EDGEPOD-157 Upgrade Aether dashboards for better support of multi-sites
Change-Id: I525b8a75739a0e8271dd0a532ba52faf2627534c
diff --git a/nem-monitoring/Chart.yaml b/nem-monitoring/Chart.yaml
index 9a4c265..44b9a31 100644
--- a/nem-monitoring/Chart.yaml
+++ b/nem-monitoring/Chart.yaml
@@ -18,4 +18,4 @@
description: Time Series Storage and Dashboard for CORD
icon: https://guide.opencord.org/logos/cord.svg
-version: 1.3.4
+version: 1.3.5
diff --git a/nem-monitoring/grafana-dashboards/aether-central-pods.json b/nem-monitoring/grafana-dashboards/aether-central-pods.json
deleted file mode 100644
index 0e2488d..0000000
--- a/nem-monitoring/grafana-dashboards/aether-central-pods.json
+++ /dev/null
@@ -1,751 +0,0 @@
-{
- "annotations": {
- "list": [
- {
- "builtIn": 1,
- "datasource": "-- Grafana --",
- "enable": true,
- "hide": true,
- "iconColor": "rgba(0, 211, 255, 1)",
- "name": "Annotations & Alerts",
- "type": "dashboard"
- }
- ]
- },
- "editable": true,
- "gnetId": null,
- "graphTooltip": 0,
- "id": 3,
- "links": [],
- "panels": [
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 0
- },
- "id": 25,
- "panels": [],
- "repeat": "pod",
- "title": "$pod usage",
- "type": "row"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": null,
- "description": "",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 5,
- "w": 6,
- "x": 0,
- "y": 1
- },
- "id": 10,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "maxPerRow": 12,
- "nullPointMode": "null",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "scopedVars": {
- "pod": {
- "selected": false,
- "text": "mme",
- "value": "mme"
- }
- },
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "100 * sum(irate(container_cpu_usage_seconds_total{pod_name=\"$pod-0\", image!=\"\", container_name!=\"POD\"}[3m])) by (pod_name) / sum(container_spec_cpu_quota{image!=\"\", container_name!=\"POD\"} / container_spec_cpu_period{image!=\"\", container_name!=\"POD\"}) by (pod_name)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "",
- "refId": "A"
- }
- ],
- "thresholds": "85,95",
- "title": "CPU usage",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": null,
- "description": "",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 5,
- "w": 6,
- "x": 6,
- "y": 1
- },
- "id": 12,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "maxPerRow": 12,
- "nullPointMode": "null",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "100 * (sum(container_memory_usage_bytes{pod_name=\"$pod-0\", container_name!=\"POD\"}) / sum(container_spec_memory_limit_bytes{pod_name=\"$pod-0\", container_name!=\"POD\"}))",
- "format": "time_series",
- "hide": false,
- "instant": false,
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A"
- }
- ],
- "thresholds": "85,95",
- "title": "Memory usage",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 2,
- "fillGradient": 0,
- "gridPos": {
- "h": 9,
- "w": 12,
- "x": 12,
- "y": 1
- },
- "id": 2,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "repeatDirection": "v",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(container_network_receive_bytes_total{pod_name=\"$pod-0\", interface!=\"tunl0\"}[3m]))by (interface)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "rx: {{ interface }}",
- "refId": "B"
- },
- {
- "expr": "-sum(rate(container_network_transmit_bytes_total{pod_name=\"$pod-0\", interface!=\"tunl0\"}[3m]))by (interface)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "tx: {{ interface }}",
- "refId": "A"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "Network I/O",
- "tooltip": {
- "shared": true,
- "sort": 1,
- "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": null,
- "show": true
- },
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "#299c46",
- "rgba(237, 129, 40, 0.89)",
- "#d44a3a"
- ],
- "datasource": null,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 4,
- "w": 3,
- "x": 0,
- "y": 6
- },
- "id": 20,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{pod=\"$pod-0\"})",
- "interval": "10s",
- "refId": "A"
- }
- ],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "CPU limits",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "#299c46",
- "rgba(237, 129, 40, 0.89)",
- "#d44a3a"
- ],
- "datasource": null,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 4,
- "w": 3,
- "x": 3,
- "y": 6
- },
- "id": 21,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\", container_name!=\"POD\", pod_name=\"$pod-0\"}[5m]))",
- "interval": "10s",
- "refId": "A"
- }
- ],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "CPU used (5m avg)",
- "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": null,
- "decimals": 2,
- "format": "decbytes",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 4,
- "w": 3,
- "x": 6,
- "y": 6
- },
- "id": 22,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{pod=\"$pod-0\"})",
- "refId": "A"
- }
- ],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "Memory limits",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "#299c46",
- "rgba(237, 129, 40, 0.89)",
- "#d44a3a"
- ],
- "datasource": null,
- "format": "decbytes",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 4,
- "w": 3,
- "x": 9,
- "y": 6
- },
- "id": 23,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(container_memory_working_set_bytes{container_name!=\"POD\", container_name!=\"\", pod_name=\"$pod-0\"})",
- "interval": "10s",
- "refId": "A"
- }
- ],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "Memory used",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "refresh": "5s",
- "schemaVersion": 20,
- "style": "dark",
- "tags": [
- "aether"
- ],
- "templating": {
- "list": [
- {
- "allValue": ".*",
- "current": {
- "tags": [],
- "text": "All",
- "value": [
- "$__all"
- ]
- },
- "hide": 0,
- "includeAll": true,
- "label": "Pod",
- "multi": true,
- "name": "pod",
- "options": [
- {
- "selected": true,
- "text": "All",
- "value": "$__all"
- },
- {
- "selected": false,
- "text": "mme",
- "value": "mme"
- },
- {
- "selected": false,
- "text": "hss",
- "value": "hss"
- },
- {
- "selected": false,
- "text": "spgwc",
- "value": "spgwc"
- }
- ],
- "query": "mme,hss,spgwc",
- "skipUrlSync": false,
- "type": "custom"
- }
- ]
- },
- "time": {
- "from": "now-5m",
- "to": "now"
- },
- "timepicker": {
- "hidden": false,
- "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": "Aether Central Pods",
- "uid": "gKU2e27Zb",
- "version": 1
-}
\ No newline at end of file
diff --git a/nem-monitoring/grafana-dashboards/aether-clusters.json b/nem-monitoring/grafana-dashboards/aether-clusters.json
deleted file mode 100644
index b558d2a..0000000
--- a/nem-monitoring/grafana-dashboards/aether-clusters.json
+++ /dev/null
@@ -1,1709 +0,0 @@
-{
- "annotations": {
- "list": [
- {
- "builtIn": 1,
- "datasource": "-- Grafana --",
- "enable": true,
- "hide": true,
- "iconColor": "rgba(0, 211, 255, 1)",
- "name": "Annotations & Alerts",
- "type": "dashboard"
- }
- ]
- },
- "description": "Monitors Kubernetes clusters using Prometheus.",
- "editable": true,
- "gnetId": 1621,
- "graphTooltip": 0,
- "iteration": 1574589072119,
- "links": [],
- "panels": [
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 0
- },
- "id": 45,
- "panels": [],
- "title": "Total usage in all data centers",
- "type": "row"
- },
- {
- "aliasColors": {},
- "bars": false,
- "cacheTimeout": null,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 0,
- "y": 1
- },
- "id": 47,
- "legend": {
- "alignAsTable": true,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pluginVersion": "6.4.2",
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "(1 - (avg(irate(node_cpu_seconds_total{mode=\"idle\",datacenter!=\"\"}[5m]))by (datacenter))) * 100",
- "interval": "10s",
- "legendFormat": "{{ datacenter }}",
- "refId": "A"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "CPU usage (5m avg)",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percent",
- "label": null,
- "logBase": 1,
- "max": "100",
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "decimals": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 8,
- "y": 1
- },
- "id": 49,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "(1 - sum(node_memory_MemAvailable_bytes{datacenter!=\"\"})by (datacenter)/ sum(node_memory_MemTotal_bytes{datacenter!=\"\"})by (datacenter)) * 100",
- "interval": "10s",
- "legendFormat": "{{ datacenter }}",
- "refId": "A"
- }
- ],
- "thresholds": [],
- "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": "percent",
- "label": null,
- "logBase": 1,
- "max": "100",
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 16,
- "y": 1
- },
- "id": 51,
- "legend": {
- "alignAsTable": true,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum (rate (node_network_receive_bytes_total{datacenter!=\"\"}[5m]))by (datacenter)",
- "interval": "10s",
- "legendFormat": "{{ datacenter }} rx",
- "refId": "A"
- },
- {
- "expr": "- sum (rate (node_network_transmit_bytes_total{datacenter!=\"\"}[5m])) by (datacenter)",
- "interval": "10s",
- "legendFormat": "{{ datacenter }} tx",
- "refId": "B"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "Network I/O pressure",
- "tooltip": {
- "shared": true,
- "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": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 9
- },
- "id": 34,
- "panels": [],
- "title": "Total usage in $datacenter node $Node",
- "type": "row"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 5,
- "w": 8,
- "x": 0,
- "y": 10
- },
- "height": "180px",
- "id": 6,
- "interval": null,
- "isNew": true,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "(1 - avg(irate(node_cpu_seconds_total{mode=\"idle\",datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"}[5m]))) * 100",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "65, 90",
- "title": "CPU usage (5m avg)",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "editable": true,
- "error": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 5,
- "w": 8,
- "x": 8,
- "y": 10
- },
- "height": "180px",
- "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,
- "options": {},
- "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": "(1 - sum(node_memory_MemAvailable_bytes{datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"})/ sum(node_memory_MemTotal_bytes{datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"})) * 100",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "65, 90",
- "title": "Memory usage",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 5,
- "w": 8,
- "x": 16,
- "y": 10
- },
- "height": "180px",
- "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,
- "options": {},
- "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!~\"rootfs|HarddiskVolume.+\",datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"})\n - sum(node_filesystem_free_bytes{device!~\"rootfs|HarddiskVolume.+\",datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"})) / sum(node_filesystem_size_bytes{device!~\"rootfs|HarddiskVolume.+\",datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"}) * 100",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "",
- "metric": "",
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "65, 90",
- "title": "Filesystem usage",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 3,
- "w": 4,
- "x": 0,
- "y": 15
- },
- "height": "1px",
- "id": 11,
- "interval": null,
- "isNew": true,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": " cores",
- "postfixFontSize": "30%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "(1 - (avg(irate(node_cpu_seconds_total{mode=\"idle\",datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"}[5m])))) * sum(machine_cpu_cores{datacenter=\"$datacenter\",kubernetes_io_hostname=~\"^$Node$\"})",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "",
- "title": "Used",
- "type": "singlestat",
- "valueFontSize": "50%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 3,
- "w": 4,
- "x": 4,
- "y": 15
- },
- "height": "1px",
- "id": 12,
- "interval": null,
- "isNew": true,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": " cores",
- "postfixFontSize": "30%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum (machine_cpu_cores{datacenter=\"$datacenter\",kubernetes_io_hostname=~\"^$Node$\"})",
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "",
- "title": "Total",
- "type": "singlestat",
- "valueFontSize": "50%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "format": "bytes",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 3,
