(prometheus_tsdb_head_chunks_created_total{instance=\"$instance\"}[$aggregation_interval])) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "created on {{ instance }}", "refId": "B" }, { "expr": "sum(rate(prometheus_tsdb_head_chunks_removed_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) * -1", "format": "time_series", "intervalFactor": 2, "legendFormat": "deleted on {{ instance }}", "refId": "C" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Head Chunks Created/Deleted per second", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 62 }, "id": 63, "panels": [ ], "repeat": null, "title": "Data maintenance", "type": "row" }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 0, "y": 63 }, "id": 33, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(prometheus_tsdb_compaction_duration_sum{instance=\"$instance\"}[30m]) / increase(prometheus_tsdb_compaction_duration_count{instance=\"$instance\"}[30m])) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ instance }}", "refId": "B" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Compaction duration", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "s", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 6, "y": 63 }, "id": 34, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(prometheus_tsdb_head_gc_duration_seconds{instance=\"$instance\"}) by (instance, quantile)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ quantile }} on {{ instance }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Go Garbage collection duration", "tooltip": { "shared": true, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "s", "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 } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 12, "y": 63 }, "id": 37, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(prometheus_tsdb_wal_truncate_duration_seconds{instance=\"$instance\"}) by (instance, quantile)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ quantile }} on {{ instance }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "WAL truncate duration seconds", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "prometheus", "fill": 1, "gridPos": { "h": 7, "w": 6, "x": 18, "y": 63 }, "id": 38, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(tsdb_wal_fsync_duration_seconds{instance=\"$instance\"}) by (instance, quantile)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{ quantile }} {{ instance }}", "refId": "A" } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "WAL fsync duration seconds", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "s", "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 } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 70 }, "id": 64, "panels": [ ], "repeat": null, "title": "RAM&CPU", "type": "row" }, { "aliasColors": { "Allocated bytes": "#7EB26D", "Allocated bytes - 1m max": "#BF1B00", "Allocated bytes - 1m min": "#BF1B00", "Allocated bytes - 5m max": "#BF1B00", "Allocated bytes - 5m min": "#BF1B00", "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833", "RSS": "#447EBC" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "decimals": null, "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 0, "y": 71 }, "id": 6, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "/-/", "fill": 0 }, { "alias": "collector heap size", "color": "#E0752D", "fill": 0, "linewidth": 2 }, { "alias": "collector kubernetes memory limit", "color": "#BF1B00", "fill": 0, "linewidth": 3 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(process_resident_memory_bytes{instance=\"$instance\"}) by (instance)", "format": "time_series", "hide": false, "intervalFactor": 2, "legendFormat": "Total resident memory - {{instance}}", "metric": "process_resident_memory_bytes", "refId": "B", "step": 1800 }, { "expr": "sum(go_memstats_alloc_bytes{instance=\"$instance\"}) by (instance)", "format": "time_series", "hide": false, "intervalFactor": 2, "legendFormat": "Total llocated bytes - {{instance}}", "metric": "go_memstats_alloc_bytes", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Memory", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { "Allocated bytes": "#F9BA8F", "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833", "RSS": "#890F02" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 8, "y": 71 }, "id": 7, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "rate(go_memstats_alloc_bytes_total{instance=\"$instance\"}[$aggregation_interval])", "format": "time_series", "intervalFactor": 2, "legendFormat": "Allocated Bytes/s", "metric": "go_memstats_alloc_bytes", "refId": "A", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Allocations per second", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "decimals": 2, "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 8, "x": 16, "y": 71 }, "id": 9, "legend": { "alignAsTable": false, "avg": false, "current": false, "hideEmpty": false, "max": false, "min": false, "rightSide": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(rate(process_cpu_seconds_total{instance=\"$instance\"}[$aggregation_interval])) by (instance)", "format": "time_series", "intervalFactor": 2, "legendFormat": "CPU/s", "metric": "prometheus_local_storage_ingested_samples_total", "refId": "B", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "CPU per second", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ "avg" ] }, "yaxes": [ { "format": "none", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 78 }, "id": 65, "panels": [ ], "repeat": null, "title": "Heapster description", "type": "row" }, { "content": "Two rows bellow can serve as example if you are running **Prometheus** in **Kubernetes** and uses **Heapster** with **InfluxDB**.