- "w": 4,
- "x": 8,
- "y": 15
- },
- "height": "1px",
- "id": 9,
- "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,
- "options": {},
- "postfix": "",
- "postfixFontSize": "20%",
- "prefix": "",
- "prefixFontSize": "20%",
- "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{device!~\"rootfs|HarddiskVolume.+\",datacenter=\"$datacenter\",kubernetes_node =~\"^$Node$\"}) - sum(node_memory_MemAvailable_bytes{device!~\"rootfs|HarddiskVolume.+\",datacenter=\"$datacenter\",kubernetes_node =~\"^$Node$\"})",
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "",
- "title": "Used",
- "type": "singlestat",
- "valueFontSize": "50%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "format": "bytes",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 3,
- "w": 4,
- "x": 12,
- "y": 15
- },
- "height": "1px",
- "id": 10,
- "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,
- "options": {},
- "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 (machine_memory_bytes{datacenter=\"$datacenter\",kubernetes_io_hostname=~\"^$Node$\"})",
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "",
- "title": "Total",
- "type": "singlestat",
- "valueFontSize": "50%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "format": "bytes",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 3,
- "w": 4,
- "x": 16,
- "y": 15
- },
- "height": "1px",
- "id": 13,
- "interval": null,
- "isNew": true,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(node_filesystem_size_bytes{device!~\"rootfs|HarddiskVolume.+\",datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"})\n - sum(node_filesystem_free_bytes{device!~\"rootfs|HarddiskVolume.+\",datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"})",
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "",
- "title": "Used",
- "type": "singlestat",
- "valueFontSize": "50%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "format": "bytes",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 3,
- "w": 4,
- "x": 20,
- "y": 15
- },
- "height": "1px",
- "id": 14,
- "interval": null,
- "isNew": true,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(node_filesystem_size_bytes{device!~\"rootfs|HarddiskVolume.+\",datacenter=\"$datacenter\",kubernetes_node=~\"^$Node$\"})",
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": "",
- "title": "Total",
- "type": "singlestat",
- "valueFontSize": "50%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "fill": 1,
- "fillGradient": 0,
- "grid": {},
- "gridPos": {
- "h": 5,
- "w": 24,
- "x": 0,
- "y": 18
- },
- "height": "200px",
- "id": 32,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": true,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": false,
- "sideWidth": 200,
- "sort": "current",
- "sortDesc": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum (rate (node_network_receive_bytes_total{datacenter=\"$datacenter\"}[5m]))by (kubernetes_node)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "{{kubernetes_node}} Received",
- "metric": "network",
- "refId": "A",
- "step": 10
- },
- {
- "expr": "- sum (rate (node_network_transmit_bytes_total{datacenter=\"$datacenter\"}[5m])) by (kubernetes_node)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "{{kubernetes_node}} Sent",
- "metric": "network",
- "refId": "B",
- "step": 10
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "Network I/O pressure",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 23
- },
- "id": 35,
- "panels": [],
- "title": "Pods CPU usage in $datacenter namespace $namespace",
- "type": "row"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "Prometheus",
- "decimals": 3,
- "editable": true,
- "error": false,
- "fill": 1,
- "fillGradient": 0,
- "grid": {},
- "gridPos": {
- "h": 7,
- "w": 24,
- "x": 0,
- "y": 24
- },
- "height": "",
- "id": 17,
- "isNew": true,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "sideWidth": null,
- "sort": "avg",
- "sortDesc": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": true,
- "targets": [
- {
- "expr": "sum (rate (container_cpu_usage_seconds_total{pod_name!=\"\", image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\", namespace=~\"^$namespace$\"}[5m])) by (pod_name)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "{{ pod_name }}",
- "metric": "container_cpu",
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "Pods CPU usage (5m avg)",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 2,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percentunit",
- "label": "cores",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 31
- },
- "id": 39,
- "panels": [],
- "title": "Pods memory usage in $datacenter namespace $namespace",
- "type": "row"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "Prometheus",
- "decimals": 2,
- "editable": true,
- "error": false,
- "fill": 1,
- "fillGradient": 0,
- "grid": {},
- "gridPos": {
- "h": 7,
- "w": 24,
- "x": 0,
- "y": 32
- },
- "id": 25,
- "isNew": true,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "sideWidth": null,
- "sort": "current",
- "sortDesc": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": true,
- "targets": [
- {
- "expr": "sum(container_memory_working_set_bytes{container_name!=\"POD\", container_name!=\"\", namespace=~\"^$namespace$\", datacenter=\"$datacenter\"})by (pod_name)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "{{ pod_name }}",
- "metric": "container_memory_usage:sort_desc",
- "refId": "A",
- "step": 10
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "Pods memory usage",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 2,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 39
- },
- "id": 41,
- "panels": [],
- "title": "Pods network I/O in $datacenter namespace $namespace",
- "type": "row"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 7,
- "w": 24,
- "x": 0,
- "y": 40
- },
- "id": 43,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum (rate (container_network_receive_bytes_total{pod_name!=\"\", container_name!=\"\", datacenter=\"$datacenter\",namespace=~\"^$namespace$\"}[5m])) by (pod_name)",
- "legendFormat": "{{ pod_name }} Rcvd",
- "refId": "A"
- },
- {
- "expr": "- sum (rate (container_network_transmit_bytes_total{pod_name!=\"\", container_name!=\"\", datacenter=\"$datacenter\",namespace=~\"^$namespace$\"}[5m])) by (pod_name)",
- "legendFormat": "{{ pod_name }} Sent",
- "refId": "B"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "Pods network I/O (5m avg)",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "decimals": null,
- "format": "Bps",
- "label": "",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- }
- ],
- "refresh": "1m",
- "schemaVersion": 20,
- "style": "dark",
- "tags": [
- "aether"
- ],
- "templating": {
- "list": [
- {
- "allValue": null,
- "datasource": "Prometheus",
- "definition": "label_values(kube_pod_info, datacenter)",
- "hide": 0,
- "includeAll": false,
- "label": "Datacenter",
- "multi": false,
- "name": "datacenter",
- "options": [],
- "query": "label_values(kube_pod_info, datacenter)",
- "refresh": 2,
- "regex": "",
- "skipUrlSync": false,
- "sort": 1,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": ".*",
- "current": {
- "tags": [],
- "text": "All",
- "value": "$__all"
- },
- "datasource": "Prometheus",
- "definition": "label_values(node_load1{datacenter=\"$datacenter\"}, kubernetes_node)",
- "hide": 0,
- "includeAll": true,
- "label": null,
- "multi": false,
- "name": "Node",
- "options": [],
- "query": "label_values(node_load1{datacenter=\"$datacenter\"}, kubernetes_node)",
- "refresh": 1,
- "regex": "",
- "skipUrlSync": false,
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": ".*",
- "current": {
- "text": "All",
- "value": "$__all"
- },
- "datasource": "Prometheus",
- "definition": "label_values(kube_pod_info{datacenter=\"$datacenter\"}, namespace)",
- "hide": 0,
- "includeAll": true,
- "label": "Namespace",
- "multi": false,
- "name": "namespace",
- "options": [
- {
- "selected": true,
- "text": "All",
- "value": "$__all"
- },
- {
- "selected": false,
- "text": "default",
- "value": "default"
- },
- {
- "selected": false,
- "text": "kube-system",
- "value": "kube-system"
- },
- {
- "selected": false,
- "text": "omec",
- "value": "omec"
- }
- ],
- "query": "label_values(kube_pod_info{datacenter=\"$datacenter\"}, namespace)",
- "refresh": 0,
- "regex": "",
- "skipUrlSync": false,
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-30m",
- "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": "Aether Clusters",
- "uid": "icjpCppik",
- "version": 6
-}
\ No newline at end of file
diff --git a/nem-monitoring/grafana-dashboards/aether-edge-pods.json b/nem-monitoring/grafana-dashboards/aether-edge-pods.json
deleted file mode 100644
index e23acf8..0000000
--- a/nem-monitoring/grafana-dashboards/aether-edge-pods.json
+++ /dev/null
@@ -1,755 +0,0 @@
-{
- "annotations": {
- "list": [
- {
- "builtIn": 1,
- "datasource": "-- Grafana --",
- "enable": true,
- "hide": true,
- "iconColor": "rgba(0, 211, 255, 1)",
- "name": "Annotations & Alerts",
- "type": "dashboard"
- }
- ]
- },
- "editable": true,
- "gnetId": null,
- "graphTooltip": 0,
- "id": 4,
- "iteration": 1575804277781,
- "links": [],
- "panels": [
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 0
- },
- "id": 25,
- "panels": [],
- "repeat": "pod",
- "title": "$pod usage",
- "type": "row"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": null,
- "description": "",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 5,
- "w": 6,
- "x": 0,
- "y": 1
- },
- "id": 10,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "maxPerRow": 12,
- "nullPointMode": "null",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "100 * sum(irate(container_cpu_usage_seconds_total{pod_name=\"$pod-0\", image!=\"\", container_name!=\"POD\"}[3m])) by (pod_name) / sum(container_spec_cpu_quota{image!=\"\", container_name!=\"POD\"} / container_spec_cpu_period{image!=\"\", container_name!=\"POD\"}) by (pod_name)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "",
- "refId": "A"
- }
- ],
- "thresholds": "85,95",
- "title": "CPU usage",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "Prometheus",
- "decimals": null,
- "description": "",
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 5,
- "w": 6,
- "x": 6,
- "y": 1
- },
- "id": 12,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "maxPerRow": 12,
- "nullPointMode": "null",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "100 * (sum(container_memory_usage_bytes{pod_name=\"$pod-0\", container_name!=\"POD\"}) / sum(container_spec_memory_limit_bytes{pod_name=\"$pod-0\", container_name!=\"POD\"}))",
- "format": "time_series",
- "hide": false,
- "instant": false,
- "interval": "10s",
- "intervalFactor": 1,
- "refId": "A"
- }
- ],
- "thresholds": "85,95",
- "title": "Memory usage",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 2,
- "fillGradient": 0,
- "gridPos": {
- "h": 9,
- "w": 12,
- "x": 12,
- "y": 1
- },
- "id": 2,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "repeat": null,
- "repeatDirection": "v",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(container_network_receive_bytes_total{pod_name=\"$pod-0\", interface!=\"tunl0\"}[3m]))by (interface)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "rx: {{ interface }}",
- "refId": "B"
- },
- {
- "expr": "-sum(rate(container_network_transmit_bytes_total{pod_name=\"$pod-0\", interface!=\"tunl0\"}[3m]))by (interface)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "tx: {{ interface }}",
- "refId": "A"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "Network I/O",
- "tooltip": {
- "shared": true,
- "sort": 1,
- "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": null,
- "show": true
- },
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "#299c46",
- "rgba(237, 129, 40, 0.89)",
- "#d44a3a"
- ],
- "datasource": null,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 4,
- "w": 3,
- "x": 0,
- "y": 6
- },
- "id": 20,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{pod=\"$pod-0\"})",
- "interval": "10s",
- "refId": "A"
- }
- ],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "CPU limits",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "#299c46",
- "rgba(237, 129, 40, 0.89)",
- "#d44a3a"
- ],
- "datasource": null,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 4,
- "w": 3,
- "x": 3,
- "y": 6
- },