\n\nThe schema and queries are very hard to generalize so you will have to tweak them but I leave them here for inspiration.", "gridPos": { "h": 3, "w": 24, "x": 0, "y": 79 }, "id": 51, "links": [ ], "mode": "markdown", "title": "Heapster rows", "type": "text" }, { "collapsed": true, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 82 }, "id": 66, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$influx_datasource", "decimals": 2, "fill": 1, "gridPos": { "h": 7, "w": 12, "x": 0, "y": 91 }, "id": 28, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "cpu/limit.mean ", "color": "#C15C17", "fill": 0 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\")/1000 as \"usage_rate.mean\" FROM \"cpu/usage_rate\" WHERE \"type\" = 'pod_container' AND \"container_name\" =~ /prometheus/ AND $timeFilter GROUP BY container_name, time($__interval) fill(null)", "rawQuery": true, "refId": "A", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "hide": false, "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\")/1000 as \"usage_rate.mean\" FROM \"cpu/usage_rate\" WHERE \"type\" = 'pod_container' AND \"container_name\" =~ /prometheus/ AND $timeFilter GROUP BY container_name, time($__interval) fill(null)", "rawQuery": true, "refId": "B", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "hide": false, "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\")/1000 as \"CPU LIMIT\" FROM \"cpu/limit\" WHERE \"type\" = 'pod_container' AND \"container_name\" =~ /prometheus/ AND $timeFilter GROUP BY container_name, time($__interval) fill(null)", "rawQuery": true, "refId": "C", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "CPU usage/s", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$influx_datasource", "fill": 1, "gridPos": { "h": 7, "w": 12, "x": 12, "y": 91 }, "id": 29, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "cpu/limit.mean ", "color": "#C15C17", "fill": 0 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"usage_rate.mean\" FROM \"memory/usage\" WHERE \"type\" = 'pod_container' AND \"container_name\" =~ /prometheus/ AND $timeFilter GROUP BY container_name, time($__interval) fill(null)", "rawQuery": true, "refId": "A", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "hide": false, "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"usage_rate.mean\" FROM \"memory/usage\" WHERE \"type\" = 'pod_container' AND \"container_name\" =~ /prometheus/ AND $timeFilter GROUP BY container_name, time($__interval) fill(null)", "rawQuery": true, "refId": "B", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "hide": false, "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"CPU LIMIT\" FROM \"memory/limit\" WHERE \"type\" = 'pod_container' AND \"container_name\" =~ /prometheus/ AND $timeFilter GROUP BY container_name, time($__interval) fill(null)", "rawQuery": true, "refId": "C", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Memory usage", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "decbytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] } ], "repeat": null, "title": "Heapster RAM&CPU", "type": "row" }, { "collapsed": true, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 83 }, "id": 67, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$influx_datasource", "fill": 1, "gridPos": { "h": 7, "w": 10, "x": 0, "y": 92 }, "id": 30, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "/errors/", "color": "#C15C17", "pointradius": 3, "points": true } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"rx_rate.mean \" FROM \"network/rx_rate\" WHERE \"labels\" =~ /app:prometheus/ AND $timeFilter GROUP BY labels, time($__interval) fill(null)", "rawQuery": true, "refId": "A", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"rx_rate.mean\" FROM \"network/rx_rate\" WHERE \"labels\" =~ /app:prometheus/ AND $timeFilter GROUP BY labels, time($__interval) fill(null)", "rawQuery": true, "refId": "B", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\")*-1 as \"tx_rate.mean \" FROM \"network/tx_rate\" WHERE \"labels\" =~ /app:prometheus/ AND $timeFilter GROUP BY labels, time($__interval) fill(null)", "rawQuery": true, "refId": "C", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\")*-1 as \"tx_rate.mean\" FROM \"network/tx_rate\" WHERE \"labels\" =~ /app:prometheus/ AND $timeFilter GROUP BY labels, time($__interval) fill(null)", "rawQuery": true, "refId": "D", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"rx_errors_rate.mean\" FROM \"network/rx_errors_rate\" WHERE \"labels\" =~ /app:prometheus/ AND $timeFilter AND \"value\">0 GROUP BY labels, time($__interval) fill(null)", "rawQuery": true, "refId": "E", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"rx_errors_rate.mean\" FROM \"network/rx_errors_rate\" WHERE \"labels\" =~ /app:prometheus/ AND $timeFilter AND \"value\">0 GROUP BY labels, time($__interval) fill(null)", "rawQuery": true, "refId": "F", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"tx_errors_rate.mean\" FROM \"network/rx_errors_rate\" WHERE \"labels\" =~ /app:prometheus/ AND $timeFilter AND \"value\">0 GROUP BY labels, time($__interval) fill(null)", "rawQuery": true, "refId": "G", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"tx_errors_rate.mean\" FROM \"network/rx_errors_rate\" WHERE \"labels\" =~ /app:prometheus/ AND $timeFilter AND \"value\">0 GROUP BY labels, time($__interval) fill(null)", "rawQuery": true, "refId": "H", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Network rx[IN] / tx[OUT] in bytes/s", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "bytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$influx_datasource", "fill": 1, "gridPos": { "h": 7, "w": 10, "x": 10, "y": 92 }, "id": 31, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "cpu/limit.mean", "color": "#C15C17", "fill": 0 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"DISK_USAGE.mean \" FROM \"filesystem/usage\" WHERE \"labels\" =~ /data-prometheus:true/ AND $timeFilter GROUP BY time($__interval) fill(null)", "rawQuery": true, "refId": "A", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "hide": false, "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"DISK_USAGE.mean \" FROM \"filesystem/usage\" WHERE \"labels\" =~ /data-prometheus:true/ AND $timeFilter GROUP BY time($__interval) fill(null)", "rawQuery": true, "refId": "B", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "hide": false, "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"DISK_LIMIT.mean \" FROM \"filesystem/limit\" WHERE \"labels\" =~ /data-prometheus:true/ AND $timeFilter GROUP BY time($__interval) fill(null)", "rawQuery": true, "refId": "C", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "hide": false, "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"DISK_LIMIT.mean\" FROM \"filesystem/limit\" WHERE \"labels\" =~ /data-prometheus:true/ AND $timeFilter GROUP BY time($__interval) fill(null)", "rawQuery": true, "refId": "D", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Disk usage", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "decbytes", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] }, { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$influx_datasource", "fill": 1, "gridPos": { "h": 7, "w": 4, "x": 20, "y": 92 }, "id": 32, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "connected", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ { "alias": "cpu/limit.mean ", "color": "#C15C17", "fill": 0 } ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"FREE_INODES.mean \" FROM \"filesystem/inodes_free\" WHERE \"labels\" =~ /data-prometheus:true/ AND $timeFilter GROUP BY time($__interval) fill(null)", "rawQuery": true, "refId": "A", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "hide": false, "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"FREE_INODES.mean\" FROM \"filesystem/inodes_free\" WHERE \"labels\" =~ /data-prometheus:true/ AND $timeFilter GROUP BY time($__interval) fill(null)", "rawQuery": true, "refId": "B", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"USED_INODES.mean \" FROM \"filesystem/inodes\" WHERE \"labels\" =~ /data-prometheus:true/ AND $timeFilter GROUP BY time($__interval) fill(null)", "rawQuery": true, "refId": "C", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] }, { "dsType": "influxdb", "groupBy": [ { "params": [ "$__interval" ], "type": "time" }, { "params": [ "null" ], "type": "fill" } ], "orderByTime": "ASC", "policy": "default", "query": "SELECT mean(\"value\") as \"USED_INODES.mean\" FROM \"filesystem/inodes\" WHERE \"labels\" =~ /data-prometheus:true/ AND $timeFilter GROUP BY time($__interval) fill(null)", "rawQuery": true, "refId": "D", "resultFormat": "time_series", "select": [ [ { "params": [ "value" ], "type": "field" }, { "params": [ ], "type": "mean" } ] ], "tags": [ ] } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Number of free INODES", "tooltip": { "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] } ], "repeat": null, "title": "Heapster host stats", "type": "row" }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, "y": 84 }, "id": 68, "panels": [ ], "repeat": null, "title": "Contrac errors", "type": "row" }, { "aliasColors": { "Chunks": "#1F78C1", "Chunks to persist": "#508642", "Max chunks": "#052B51", "Max to persist": "#3F6833" }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "editable": true, "error": false, "fill": 1, "gridPos": { "h": 7, "w": 24, "x": 0, "y": 85 }, "id": 47, "legend": { "avg": false, "current": false, "max": false, "min": false, "show": false, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "sum(increase(net_conntrack_dialer_conn_failed_total{instance=\"$instance\"}[$aggregation_interval])) by (instance) > 0", "format": "time_series", "hide": false, "interval": "", "intervalFactor": 2, "legendFormat": "conntrack_dialer_conn_failed on {{ instance }}", "metric": "prometheus_local_storage_chunk_ops_total", "refId": "M", "step": 1800 } ], "thresholds": [ ], "timeFrom": null, "timeRegions": [ ], "timeShift": null, "title": "Net errors", "tooltip": { "msResolution": false, "shared": true, "sort": 2, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": "0", "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ], "yaxis": { "align": false, "alignLevel": null } } ], "refresh": "5m", "schemaVersion": 16, "style": "dark", "tags": [ "custom" ], "templating": { "list": [ { "auto": true, "auto_count": 30, "auto_min": "2m", "current": { "text": "auto", "value": "$__auto_interval_aggregation_interval" }, "hide": 0, "label": "aggregation intarval", "name": "aggregation_interval", "options": [ { "selected": true, "text": "auto", "value": "$__auto_interval_aggregation_interval" }, { "selected": false, "text": "1m", "value": "1m" }, { "selected": false, "text": "10m", "value": "10m" }, { "selected": false, "text": "30m", "value": "30m" }, { "selected": false, "text": "1h", "value": "1h" }, { "selected": false, "text": "6h", "value": "6h" }, { "selected": false, "text": "12h", "value": "12h" }, { "selected": false, "text": "1d", "value": "1d" }, { "selected": false, "text": "7d", "value": "7d" }, { "selected": false, "text": "14d", "value": "14d" }, { "selected": false, "text": "30d", "value": "30d" } ], "query": "1m,10m,30m,1h,6h,12h,1d,7d,14d,30d", "refresh": 2, "skipUrlSync": false, "type": "interval" }, { "allValue": null, "current": { "text": "10.32.0.53:9090", "value": "10.32.0.53:9090" }, "datasource": "$datasource", "definition": "", "hide": 0, "includeAll": false, "label": "Instance", "multi": false, "name": "instance", "options": [ ], "query": "label_values(prometheus_build_info, instance)", "refresh": 2, "regex": "", "skipUrlSync": false, "sort": 2, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "current": { "text": "60", "value": "60" }, "hide": 0, "label": "Scrape interval seconds", "name": "scrape_interval", "options": [ { "text": "60", "value": "60" } ], "query": "60", "skipUrlSync": false, "type": "constant" }, { "current": { "text": "prometheus", "value": "prometheus" }, "hide": 0, "label": "Prometheus datasource", "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "skipUrlSync": false, "type": "datasource" }, { "current": { "text": "No data sources found", "value": "" }, "hide": 0, "label": "InfluxDB datasource", "name": "influx_datasource", "options": [ ], "query": "influxdb", "refresh": 1, "regex": "", "skipUrlSync": false, "type": "datasource" } ] }, "time": { "from": "now-3h", "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": "Prometheus2.0", "uid": "XmsJC9mRz", "version": 1 } kind: ConfigMap metadata: name: grafana-dashboard-prometheus-dashboard namespace: monitoring - apiVersion: v1 data: statefulset.json: |- { "__inputs": [ ], "__requires": [ ], "annotations": { "list": [ ] }, "editable": false, "gnetId": null, "graphTooltip": 0, "hideControls": false, "id": null, "links": [ ], "refresh": "", "rows": [ { "collapse": false, "collapsed": false, "panels": [ { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "$datasource", "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 2, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "cores", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 4, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "lineColor": "rgb(31, 120, 193)", "show": true }, "tableColumn": "", "targets": [ { "expr": "sum(rate(container_cpu_usage_seconds_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}[3m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "", "title": "CPU", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "0", "value": "null" } ], "valueName": "current" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "$datasource", "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 3, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "GB", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 4, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "lineColor": "rgb(31, 120, 193)", "show": true }, "tableColumn": "", "targets": [ { "expr": "sum(container_memory_usage_bytes{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}) / 1024^3", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "", "title": "Memory", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "0", "value": "null" } ], "valueName": "current" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "$datasource", "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 4, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "Bps", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 4, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "lineColor": "rgb(31, 120, 193)", "show": true }, "tableColumn": "", "targets": [ { "expr": "sum(rate(container_network_transmit_bytes_total{job=\"kubelet\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=\"$namespace\",pod_name=~\"$statefulset.*\"}[3m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "", "title": "Network", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "0", "value": "null" } ], "valueName": "current" } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "height": "100px", "panels": [ { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "$datasource", "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 5, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 3, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "max(kube_statefulset_replicas{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", statefulset=\"$statefulset\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "", "title": "Desired Replicas", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "0", "value": "null" } ], "valueName": "current" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "$datasource", "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 6, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 3, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "min(kube_statefulset_status_replicas_current{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", statefulset=\"$statefulset\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "", "title": "Replicas