- "id": 21,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\", container_name!=\"POD\", pod_name=\"$pod-0\"}[5m]))",
- "interval": "10s",
- "refId": "A"
- }
- ],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "CPU used (5m avg)",
- "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": null,
- "decimals": 2,
- "format": "decbytes",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 4,
- "w": 3,
- "x": 6,
- "y": 6
- },
- "id": 22,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{pod=\"$pod-0\"})",
- "refId": "A"
- }
- ],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "Memory limits",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "#299c46",
- "rgba(237, 129, 40, 0.89)",
- "#d44a3a"
- ],
- "datasource": null,
- "format": "decbytes",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "gridPos": {
- "h": 4,
- "w": 3,
- "x": 9,
- "y": 6
- },
- "id": 23,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "options": {},
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "repeat": null,
- "repeatDirection": "v",
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(container_memory_working_set_bytes{container_name!=\"POD\", container_name!=\"\", pod_name=\"$pod-0\"})",
- "interval": "10s",
- "refId": "A"
- }
- ],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "Memory used",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "refresh": "5s",
- "schemaVersion": 20,
- "style": "dark",
- "tags": [
- "aether"
- ],
- "templating": {
- "list": [
- {
- "allValue": ".*",
- "current": {
- "tags": [],
- "text": "All",
- "value": [
- "$__all"
- ]
- },
- "hide": 0,
- "includeAll": true,
- "label": "Pod",
- "multi": true,
- "name": "pod",
- "options": [
- {
- "selected": false,
- "text": "All",
- "value": "$__all"
- },
- {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-ric",
- "value": "accelleran-cbrs-ric"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-nats",
- "value": "accelleran-cbrs-nats"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-redis",
- "value": "accelleran-cbrs-redis"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-ricweb",
- "value": "accelleran-cbrs-ricweb"
- }
- ],
- "query": "spgwu,accelleran-cbrs-ric,accelleran-cbrs-nats,accelleran-cbrs-redis,accelleran-cbrs-ricweb",
- "skipUrlSync": false,
- "type": "custom"
- }
- ]
- },
- "time": {
- "from": "now-5m",
- "to": "now"
- },
- "timepicker": {
- "hidden": false,
- "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": "Aether Edge Pods",
- "uid": "gKU2e27Zc",
- "version": 1
-}
diff --git a/nem-monitoring/grafana-dashboards/multi-site-k8s.json b/nem-monitoring/grafana-dashboards/multi-site-k8s.json
new file mode 100644
index 0000000..acc4fe6
--- /dev/null
+++ b/nem-monitoring/grafana-dashboards/multi-site-k8s.json
@@ -0,0 +1,2329 @@
+{
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "Monitors K8S cluster using Prometheus.",
+ "editable": true,
+ "gnetId": 8721,
+ "graphTooltip": 1,
+ "id": 5,
+ "iteration": 1580676696605,
+ "links": [],
+ "panels": [
+ {
+ "collapsed": false,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 34,
+ "panels": [],
+ "title": "Total usage",
+ "type": "row"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 8,
+ "x": 0,
+ "y": 1
+ },
+ "height": "180px",
+ "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,
+ "options": {},
+ "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 (rate (container_cpu_usage_seconds_total{id=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}[5m])) / sum (machine_cpu_cores{datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}) * 100",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "65, 90",
+ "title": "Cluster CPU usage (5m avg)",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "editable": true,
+ "error": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 8,
+ "x": 8,
+ "y": 1
+ },
+ "height": "180px",
+ "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,
+ "options": {},
+ "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 (container_memory_working_set_bytes{id=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}) / sum (machine_memory_bytes{datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}) * 100",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "65, 90",
+ "title": "Cluster memory usage",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 8,
+ "x": 16,
+ "y": 1
+ },
+ "height": "180px",
+ "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,
+ "options": {},
+ "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 (container_fs_usage_bytes{device=~\"^/dev/[sv]d[a-z][1-9]$\",id=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}) / sum (container_fs_limit_bytes{device=~\"^/dev/[sv]d[a-z][1-9]$\",id=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}) * 100",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "metric": "",
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "65, 90",
+ "title": "Cluster filesystem usage",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 4,
+ "x": 0,
+ "y": 6
+ },
+ "height": "1px",
+ "id": 11,
+ "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,
+ "options": {},
+ "postfix": " cores",
+ "postfixFontSize": "30%",
+ "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 (rate (container_cpu_usage_seconds_total{id=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}[5m]))",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "",
+ "title": "Used",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 4,
+ "x": 4,
+ "y": 6
+ },
+ "height": "1px",
+ "id": 12,
+ "interval": null,
+ "isNew": true,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": " cores",
+ "postfixFontSize": "30%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum (machine_cpu_cores{datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"})",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "",
+ "title": "Total",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 4,
+ "x": 8,
+ "y": 6
+ },
+ "height": "1px",
+ "id": 9,
+ "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,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "20%",
+ "prefix": "",
+ "prefixFontSize": "20%",
+ "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 (container_memory_working_set_bytes{id=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"})",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "",
+ "title": "Used",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 4,
+ "x": 12,
+ "y": 6
+ },
+ "height": "1px",
+ "id": 10,
+ "interval": null,
+ "isNew": true,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum (machine_memory_bytes{datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"})",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "",
+ "title": "Total",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 4,
+ "x": 16,
+ "y": 6
+ },
+ "height": "1px",
+ "id": 13,
+ "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,
+ "options": {},
+ "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 (container_fs_usage_bytes{device=~\"^/dev/[sv]d[a-z][1-9]$\",id=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"})",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "",
+ "title": "Used",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 4,
+ "x": 20,
+ "y": 6
+ },
+ "height": "1px",
+ "id": 14,
+ "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,
+ "options": {},
+ "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 (container_fs_limit_bytes{device=~\"^overlay$\",id=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"})",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": "",
+ "title": "Total",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "collapsed": false,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 9
+ },
+ "id": 33,
+ "panels": [],
+ "title": "Usage total",
+ "type": "row"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 3,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 0,
+ "y": 10
+ },
+ "height": "200px",
+ "id": 32,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": false,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": false,
+ "sideWidth": 200,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (rate (container_network_receive_bytes_total{datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}[5m]))",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 4,
+ "legendFormat": "Received",
+ "metric": "network",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "- sum (rate (container_network_transmit_bytes_total{datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\"}[5m]))",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 4,
+ "legendFormat": "Sent",
+ "metric": "network",
+ "refId": "B",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Network I/O pressure",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 0,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 8,
+ "y": 10
+ },
+ "height": "200px",
+ "id": 45,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": false,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": false,
+ "sideWidth": 200,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name=~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (namespace)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 4,
+ "legendFormat": "{{ namespace}}",
+ "metric": "network",
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "CPU use by namespace",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "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": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 0,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 16,
+ "y": 10
+ },
+ "height": "200px",
+ "id": 46,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": false,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": false,
+ "sideWidth": 200,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",name=~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (namespace)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 4,
+ "legendFormat": "{{ namespace}}",
+ "metric": "network",
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Memory use by namespace",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "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": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "collapsed": false,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 17
+ },
+ "id": 35,
+ "panels": [],
+ "title": "Pods CPU usage",
+ "type": "row"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 3,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 24,
+ "x": 0,
+ "y": 18
+ },
+ "height": "",
+ "id": 17,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name=~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (pod_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "{{ pod_name }}",
+ "metric": "container_cpu",
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Pods CPU usage (5m avg)",
+ "tooltip": {
+ "msResolution": true,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": "cores",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "collapsed": true,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 25
+ },
+ "id": 37,
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 3,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 24,
+ "x": 0,
+ "y": 26
+ },
+ "height": "",
+ "id": 24,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name=~\"^k8s_.*\",container_name!=\"POD\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (container_name, pod_name)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "pod: {{ pod_name }} | {{ container_name }}",
+ "metric": "container_cpu",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name!~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "container_cpu",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{rkt_container_name!=\"\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "container_cpu",
+ "refId": "C",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers CPU usage (5m avg)",
+ "tooltip": {
+ "msResolution": true,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": "cores",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "title": "Containers CPU usage",
+ "type": "row"
+ },
+ {
+ "collapsed": true,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 26
+ },
+ "id": 38,
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 3,
+ "editable": true,
+ "error": false,
+ "fill": 0,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 13,
+ "w": 24,
+ "x": 0,
+ "y": 27
+ },
+ "id": 20,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{id!=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (id)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 4,
+ "legendFormat": "{{ id }}",
+ "metric": "container_cpu",
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "All processes CPU usage (5m avg)",
+ "tooltip": {
+ "msResolution": true,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": "cores",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "repeat": null,
+ "title": "All processes CPU usage",
+ "type": "row"
+ },
+ {
+ "collapsed": false,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 27
+ },
+ "id": 39,
+ "panels": [],
+ "title": "Pods memory usage",
+ "type": "row"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 24,
+ "x": 0,
+ "y": 28
+ },
+ "id": 25,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 1,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",name=~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (pod_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "{{ pod_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Pods memory usage",
+ "tooltip": {
+ "msResolution": false,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "collapsed": true,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 35
+ },
+ "id": 41,
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 24,
+ "x": 0,
+ "y": 36
+ },
+ "id": 27,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",name=~\"^k8s_.*\",container_name!=\"POD\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (container_name, pod_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "pod: {{ pod_name }} | {{ container_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",name!~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum (container_memory_working_set_bytes{rkt_container_name!=\"\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "C",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers memory usage",
+ "tooltip": {
+ "msResolution": false,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "title": "Containers memory usage",
+ "type": "row"
+ },
+ {
+ "collapsed": true,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 36
+ },
+ "id": 42,
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 0,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 13,
+ "w": 24,
+ "x": 0,
+ "y": 37
+ },
+ "id": 28,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": 200,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (container_memory_working_set_bytes{id!=\"/\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (id)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 4,
+ "legendFormat": "{{ id }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "A",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "All processes memory usage",
+ "tooltip": {
+ "msResolution": false,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "title": "All processes memory usage",
+ "type": "row"
+ },
+ {
+ "collapsed": false,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 37
+ },
+ "id": 43,
+ "panels": [],
+ "title": "Pods network I/O",
+ "type": "row"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 24,
+ "x": 0,
+ "y": 38
+ },
+ "id": 16,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null as zero",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 1,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name=~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (pod_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "-> {{ pod_name }}",
+ "metric": "network",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name=~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (pod_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "<- {{ pod_name }}",
+ "metric": "network",
+ "refId": "B",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Pods network I/O (5m avg)",
+ "tooltip": {
+ "msResolution": false,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "collapsed": true,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 45
+ },
+ "id": 44,
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 7,
+ "w": 24,
+ "x": 0,
+ "y": 46
+ },
+ "id": 30,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null as zero",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name=~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (container_name, pod_name)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "-> pod: {{ pod_name }} | {{ container_name }}",
+ "metric": "network",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name=~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (container_name, pod_name)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "<- pod: {{ pod_name }} | {{ container_name }}",
+ "metric": "network",
+ "refId": "D",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name!~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "-> docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "network",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name!~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "<- docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "network",
+ "refId": "C",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_network_transmit_bytes_total{rkt_container_name!=\"\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "-> rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "network",
+ "refId": "E",
+ "step": 10
+ },
+ {
+ "expr": "- sum (rate (container_network_transmit_bytes_total{rkt_container_name!=\"\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "<- rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "network",
+ "refId": "F",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers network I/O (5m avg)",
+ "tooltip": {
+ "msResolution": false,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "title": "Containers network I/O",
+ "type": "row"
+ }
+ ],
+ "refresh": "1m",
+ "schemaVersion": 20,
+ "style": "dark",
+ "tags": [
+ "aether"
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "datasource": "Prometheus",
+ "definition": "label_values(kube_pod_info, datacenter)",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Site",
+ "multi": false,
+ "name": "Site",
+ "query": "label_values(kube_pod_info, datacenter)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": ".*",
+ "current": {
+ "text": "All",
+ "value": "$__all"
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(node_load1{datacenter=\"$Site\"}, kubernetes_node)",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "Node",
+ "options": [],
+ "query": "label_values(node_load1{datacenter=\"$Site\"}, kubernetes_node)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": ".*",
+ "current": {
+ "text": "All",
+ "value": "$__all"
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(kube_pod_info{datacenter=\"$Site\"}, namespace)",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": true,
+ "name": "Namespace",
+ "options": [],
+ "query": "label_values(kube_pod_info{datacenter=\"$Site\"}, namespace)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "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": "Multi-Site K8S Cluster Monitoring",
+ "uid": "1bI4P-xiz",
+ "version": 1
+}
diff --git a/nem-monitoring/grafana-dashboards/multi-site-system-summary.json b/nem-monitoring/grafana-dashboards/multi-site-system-summary.json
new file mode 100644
index 0000000..64b6bb5
--- /dev/null
+++ b/nem-monitoring/grafana-dashboards/multi-site-system-summary.json
@@ -0,0 +1,508 @@
+{
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "Support Node Exporter v0.16 and above.",
+ "editable": true,
+ "gnetId": 11074,
+ "graphTooltip": 0,
+ "id": 9,
+ "iteration": 1580631219620,
+ "links": [],
+ "panels": [
+ {
+ "collapsed": false,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 179,
+ "panels": [],
+ "repeat": "site",
+ "title": "$site",
+ "type": "row"
+ },
+ {
+ "datasource": "Prometheus",
+ "gridPos": {
+ "h": 10,
+ "w": 8,
+ "x": 0,
+ "y": 1
+ },
+ "id": 177,
+ "options": {
+ "displayMode": "lcd",
+ "fieldOptions": {
+ "calcs": [
+ "last"
+ ],
+ "defaults": {
+ "mappings": [],
+ "max": 100,
+ "min": 0,
+ "thresholds": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "#EAB839",
+ "value": 60
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ],
+ "title": "",
+ "unit": "percent"
+ },
+ "override": {},
+ "values": false
+ },
+ "orientation": "horizontal"
+ },
+ "pluginVersion": "6.4.2",
+ "repeat": null,
+ "repeatDirection": "v",
+ "targets": [
+ {
+ "expr": "100 - (avg(irate(node_cpu_seconds_total{datacenter=\"$site\",mode=\"idle\"}[30m])) * 100)",
+ "instant": true,
+ "legendFormat": "CPU Busy",
+ "refId": "A"
+ },
+ {
+ "expr": "avg(irate(node_cpu_seconds_total{datacenter=\"$site\",mode=\"iowait\"}[30m])) * 100",
+ "hide": true,
+ "instant": true,
+ "legendFormat": "Busy Iowait",
+ "refId": "C"
+ },
+ {
+ "expr": "(1 - (node_memory_MemAvailable_bytes{datacenter=\"$site\"} / (node_memory_MemTotal_bytes{datacenter=\"$site\"})))* 100",
+ "instant": true,
+ "legendFormat": "Memory Usage ({{kubernetes_node}})",
+ "refId": "B"
+ },
+ {
+ "expr": "100 - ((node_filesystem_avail_bytes{datacenter=\"$site\",mountpoint=\"$maxmount\",fstype=~\"ext4|xfs\"} * 100) / node_filesystem_size_bytes{datacenter=\"$site\",mountpoint=\"$maxmount\",fstype=~\"ext4|xfs\"})",
+ "hide": false,
+ "instant": true,
+ "legendFormat": "Disk Usage ({{kubernetes_node}})",
+ "refId": "D"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "",
+ "type": "bargauge"
+ },
+ {
+ "aliasColors": {
+ "15分钟": "#6ED0E0",
+ "1分钟": "#BF1B00",
+ "5分钟": "#CCA300"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 1,
+ "grid": {},
+ "gridPos": {
+ "h": 10,
+ "w": 8,
+ "x": 8,
+ "y": 1
+ },
+ "height": "300",
+ "id": 13,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "maxPerRow": 6,
+ "nullPointMode": "null as zero",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_load1{datacenter=\"$site\"}",
+ "format": "time_series",
+ "hide": true,
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_1m",
+ "metric": "",
+ "refId": "A",
+ "step": 20,
+ "target": ""
+ },
+ {
+ "expr": "node_load5{datacenter=\"$site\"}",
+ "format": "time_series",
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}",
+ "refId": "B",
+ "step": 20
+ },
+ {
+ "expr": "node_load15{datacenter=\"$site\"}",
+ "format": "time_series",
+ "hide": true,
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_15m",
+ "refId": "C",
+ "step": 20
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "System Load (5m average)",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "192.168.10.227:9100_em1_in下载": "super-light-green",
+ "192.168.10.227:9100_em1_out上传": "dark-blue"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "fill": 1,
+ "fillGradient": 3,
+ "gridPos": {
+ "h": 10,
+ "w": 8,
+ "x": 16,
+ "y": 1
+ },
+ "height": "300",
+ "id": 157,
+ "legend": {
+ "alignAsTable": true,
+ "avg": false,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/.*_transmit$/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(node_network_receive_bytes_total{datacenter=\"$site\",device!~'tap.*|veth.*|br.*|docker.*|virbr*|lo*|tunl0|cali.*'}[30m])*8",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_receive",
+ "refId": "A",
+ "step": 4
+ },
+ {
+ "expr": "irate(node_network_transmit_bytes_total{datacenter=\"$site\",device!~'tap.*|veth.*|br.*|docker.*|virbr*|lo*|tunl0|cali.*'}[30m])*8",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_transmit",
+ "refId": "B",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Network Traffic",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bps",
+ "label": "transmit（-）/receive（+）",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "refresh": false,
+ "schemaVersion": 20,
+ "style": "dark",
+ "tags": [
+ "aether"
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+ "tags": [],
+ "text": "All",
+ "value": [
+ "$__all"
+ ]
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(node_uname_info, datacenter)",
+ "hide": 0,
+ "includeAll": true,
+ "label": "Site",
+ "multi": true,
+ "name": "site",
+ "query": "label_values(node_uname_info, datacenter)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "/",
+ "value": "/"
+ },
+ "datasource": "Prometheus",
+ "definition": "query_result(topk(1,sort_desc (max(node_filesystem_size_bytes{fstype=~\"ext4|xfs\"}) by (mountpoint))))",
+ "hide": 2,
+ "includeAll": false,