of current version", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "0", "value": "null" } ], "valueName": "current" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "$datasource", "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 7, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 3, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "max(kube_statefulset_status_observed_generation{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", statefulset=\"$statefulset\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "", "title": "Observed Generation", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "0", "value": "null" } ], "valueName": "current" }, { "cacheTimeout": null, "colorBackground": false, "colorValue": false, "colors": [ "#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a" ], "datasource": "$datasource", "format": "none", "gauge": { "maxValue": 100, "minValue": 0, "show": false, "thresholdLabels": false, "thresholdMarkers": true }, "gridPos": { }, "id": 8, "interval": null, "links": [ ], "mappingType": 1, "mappingTypes": [ { "name": "value to text", "value": 1 }, { "name": "range to text", "value": 2 } ], "maxDataPoints": 100, "nullPointMode": "connected", "nullText": null, "postfix": "", "postfixFontSize": "50%", "prefix": "", "prefixFontSize": "50%", "rangeMaps": [ { "from": "null", "text": "N/A", "to": "null" } ], "span": 3, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, "lineColor": "rgb(31, 120, 193)", "show": false }, "tableColumn": "", "targets": [ { "expr": "max(kube_statefulset_metadata_generation{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "", "refId": "A" } ], "thresholds": "", "title": "Metadata Generation", "tooltip": { "shared": false }, "type": "singlestat", "valueFontSize": "80%", "valueMaps": [ { "op": "=", "text": "0", "value": "null" } ], "valueName": "current" } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" }, { "collapse": false, "collapsed": false, "panels": [ { "aliasColors": { }, "bars": false, "dashLength": 10, "dashes": false, "datasource": "$datasource", "fill": 1, "gridPos": { }, "id": 9, "legend": { "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, "rightSide": false, "show": true, "total": false, "values": false }, "lines": true, "linewidth": 1, "links": [ ], "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, "renderer": "flot", "repeat": null, "seriesOverrides": [ ], "spaceLength": 10, "stack": false, "steppedLine": false, "targets": [ { "expr": "max(kube_statefulset_replicas{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "replicas specified", "refId": "A" }, { "expr": "max(kube_statefulset_status_replicas{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "replicas created", "refId": "B" }, { "expr": "min(kube_statefulset_status_replicas_ready{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "ready", "refId": "C" }, { "expr": "min(kube_statefulset_status_replicas_current{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "replicas of current version", "refId": "D" }, { "expr": "min(kube_statefulset_status_replicas_updated{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "updated", "refId": "E" } ], "thresholds": [ ], "timeFrom": null, "timeShift": null, "title": "Replicas", "tooltip": { "shared": false, "sort": 0, "value_type": "individual" }, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, "show": true, "values": [ ] }, "yaxes": [ { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true }, { "format": "short", "label": null, "logBase": 1, "max": null, "min": null, "show": true } ] } ], "repeat": null, "repeatIteration": null, "repeatRowId": null, "showTitle": false, "title": "Dashboard Row", "titleSize": "h6", "type": "row" } ], "schemaVersion": 14, "style": "dark", "tags": [ "kubernetes-mixin" ], "templating": { "list": [ { "current": { "text": "Prometheus", "value": "Prometheus" }, "hide": 0, "label": null, "name": "datasource", "options": [ ], "query": "prometheus", "refresh": 1, "regex": "", "type": "datasource" }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 2, "includeAll": false, "label": "cluster", "multi": false, "name": "cluster", "options": [ ], "query": "label_values(kube_statefulset_metadata_generation, cluster)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "Namespace", "multi": false, "name": "namespace", "options": [ ], "query": "label_values(kube_statefulset_metadata_generation{job=\"kube-state-metrics\"}, namespace)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false }, { "allValue": null, "current": { }, "datasource": "$datasource", "hide": 0, "includeAll": false, "label": "Name", "multi": false, "name": "statefulset", "options": [ ], "query": "label_values(kube_statefulset_metadata_generation{job=\"kube-state-metrics\", namespace=\"$namespace\"}, statefulset)", "refresh": 2, "regex": "", "sort": 0, "tagValuesQuery": "", "tags": [ ], "tagsQuery": "", "type": "query", "useTags": false } ] }, "time": { "from": "now-1h", "to": "now" }, "timepicker": { "refresh_intervals": [ "5s", "10s", "30s", "1m", "5m", "15m", "30m", "1h", "2h", "1d" ], "time_options": [ "5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d" ] }, "timezone": "", "title": "Kubernetes / StatefulSets", "uid": "a31c1f46e6f727cb37c0d731a7245005", "version": 0 } kind: ConfigMap metadata: name: grafana-dashboard-statefulset namespace: monitoring kind: ConfigMapList