+ "label": "",
+ "multi": false,
+ "name": "maxmount",
+ "options": [],
+ "query": "query_result(topk(1,sort_desc (max(node_filesystem_size_bytes{fstype=~\"ext4|xfs\"}) by (mountpoint))))",
+ "refresh": 2,
+ "regex": "/.*\\\"(.*)\\\".*/",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allFormat": "glob",
+ "allValue": null,
+ "current": {
+ "isNone": true,
+ "text": "None",
+ "value": ""
+ },
+ "datasource": "Prometheus",
+ "definition": "",
+ "hide": 2,
+ "includeAll": false,
+ "label": "环境",
+ "multi": false,
+ "multiFormat": "regex values",
+ "name": "env",
+ "options": [],
+ "query": "label_values(node_exporter_build_info,env)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allFormat": "glob",
+ "allValue": "",
+ "current": {
+ "isNone": true,
+ "text": "None",
+ "value": ""
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(node_exporter_build_info{env=~'$env'},name)",
+ "hide": 2,
+ "includeAll": false,
+ "label": "名称",
+ "multi": true,
+ "multiFormat": "regex values",
+ "name": "name",
+ "options": [],
+ "query": "label_values(node_exporter_build_info{env=~'$env'},name)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 1,
+ "tagValuesQuery": "/.*/",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-24h",
+ "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": "browser",
+ "title": "Multi-Site System Monitoring Summary",
+ "uid": "hb7fSE0Za",
+ "version": 7
+}
diff --git a/nem-monitoring/grafana-dashboards/multi-site-system.json b/nem-monitoring/grafana-dashboards/multi-site-system.json
new file mode 100644
index 0000000..31db6b3
--- /dev/null
+++ b/nem-monitoring/grafana-dashboards/multi-site-system.json
@@ -0,0 +1,2255 @@
+{
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "Support Node Exporter v0.16 and above.",
+ "editable": true,
+ "gnetId": 11074,
+ "graphTooltip": 0,
+ "id": 7,
+ "iteration": 1580631029618,
+ "links": [],
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorPostfix": false,
+ "colorPrefix": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 1,
+ "description": "",
+ "format": "s",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 2,
+ "x": 0,
+ "y": 0
+ },
+ "hideTimeOverride": true,
+ "id": 15,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "null",
+ "nullText": null,
+ "options": {},
+ "pluginVersion": "6.4.2",
+ "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(time() - node_boot_time_seconds{datacenter=\"$site\",kubernetes_node=~\"$hostname\"})",
+ "format": "time_series",
+ "hide": false,
+ "instant": true,
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 40
+ }
+ ],
+ "thresholds": "1,2",
+ "title": "System Uptime",
+ "type": "singlestat",
+ "valueFontSize": "70%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "description": "",
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 2,
+ "x": 2,
+ "y": 0
+ },
+ "id": 75,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "70%",
+ "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{datacenter=\"$site\",kubernetes_node=~\"$hostname\"})",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 20
+ }
+ ],
+ "thresholds": "2,3",
+ "title": "Total RAM",
+ "type": "singlestat",
+ "valueFontSize": "70%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "datasource": "Prometheus",
+ "gridPos": {
+ "h": 6,
+ "w": 4,
+ "x": 4,
+ "y": 0
+ },
+ "id": 177,
+ "options": {
+ "displayMode": "lcd",
+ "fieldOptions": {
+ "calcs": [
+ "last"
+ ],
+ "defaults": {
+ "mappings": [],
+ "max": 100,
+ "min": 0,
+ "thresholds": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "#EAB839",
+ "value": 60
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ],
+ "title": "",
+ "unit": "percent"
+ },
+ "override": {},
+ "values": false
+ },
+ "orientation": "horizontal"
+ },
+ "pluginVersion": "6.4.2",
+ "targets": [
+ {
+ "expr": "100 - (avg(irate(node_cpu_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mode=\"idle\"}[30m])) * 100)",
+ "instant": true,
+ "legendFormat": "CPU Busy",
+ "refId": "A"
+ },
+ {
+ "expr": "avg(irate(node_cpu_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mode=\"iowait\"}[30m])) * 100",
+ "hide": true,
+ "instant": true,
+ "legendFormat": "Busy Iowait",
+ "refId": "C"
+ },
+ {
+ "expr": "(1 - (node_memory_MemAvailable_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\"} / (node_memory_MemTotal_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\"})))* 100",
+ "instant": true,
+ "legendFormat": "Memory Usage_{{kubernetes_node}}",
+ "refId": "B"
+ },
+ {
+ "expr": "100 - ((node_filesystem_avail_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mountpoint=\"$maxmount\",fstype=~\"ext4|xfs\"} * 100) / node_filesystem_size_bytes {datacenter=\"$site\",kubernetes_node=~\"$hostname\",mountpoint=\"$maxmount\",fstype=~\"ext4|xfs\"})",
+ "hide": false,
+ "instant": true,
+ "legendFormat": "Disk Usage($maxmount)_{{kubernetes_node}}",
+ "refId": "D"
+ },
+ {
+ "expr": "(1 - (node_memory_SwapFree_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\"} / node_memory_SwapTotal_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\"})) * 100",
+ "hide": true,
+ "instant": true,
+ "legendFormat": "SWAP Usage_{{kubernetes_node}}",
+ "refId": "E"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "",
+ "type": "bargauge"
+ },
+ {
+ "columns": [],
+ "datasource": "Prometheus",
+ "fontSize": "100%",
+ "gridPos": {
+ "h": 6,
+ "w": 8,
+ "x": 8,
+ "y": 0
+ },
+ "id": 164,
+ "links": [],
+ "options": {},
+ "pageSize": null,
+ "scroll": true,
+ "showHeader": true,
+ "sort": {
+ "col": 6,
+ "desc": false
+ },
+ "styles": [
+ {
+ "alias": "Mounted on",
+ "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,
+ "mappingType": 1,
+ "pattern": "mountpoint",
+ "thresholds": [
+ ""
+ ],
+ "type": "string",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Avail",
+ "colorMode": "value",
+ "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,
+ "mappingType": 1,
+ "pattern": "Value #A",
+ "thresholds": [
+ "10000000000",
+ "20000000000"
+ ],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Used",
+ "colorMode": "cell",
+ "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,
+ "mappingType": 1,
+ "pattern": "Value #B",
+ "thresholds": [
+ "0.6",
+ "0.8"
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ },
+ {
+ "alias": "Size",
+ "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": 1,
+ "link": false,
+ "mappingType": 1,
+ "pattern": "Value #C",
+ "thresholds": [],
+ "type": "number",
+ "unit": "bytes"
+ },
+ {
+ "alias": "Filesystem",
+ "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,
+ "link": false,
+ "mappingType": 1,
+ "pattern": "fstype",
+ "thresholds": [],
+ "type": "string",
+ "unit": "short"
+ },
+ {
+ "alias": "IP",
+ "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,
+ "link": false,
+ "mappingType": 1,
+ "pattern": "kubernetes_node",
+ "preserveFormat": false,
+ "sanitize": false,
+ "thresholds": [],
+ "type": "string",
+ "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": "/.*/",
+ "preserveFormat": true,
+ "sanitize": false,
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ }
+ ],
+ "targets": [
+ {
+ "expr": "node_filesystem_size_bytes{datacenter=\"$site\",kubernetes_node=~'$hostname',fstype=~\"ext4|xfs\"}-0",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "C"
+ },
+ {
+ "expr": "node_filesystem_avail_bytes {datacenter=\"$site\",kubernetes_node=~'$hostname',fstype=~\"ext4|xfs\"}-0",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ },
+ {
+ "expr": "1-(node_filesystem_free_bytes{datacenter=\"$site\",kubernetes_node=~'$hostname',fstype=~\"ext4|xfs\"} / node_filesystem_size_bytes{datacenter=\"$site\",kubernetes_node=~'$hostname',fstype=~\"ext4|xfs\"})",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "B"
+ }
+ ],
+ "title": "Disk Space Used Basic(EXT4/XFS)",
+ "transform": "table",
+ "type": "table"
+ },
+ {
+ "aliasColors": {
+ "filefd_192.168.200.241:9100": "super-light-green",
+ "switches_192.168.200.241:9100": "semi-dark-red"
+ },
+ "bars": false,
+ "cacheTimeout": null,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fill": 0,
+ "fillGradient": 1,
+ "gridPos": {
+ "h": 6,
+ "w": 8,
+ "x": 16,
+ "y": 0
+ },
+ "hideTimeOverride": false,
+ "id": 16,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": false,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pluginVersion": "6.4.2",
+ "pointradius": 1,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/filefd_.*/",
+ "lines": false,
+ "pointradius": 1,
+ "points": true
+ },
+ {
+ "alias": "/switches_.*/",
+ "color": "#F2495C",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_filefd_allocated{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 5,
+ "legendFormat": "filefd_{{kubernetes_node}}",
+ "refId": "B"
+ },
+ {
+ "expr": "irate(node_context_switches_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m])",
+ "intervalFactor": 5,
+ "legendFormat": "switches_{{kubernetes_node}}",
+ "refId": "A"
+ },
+ {
+ "expr": "node_filefd_maximum{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "hide": true,
+ "refId": "C"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Open File Descriptor(left)/Context switches(right)",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": "context_switches",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorPostfix": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "Prometheus",
+ "description": "",
+ "format": "short",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 2,
+ "x": 0,
+ "y": 3
+ },
+ "id": 14,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(count(node_cpu_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mode='system'}) by (cpu))",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 20
+ }
+ ],
+ "thresholds": "1,2",
+ "title": "CPU Cores",
+ "type": "singlestat",
+ "valueFontSize": "70%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "description": "",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 2,
+ "x": 2,
+ "y": 3
+ },
+ "id": 20,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "pluginVersion": "6.4.2",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "#3274D9",
+ "show": true,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "avg(irate(node_cpu_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mode=\"iowait\"}[30m])) * 100",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 20
+ }
+ ],
+ "thresholds": "20,50",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU IOwait",
+ "type": "singlestat",
+ "valueFontSize": "70%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "aliasColors": {
+ "15分钟": "#6ED0E0",
+ "1分钟": "#BF1B00",
+ "5分钟": "#CCA300"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 1,
+ "grid": {},
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 0,
+ "y": 6
+ },
+ "height": "300",
+ "id": 13,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "maxPerRow": 6,
+ "nullPointMode": "null as zero",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_load1{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_1m",
+ "metric": "",
+ "refId": "A",
+ "step": 20,
+ "target": ""
+ },
+ {
+ "expr": "node_load5{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_5m",
+ "refId": "B",
+ "step": 20
+ },
+ {
+ "expr": "node_load15{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_15m",
+ "refId": "C",
+ "step": 20
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "System Load",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "192.168.200.241:9100_Total": "dark-red",
+ "Idle - Waiting for something to happen": "#052B51",
+ "guest": "#9AC48A",
+ "idle": "#052B51",
+ "iowait": "#EAB839",
+ "irq": "#BF1B00",
+ "nice": "#C15C17",
+ "sdb_每秒I/O操作%": "#d683ce",
+ "softirq": "#E24D42",
+ "steal": "#FCE2DE",
+ "system": "#508642",
+ "user": "#5195CE",
+ "磁盘花费在I/O操作占比": "#ba43a9"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "description": "",
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 8,
+ "y": 6
+ },
+ "id": 7,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*_Total/",
+ "color": "#C4162A",
+ "fill": 0
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "avg(irate(node_cpu_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mode=\"system\"}[30m])) by (kubernetes_node)",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_System",
+ "refId": "A",
+ "step": 20
+ },
+ {
+ "expr": "avg(irate(node_cpu_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mode=\"user\"}[30m])) by (kubernetes_node)",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_User",
+ "refId": "B",
+ "step": 240
+ },
+ {
+ "expr": "avg(irate(node_cpu_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mode=\"iowait\"}[30m])) by (kubernetes_node)",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_Iowait",
+ "refId": "D",
+ "step": 240
+ },
+ {
+ "expr": "1 - avg(irate(node_cpu_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",mode=\"idle\"}[30m])) by (kubernetes_node)",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_Total",
+ "refId": "F",
+ "step": 240
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "CPU Basic",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "decimals": 2,
+ "format": "percentunit",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "192.168.10.227:9100_em1_in下载": "super-light-green",
+ "192.168.10.227:9100_em1_out上传": "dark-blue"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "fill": 1,
+ "fillGradient": 3,
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 16,
+ "y": 6
+ },
+ "height": "300",
+ "id": 157,
+ "legend": {
+ "alignAsTable": true,
+ "avg": false,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/.*_transmit$/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(node_network_receive_bytes_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",device!~'tap.*|veth.*|br.*|docker.*|virbr*|lo*'}[30m])*8",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_receive",
+ "refId": "A",
+ "step": 4
+ },
+ {
+ "expr": "irate(node_network_transmit_bytes_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\",device!~'tap.*|veth.*|br.*|docker.*|virbr*|lo*'}[30m])*8",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_transmit",
+ "refId": "B",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Network Traffic Basic",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bps",
+ "label": "transmit（-）/receive（+）",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "fill": 1,
+ "fillGradient": 3,
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 0,
+ "y": 14
+ },
+ "id": 174,
+ "legend": {
+ "alignAsTable": true,
+ "avg": false,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/Inodes.*/",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "1-(node_filesystem_free_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\",fstype=~\"ext4|xfs\"} / node_filesystem_size_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\",fstype=~\"ext4|xfs\"})",
+ "format": "time_series",
+ "instant": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}：{{mountpoint}}",
+ "refId": "A"
+ },
+ {
+ "expr": "node_filesystem_files_free{datacenter=\"$site\",kubernetes_node=~\"$hostname\",fstype=~\"ext4|xfs\"} / node_filesystem_files{datacenter=\"$site\",kubernetes_node=~\"$hostname\",fstype=~\"ext4|xfs\"}",
+ "hide": true,
+ "legendFormat": "Inodes：{{instance}}：{{mountpoint}}",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Disk Space Used Basic",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "decimals": 2,
+ "format": "percentunit",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "decimals": 2,
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": "1",
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "192.168.200.241:9100_总内存": "dark-red",
+ "内存_Avaliable": "#6ED0E0",
+ "内存_Cached": "#EF843C",
+ "内存_Free": "#629E51",
+ "内存_Total": "#6d1f62",
+ "内存_Used": "#eab839",
+ "可用": "#9ac48a",
+ "总内存": "#bf1b00"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 8,
+ "y": 14
+ },
+ "height": "300",
+ "id": 156,
+ "legend": {
+ "alignAsTable": true,
+ "avg": false,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/.*_Total/",
+ "fill": 0
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_memory_MemTotal_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_Total",
+ "refId": "A",
+ "step": 4
+ },
+ {
+ "expr": "node_memory_MemTotal_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\"} - node_memory_MemAvailable_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_Used",
+ "refId": "B",
+ "step": 4
+ },
+ {
+ "expr": "node_memory_MemAvailable_bytes{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_Avaliable",
+ "refId": "F",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Memory Basic",
+ "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": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "Idle - Waiting for something to happen": "#052B51",
+ "guest": "#9AC48A",
+ "idle": "#052B51",
+ "iowait": "#EAB839",
+ "irq": "#BF1B00",
+ "nice": "#C15C17",
+ "sdb_每秒I/O操作%": "#d683ce",
+ "softirq": "#E24D42",
+ "steal": "#FCE2DE",
+ "system": "#508642",
+ "user": "#5195CE",
+ "磁盘花费在I/O操作占比": "#ba43a9"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": null,
+ "description": "The time spent on I/O in the natural time of each second.（wall-clock time）",
+ "fill": 1,
+ "fillGradient": 5,
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 16,
+ "y": 14
+ },
+ "id": 175,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "sort": null,
+ "sortDesc": null,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(node_disk_io_time_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_ IO time",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Time Spent Doing I/Os",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "decimals": null,
+ "format": "s",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "vda_write": "#6ED0E0"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "Read/write completions per second",
+ "fill": 1,
+ "fillGradient": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 8,
+ "x": 0,
+ "y": 22
+ },
+ "height": "300",
+ "id": 161,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/.*_Reads completed$/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(node_disk_reads_completed_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m])",
+ "format": "time_series",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_Reads completed",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "irate(node_disk_writes_completed_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m])",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_Writes completed",
+ "refId": "B",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Disk IOps Completed",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "decimals": null,
+ "format": "iops",
+ "label": "IO read (-) / write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "vda_write": "#6ED0E0"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "Per second read / write bytes ",
+ "fill": 1,
+ "fillGradient": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 8,
+ "x": 8,
+ "y": 22
+ },
+ "height": "300",
+ "id": 168,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/.*_Read bytes$/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(node_disk_read_bytes_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m])",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_Read bytes",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "irate(node_disk_written_bytes_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m])",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_Written bytes",
+ "refId": "B",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Disk R/W Data",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "decimals": null,
+ "format": "Bps",
+ "label": "Bytes read (-) / write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "vda": "#6ED0E0"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "Time spent on each read/write operation",
+ "fill": 1,
+ "fillGradient": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 8,
+ "x": 16,
+ "y": 22
+ },
+ "height": "300",
+ "id": 160,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null as zero",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/,*_Read time$/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(node_disk_read_time_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m]) / irate(node_disk_reads_completed_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m])",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_Read time",
+ "refId": "B"
+ },
+ {
+ "expr": "irate(node_disk_write_time_seconds_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m]) / irate(node_disk_writes_completed_total{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}[30m])",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_{{device}}_Write time",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Disk R/W Time(Reference: less than 100ms)(beta)",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": "Time. read (-) / write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "TCP": "#6ED0E0"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "TCP_alloc - Allocated sockets\n\nCurrEstab - TCP connections for which the current state is either ESTABLISHED or CLOSE- WAIT\n\nTCP_tw - Sockets wating close\n\nUDP_inuse - Udp sockets currently in use\n\nSockets_used - Sockets currently in use",
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 12,
+ "w": 12,
+ "x": 0,
+ "y": 31
+ },
+ "height": "300",
+ "id": 158,
+ "interval": "",
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/.*_Sockets_used/",
+ "color": "#C4162A",
+ "fill": 0
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_netstat_Tcp_CurrEstab{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_CurrEstab",
+ "refId": "A",
+ "step": 20
+ },
+ {
+ "expr": "node_sockstat_TCP_tw{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{kubernetes_node}}_TCP_tw",
+ "refId": "D"
+ },
+ {
+ "expr": "node_sockstat_sockets_used{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "legendFormat": "{{kubernetes_node}}_Sockets_used",
+ "refId": "B"
+ },
+ {
+ "expr": "node_sockstat_UDP_inuse{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "legendFormat": "{{kubernetes_node}}_UDP_inuse",
+ "refId": "C"
+ },
+ {
+ "expr": "node_sockstat_TCP_alloc{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "legendFormat": "{{kubernetes_node}}_TCP_alloc",
+ "refId": "E"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Network Sockstat",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fill": 0,
+ "fillGradient": 1,
+ "gridPos": {
+ "h": 12,
+ "w": 12,
+ "x": 12,
+ "y": 31
+ },
+ "id": 169,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": true,
+ "hideZero": true,
+ "max": true,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_hwmon_temp_celsius{datacenter=\"$site\",kubernetes_node=~\"$hostname\"}",
+ "format": "time_series",
+ "intervalFactor": 10,
+ "legendFormat": "{{kubernetes_node}}_{{chip}}_{{sensor}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Hardware Temperature(VM may not display the metrics)",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "celsius",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "refresh": false,
+ "schemaVersion": 20,
+ "style": "dark",
+ "tags": [
+ "aether"
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "datasource": "Prometheus",
+ "definition": "label_values(node_uname_info, datacenter)",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Site",
+ "multi": false,
+ "name": "site",
+ "query": "label_values(node_uname_info, datacenter)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "All",
+ "value": "$__all"
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(node_uname_info{datacenter=\"$site\",job=~\"kubernetes-service-endpoints\"}, kubernetes_node)",
+ "hide": 0,
+ "includeAll": true,
+ "label": "Host",
+ "multi": true,
+ "name": "hostname",
+ "options": [],
+ "query": "label_values(node_uname_info{datacenter=\"$site\",job=~\"kubernetes-service-endpoints\"}, kubernetes_node)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "/",
+ "value": "/"
+ },
+ "datasource": "Prometheus",
+ "definition": "query_result(topk(1,sort_desc (max(node_filesystem_size_bytes{kubernetes_node=~'$hostname',fstype=~\"ext4|xfs\"}) by (mountpoint))))",
+ "hide": 2,
+ "includeAll": false,
+ "label": "",
+ "multi": false,
+ "name": "maxmount",
+ "options": [],
+ "query": "query_result(topk(1,sort_desc (max(node_filesystem_size_bytes{kubernetes_node=~'$hostname',fstype=~\"ext4|xfs\"}) by (mountpoint))))",
+ "refresh": 2,
+ "regex": "/.*\\\"(.*)\\\".*/",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allFormat": "glob",
+ "allValue": null,
+ "current": {
+ "isNone": true,
+ "text": "None",
+ "value": ""
+ },
+ "datasource": "Prometheus",
+ "definition": "",
+ "hide": 2,
+ "includeAll": false,
+ "label": "环境",
+ "multi": false,
+ "multiFormat": "regex values",
+ "name": "env",
+ "options": [],
+ "query": "label_values(node_exporter_build_info,env)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allFormat": "glob",
+ "allValue": "",
+ "current": {
+ "isNone": true,
+ "text": "None",
+ "value": ""
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(node_exporter_build_info{env=~'$env'},name)",
+ "hide": 2,
+ "includeAll": false,
+ "label": "名称",
+ "multi": true,
+ "multiFormat": "regex values",
+ "name": "name",
+ "options": [],
+ "query": "label_values(node_exporter_build_info{env=~'$env'},name)",
+ "refresh": 2,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 1,
+ "tagValuesQuery": "/.*/",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-24h",
+ "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": "browser",
+ "title": "Multi-Site System Monitoring",
+ "uid": "hb7fSE0Zz",
+ "version": 11
+}
diff --git a/nem-monitoring/grafana-dashboards/aether-edges.json b/nem-monitoring/grafana-dashboards/omec-control-plane.json
similarity index 73%
rename from nem-monitoring/grafana-dashboards/aether-edges.json
rename to nem-monitoring/grafana-dashboards/omec-control-plane.json
index 0deddf7..9111bd4 100644
--- a/nem-monitoring/grafana-dashboards/aether-edges.json
+++ b/nem-monitoring/grafana-dashboards/omec-control-plane.json
@@ -15,7 +15,8 @@
"editable": true,
"gnetId": null,
"graphTooltip": 0,
- "id": 3,
+ "id": 1,
+ "iteration": 1580635125090,
"links": [],
"panels": [
{
@@ -27,321 +28,10 @@
"x": 0,
"y": 0
},
- "id": 4,
- "panels": [],
- "repeat": "pod",
- "scopedVars": {
- "pod": {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- }
- },
- "title": "$pod",
- "type": "row"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 0,
- "y": 1
- },
- "id": 16,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "scopedVars": {
- "pod": {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- }
- },
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": true,
- "targets": [
- {
- "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\", container_name!=\"POD\", pod_name=\"$pod-0\"}[5m])) by (datacenter)",
- "interval": "10s",
- "legendFormat": "{{ datacenter }}",
- "refId": "A"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "$pod CPU usage in all Edges",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "decimals": 2,
- "format": "percentunit",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 8,
- "y": 1
- },
- "id": 17,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "scopedVars": {
- "pod": {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- }
- },
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(container_memory_working_set_bytes{container_name!=\"POD\", container_name!=\"\", pod_name=\"$pod-0\"}) by (datacenter)",
- "interval": "10s",
- "legendFormat": "{{ datacenter }}",
- "refId": "A"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "$pod memory usage in all Edges",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "decimals": 2,
- "format": "bytes",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 16,
- "y": 1
- },
- "id": 18,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "scopedVars": {
- "pod": {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- }
- },
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(container_network_receive_bytes_total{pod_name=\"$pod-0\",container_name!=\"\"}[5m]))by (datacenter)",
- "interval": "10s",
- "legendFormat": "{{ datacenter }} rx",
- "refId": "A"
- },
- {
- "expr": "- sum(rate(container_network_transmit_bytes_total{pod_name=\"$pod-0\",container_name!=\"\"}[5m]))by (datacenter)",
- "legendFormat": "{{ datacenter }} tx",
- "refId": "B"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "$pod network I/O in all Edges",
- "tooltip": {
- "shared": true,
- "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": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 27
- },
"id": 25,
"panels": [],
- "repeat": "datacenter",
- "title": "$pod usage in $datacenter",
+ "repeat": "pod",
+ "title": "$pod Resource Usage",
"type": "row"
},
{
@@ -368,7 +58,7 @@
"h": 5,
"w": 6,
"x": 0,
- "y": 28
+ "y": 1
},
"id": 10,
"interval": null,
@@ -411,7 +101,7 @@
"tableColumn": "",
"targets": [
{
- "expr": "100 * sum(irate(container_cpu_usage_seconds_total{pod_name=\"$pod-0\", image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\"}[3m])) by (pod_name) / sum(container_spec_cpu_quota{image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\"} / container_spec_cpu_period{image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\"}) by (pod_name)",
+ "expr": "sum(rate(container_cpu_usage_seconds_total{datacenter=\"$site\", image!=\"\", container_name!=\"POD\", pod_name=~\"$pod.*\"}[5m])) / sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$site\", pod=~\"$pod.*\"}) * 100",
"format": "time_series",
"interval": "10s",
"intervalFactor": 1,
@@ -420,7 +110,7 @@
}
],
"thresholds": "85,95",
- "title": "$pod CPU usage in $datacenter",
+ "title": "CPU usage",
"type": "singlestat",
"valueFontSize": "80%",
"valueMaps": [
@@ -456,7 +146,7 @@
"h": 5,
"w": 6,
"x": 6,
- "y": 28
+ "y": 1
},
"id": 12,
"interval": null,
@@ -499,7 +189,7 @@
"tableColumn": "",
"targets": [
{
- "expr": "100 * (sum(container_memory_usage_bytes{pod_name=\"$pod-0\", container_name!=\"POD\", datacenter=\"$datacenter\"}) / sum(container_spec_memory_limit_bytes{pod_name=\"$pod-0\", container_name!=\"POD\", datacenter=\"$datacenter\"}))",
+ "expr": "100 * (sum(container_memory_usage_bytes{datacenter=\"$site\", pod_name=~\"$pod.*\", container_name!=\"POD\"}) / sum(container_spec_memory_limit_bytes{datacenter=\"$site\", pod_name=~\"$pod.*\", container_name!=\"POD\"}))",
"format": "time_series",
"hide": false,
"instant": false,
@@ -509,7 +199,7 @@
}
],
"thresholds": "85,95",
- "title": "$pod memory usage in $datacenter",
+ "title": "Memory usage",
"type": "singlestat",
"valueFontSize": "80%",
"valueMaps": [
@@ -530,10 +220,10 @@
"fill": 2,
"fillGradient": 0,
"gridPos": {
- "h": 9,
+ "h": 8,
"w": 12,
"x": 12,
- "y": 28
+ "y": 1
},
"id": 2,
"legend": {
@@ -548,14 +238,14 @@
"values": true
},
"lines": true,
- "linewidth": 2,
+ "linewidth": 1,
"links": [],
"nullPointMode": "null",
"options": {
"dataLinks": []
},
"percentage": false,
- "pointradius": 5,
+ "pointradius": 1,
"points": false,
"renderer": "flot",
"repeat": null,
@@ -566,19 +256,19 @@
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(container_network_receive_bytes_total{pod_name=\"$pod-0\", datacenter=\"$datacenter\"}[3m]))by (interface)",
+ "expr": "sum(rate(container_network_receive_bytes_total{pod_name=~\"$pod.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
"format": "time_series",
"interval": "10s",
"intervalFactor": 1,
- "legendFormat": "rx: {{ interface }}",
+ "legendFormat": "rx: {{pod_name}}| {{ interface }}",
"refId": "B"
},
{
- "expr": "-sum(rate(container_network_transmit_bytes_total{pod_name=\"$pod-0\", datacenter=\"$datacenter\"}[3m]))by (interface)",
+ "expr": "-sum(rate(container_network_transmit_bytes_total{pod_name=~\"$pod.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
"format": "time_series",
"interval": "10s",
"intervalFactor": 1,
- "legendFormat": "tx: {{ interface }}",
+ "legendFormat": "tx: {{pod_name}}| {{ interface }}",
"refId": "A"
}
],
@@ -586,7 +276,7 @@
"timeFrom": null,
"timeRegions": [],
"timeShift": null,
- "title": "$pod network I/O in $datacenter",
+ "title": "Network I/O",
"tooltip": {
"shared": true,
"sort": 1,
@@ -642,10 +332,10 @@
"thresholdMarkers": true
},
"gridPos": {
- "h": 4,
+ "h": 3,
"w": 3,
"x": 0,
- "y": 33
+ "y": 6
},
"id": 20,
"interval": null,
@@ -689,7 +379,7 @@
"tableColumn": "",
"targets": [
{
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$datacenter\", pod=\"$pod-0\"})",
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$site\", pod=~\"$pod.*\"})",
"interval": "10s",
"refId": "A"
}
@@ -728,10 +418,10 @@
"thresholdMarkers": true
},
"gridPos": {
- "h": 4,
+ "h": 3,
"w": 3,
"x": 3,
- "y": 33
+ "y": 6
},
"id": 21,
"interval": null,
@@ -775,7 +465,7 @@
"tableColumn": "",
"targets": [
{
- "expr": "sum (irate (container_cpu_usage_seconds_total{image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\", pod_name=\"$pod-0\"}[5m]))",
+ "expr": "sum (rate(container_cpu_usage_seconds_total{datacenter=\"$site\", image!=\"\", container_name!=\"POD\", pod_name=~\"$pod.*\"}[5m]))",
"interval": "10s",
"refId": "A"
}
@@ -815,10 +505,10 @@
"thresholdMarkers": true
},
"gridPos": {
- "h": 4,
+ "h": 3,
"w": 3,
"x": 6,
- "y": 33
+ "y": 6
},
"id": 22,
"interval": null,
@@ -862,7 +552,7 @@
"tableColumn": "",
"targets": [
{
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{datacenter=\"$datacenter\", pod=\"$pod-0\"})",
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{datacenter=\"$site\", pod=~\"$pod.*\"})",
"refId": "A"
}
],
@@ -891,6 +581,7 @@
"#d44a3a"
],
"datasource": null,
+ "decimals": null,
"format": "decbytes",
"gauge": {
"maxValue": 100,
@@ -900,10 +591,10 @@
"thresholdMarkers": true
},
"gridPos": {
- "h": 4,
+ "h": 3,
"w": 3,
"x": 9,
- "y": 33
+ "y": 6
},
"id": 23,
"interval": null,
@@ -947,7 +638,7 @@
"tableColumn": "",
"targets": [
{
- "expr": "sum(container_memory_working_set_bytes{container_name!=\"POD\", container_name!=\"\", datacenter=\"$datacenter\", pod_name=\"$pod-0\"})",
+ "expr": "sum(container_memory_working_set_bytes{datacenter=\"$site\", container_name!=\"POD\", container_name!=\"\", pod_name=~\"$pod.*\"})",
"interval": "10s",
"refId": "A"
}
@@ -966,91 +657,340 @@
}
],
"valueName": "avg"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 3,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 6,
+ "w": 12,
+ "x": 0,
+ "y": 9
+ },
+ "height": "",
+ "id": 27,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod_name=~\"^$pod-.*\",name=~\"^k8s_.*\",container_name!=\"POD\",datacenter=\"$site\"}[5m])) by (container_name, pod_name)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "pod: {{ pod_name }} | {{ container_name }}",
+ "metric": "container_cpu",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod_name=\"$pod-0\",name!~\"^k8s_.*\",datacenter=\"$site\"}[5m])) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "container_cpu",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{rkt_container_name!=\"\",datacenter=\"$site\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "container_cpu",
+ "refId": "C",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers CPU usage (5m avg)",
+ "tooltip": {
+ "msResolution": true,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": "cores",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 6,
+ "w": 12,
+ "x": 12,
+ "y": 9
+ },
+ "id": 29,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",datacenter=\"$site\",pod_name=~\"^$pod.*\",name=~\"^k8s_.*\",container_name!=\"POD\"}) by (container_name, pod_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "pod: {{ pod_name }} | {{ container_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",name!~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum (container_memory_working_set_bytes{rkt_container_name!=\"\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "C",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers memory usage",
+ "tooltip": {
+ "msResolution": false,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
}
],
- "refresh": "5s",
+ "refresh": "1m",
"schemaVersion": 20,
"style": "dark",
"tags": [
- "aether"
+ "aether",
+ "omec"
],
"templating": {
"list": [
{
- "allValue": ".*",
- "current": {
- "text": "spgwu",
- "value": [
- "spgwu"
- ]
- },
- "hide": 0,
- "includeAll": false,
- "label": "Pod",
- "multi": false,
- "name": "pod",
- "options": [
- {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-ric",
- "value": "accelleran-cbrs-ric"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-nats",
- "value": "accelleran-cbrs-nats"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-redis",
- "value": "accelleran-cbrs-redis"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-ricweb",
- "value": "accelleran-cbrs-ricweb"
- }
- ],
- "query": "spgwu,accelleran-cbrs-ric,accelleran-cbrs-nats,accelleran-cbrs-redis,accelleran-cbrs-ricweb",
- "skipUrlSync": false,
- "type": "custom"
- },
- {
- "allValue": ".*",
- "current": {
- "text": "All",
- "value": [
- "$__all"
- ]
- },
+ "allValue": null,
"datasource": "Prometheus",
"definition": "label_values(kube_pod_info, datacenter)",
"hide": 0,
- "includeAll": true,
- "label": "Datacenter",
- "multi": true,
- "name": "datacenter",
- "options": [],
+ "includeAll": false,
+ "label": "Site",
+ "multi": false,
+ "name": "site",
"query": "label_values(kube_pod_info, datacenter)",
"refresh": 2,
- "regex": "/.*edge/",
+ "regex": "/.*central.*/",
"skipUrlSync": false,
- "sort": 1,
+ "sort": 0,
"tagValuesQuery": "",
"tags": [],
"tagsQuery": "",
"type": "query",
"useTags": false
+ },
+ {
+ "allValue": ".*",
+ "current": {
+ "tags": [],
+ "text": "All",
+ "value": [
+ "$__all"
+ ]
+ },
+ "hide": 0,
+ "includeAll": true,
+ "label": "Pod",
+ "multi": true,
+ "name": "pod",
+ "options": [
+ {
+ "selected": false,
+ "text": "All",
+ "value": "$__all"
+ },
+ {
+ "selected": true,
+ "text": "mme",
+ "value": "mme"
+ },
+ {
+ "selected": false,
+ "text": "spgwc",
+ "value": "spgwc"
+ },
+ {
+ "selected": false,
+ "text": "hss",
+ "value": "hss"
+ },
+ {
+ "selected": false,
+ "text": "cassandra",
+ "value": "cassandra"
+ }
+ ],
+ "query": "mme,spgwc,hss,cassandra",
+ "skipUrlSync": false,
+ "type": "custom"
}
]
},
"time": {
- "from": "now-5m",
+ "from": "now-3h",
"to": "now"
},
"timepicker": {
@@ -1080,7 +1020,7 @@
]
},
"timezone": "",
- "title": "Aether Edges",
- "uid": "gKU2e27Za",
- "version": 3
-}
\ No newline at end of file
+ "title": "OMEC Control Plane",
+ "uid": "gKU2e27Zb",
+ "version": 1
+}
diff --git a/nem-monitoring/grafana-dashboards/aether-mme-exporter.json b/nem-monitoring/grafana-dashboards/omec-imsi-list.json
similarity index 98%
rename from nem-monitoring/grafana-dashboards/aether-mme-exporter.json
rename to nem-monitoring/grafana-dashboards/omec-imsi-list.json
index 7d00b4b..ca60507 100644
--- a/nem-monitoring/grafana-dashboards/aether-mme-exporter.json
+++ b/nem-monitoring/grafana-dashboards/omec-imsi-list.json
@@ -251,7 +251,10 @@
],
"schemaVersion": 20,
"style": "dark",
- "tags": [],
+ "tags": [
+ "aether",
+ "omec"
+ ],
"templating": {
"list": []
},
@@ -277,4 +280,4 @@
"title": "IMSI list",
"uid": "pWy5BgyWz",
"version": 5
- }
\ No newline at end of file
+ }
diff --git a/nem-monitoring/grafana-dashboards/aether-edges.json b/nem-monitoring/grafana-dashboards/omec-user-plane-summary.json
similarity index 64%
copy from nem-monitoring/grafana-dashboards/aether-edges.json
copy to nem-monitoring/grafana-dashboards/omec-user-plane-summary.json
index 0deddf7..c386151 100644
--- a/nem-monitoring/grafana-dashboards/aether-edges.json
+++ b/nem-monitoring/grafana-dashboards/omec-user-plane-summary.json
@@ -15,7 +15,8 @@
"editable": true,
"gnetId": null,
"graphTooltip": 0,
- "id": 3,
+ "id": 5,
+ "iteration": 1580754449762,
"links": [],
"panels": [
{
@@ -27,350 +28,37 @@
"x": 0,
"y": 0
},
- "id": 4,
+ "id": 31,
"panels": [],
- "repeat": "pod",
- "scopedVars": {
- "pod": {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- }
- },
- "title": "$pod",
- "type": "row"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 0,
- "y": 1
- },
- "id": 16,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "scopedVars": {
- "pod": {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- }
- },
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": true,
- "targets": [
- {
- "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\", container_name!=\"POD\", pod_name=\"$pod-0\"}[5m])) by (datacenter)",
- "interval": "10s",
- "legendFormat": "{{ datacenter }}",
- "refId": "A"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "$pod CPU usage in all Edges",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "decimals": 2,
- "format": "percentunit",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 8,
- "y": 1
- },
- "id": 17,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "scopedVars": {
- "pod": {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- }
- },
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(container_memory_working_set_bytes{container_name!=\"POD\", container_name!=\"\", pod_name=\"$pod-0\"}) by (datacenter)",
- "interval": "10s",
- "legendFormat": "{{ datacenter }}",
- "refId": "A"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "$pod memory usage in all Edges",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "decimals": 2,
- "format": "bytes",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": null,
- "fill": 1,
- "fillGradient": 0,
- "gridPos": {
- "h": 8,
- "w": 8,
- "x": 16,
- "y": 1
- },
- "id": 18,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "nullPointMode": "null",
- "options": {
- "dataLinks": []
- },
- "percentage": false,
- "pointradius": 2,
- "points": false,
- "renderer": "flot",
- "scopedVars": {
- "pod": {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- }
- },
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(container_network_receive_bytes_total{pod_name=\"$pod-0\",container_name!=\"\"}[5m]))by (datacenter)",
- "interval": "10s",
- "legendFormat": "{{ datacenter }} rx",
- "refId": "A"
- },
- {
- "expr": "- sum(rate(container_network_transmit_bytes_total{pod_name=\"$pod-0\",container_name!=\"\"}[5m]))by (datacenter)",
- "legendFormat": "{{ datacenter }} tx",
- "refId": "B"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeRegions": [],
- "timeShift": null,
- "title": "$pod network I/O in all Edges",
- "tooltip": {
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ],
- "yaxis": {
- "align": false,
- "alignLevel": null
- }
- },
- {
- "collapsed": false,
- "datasource": null,
- "gridPos": {
- "h": 1,
- "w": 24,
- "x": 0,
- "y": 27
- },
- "id": 25,
- "panels": [],
- "repeat": "datacenter",
- "title": "$pod usage in $datacenter",
+ "repeat": "site",
+ "title": "$site",
"type": "row"
},
{
"cacheTimeout": null,
"colorBackground": false,
- "colorValue": true,
+ "colorValue": false,
"colors": [
- "rgba(50, 172, 45, 0.97)",
+ "#299c46",
"rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
+ "#d44a3a"
],
- "datasource": "Prometheus",
- "decimals": null,
- "description": "",
- "format": "percent",
+ "datasource": null,
+ "format": "s",
"gauge": {
"maxValue": 100,
"minValue": 0,
- "show": true,
+ "show": false,
"thresholdLabels": false,
"thresholdMarkers": true
},
"gridPos": {
- "h": 5,
- "w": 6,
+ "h": 4,
+ "w": 4,
"x": 0,
- "y": 28
+ "y": 1
},
- "id": 10,
+ "id": 29,
"interval": null,
"links": [],
"mappingType": 1,
@@ -385,8 +73,7 @@
}
],
"maxDataPoints": 100,
- "maxPerRow": 12,
- "nullPointMode": "null",
+ "nullPointMode": "connected",
"nullText": null,
"options": {},
"postfix": "",
@@ -406,23 +93,23 @@
"fillColor": "rgba(31, 118, 189, 0.18)",
"full": false,
"lineColor": "rgb(31, 120, 193)",
- "show": true
+ "show": false,
+ "ymax": null,
+ "ymin": null
},
"tableColumn": "",
"targets": [
{
- "expr": "100 * sum(irate(container_cpu_usage_seconds_total{pod_name=\"$pod-0\", image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\"}[3m])) by (pod_name) / sum(container_spec_cpu_quota{image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\"} / container_spec_cpu_period{image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\"}) by (pod_name)",
- "format": "time_series",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "",
+ "expr": "sum(time() - kube_pod_created{datacenter=\"$site\",namespace=~\"omec|omec-cbrs\",pod=~\"spgwu.*\"})",
"refId": "A"
}
],
- "thresholds": "85,95",
- "title": "$pod CPU usage in $datacenter",
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Age",
"type": "singlestat",
- "valueFontSize": "80%",
+ "valueFontSize": "100%",
"valueMaps": [
{
"op": "=",
@@ -430,7 +117,62 @@
"value": "null"
}
],
- "valueName": "current"
+ "valueName": "avg"
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "description": "",
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 4,
+ "y": 1
+ },
+ "id": 10,
+ "links": [],
+ "options": {
+ "fieldOptions": {
+ "calcs": [
+ "mean"
+ ],
+ "defaults": {
+ "mappings": [],
+ "max": 100,
+ "min": 0,
+ "thresholds": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "override": {},
+ "values": false
+ },
+ "orientation": "horizontal",
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true
+ },
+ "pluginVersion": "6.4.2",
+ "repeat": null,
+ "repeatDirection": "v",
+ "targets": [
+ {
+ "expr": "sum(rate(container_cpu_usage_seconds_total{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", image!=\"\", container_name!=\"POD\", pod_name=~\"spgwu.*\"}[5m])) / sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod=~\"spgwu.*\"}) * 100",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "SPGWU CPU Utilization",
+ "type": "gauge"
},
{
"cacheTimeout": null,
@@ -453,10 +195,10 @@
"thresholdMarkers": true
},
"gridPos": {
- "h": 5,
- "w": 6,
- "x": 6,
- "y": 28
+ "h": 8,
+ "w": 4,
+ "x": 8,
+ "y": 1
},
"id": 12,
"interval": null,
@@ -499,7 +241,7 @@
"tableColumn": "",
"targets": [
{
- "expr": "100 * (sum(container_memory_usage_bytes{pod_name=\"$pod-0\", container_name!=\"POD\", datacenter=\"$datacenter\"}) / sum(container_spec_memory_limit_bytes{pod_name=\"$pod-0\", container_name!=\"POD\", datacenter=\"$datacenter\"}))",
+ "expr": "100 * (sum(container_memory_usage_bytes{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod_name=~\"spgwu.*\", container_name!=\"POD\"}) / sum(container_spec_memory_limit_bytes{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod_name=~\"spgwu.*\", container_name!=\"POD\"}))",
"format": "time_series",
"hide": false,
"instant": false,
@@ -509,7 +251,7 @@
}
],
"thresholds": "85,95",
- "title": "$pod memory usage in $datacenter",
+ "title": "SPGWU Memory Utilization",
"type": "singlestat",
"valueFontSize": "80%",
"valueMaps": [
@@ -530,10 +272,10 @@
"fill": 2,
"fillGradient": 0,
"gridPos": {
- "h": 9,
+ "h": 8,
"w": 12,
"x": 12,
- "y": 28
+ "y": 1
},
"id": 2,
"legend": {
@@ -548,14 +290,14 @@
"values": true
},
"lines": true,
- "linewidth": 2,
+ "linewidth": 1,
"links": [],
"nullPointMode": "null",
"options": {
"dataLinks": []
},
"percentage": false,
- "pointradius": 5,
+ "pointradius": 1,
"points": false,
"renderer": "flot",
"repeat": null,
@@ -566,7 +308,7 @@
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(container_network_receive_bytes_total{pod_name=\"$pod-0\", datacenter=\"$datacenter\"}[3m]))by (interface)",
+ "expr": "sum(rate(container_network_receive_bytes_total{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod_name=~\"spgwu.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
"format": "time_series",
"interval": "10s",
"intervalFactor": 1,
@@ -574,7 +316,7 @@
"refId": "B"
},
{
- "expr": "-sum(rate(container_network_transmit_bytes_total{pod_name=\"$pod-0\", datacenter=\"$datacenter\"}[3m]))by (interface)",
+ "expr": "-sum(rate(container_network_transmit_bytes_total{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod_name=~\"spgwu.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
"format": "time_series",
"interval": "10s",
"intervalFactor": 1,
@@ -586,7 +328,7 @@
"timeFrom": null,
"timeRegions": [],
"timeShift": null,
- "title": "$pod network I/O in $datacenter",
+ "title": "SPGWU Network I/O",
"tooltip": {
"shared": true,
"sort": 1,
@@ -633,7 +375,7 @@
"#d44a3a"
],
"datasource": null,
- "format": "none",
+ "format": "s",
"gauge": {
"maxValue": 100,
"minValue": 0,
@@ -643,11 +385,11 @@
},
"gridPos": {
"h": 4,
- "w": 3,
+ "w": 4,
"x": 0,
- "y": 33
+ "y": 5
},
- "id": 20,
+ "id": 32,
"interval": null,
"links": [],
"mappingType": 1,
@@ -676,7 +418,6 @@
"to": "null"
}
],
- "repeat": null,
"repeatDirection": "v",
"sparkline": {
"fillColor": "rgba(31, 118, 189, 0.18)",
@@ -689,17 +430,16 @@
"tableColumn": "",
"targets": [
{
- "expr": "sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$datacenter\", pod=\"$pod-0\"})",
- "interval": "10s",
+ "expr": "count(ue_info{datacenter=\"$site\"})",
"refId": "A"
}
],
"thresholds": "",
"timeFrom": null,
"timeShift": null,
- "title": "CPU limits",
+ "title": "Number of Subscribers",
"type": "singlestat",
- "valueFontSize": "80%",
+ "valueFontSize": "100%",
"valueMaps": [
{
"op": "=",
@@ -711,7 +451,8 @@
},
{
"cacheTimeout": null,
- "colorBackground": false,
+ "colorBackground": true,
+ "colorPrefix": false,
"colorValue": false,
"colors": [
"#299c46",
@@ -729,11 +470,11 @@
},
"gridPos": {
"h": 4,
- "w": 3,
- "x": 3,
- "y": 33
+ "w": 4,
+ "x": 0,
+ "y": 9
},
- "id": 21,
+ "id": 36,
"interval": null,
"links": [],
"mappingType": 1,
@@ -762,7 +503,6 @@
"to": "null"
}
],
- "repeat": null,
"repeatDirection": "v",
"sparkline": {
"fillColor": "rgba(31, 118, 189, 0.18)",
@@ -775,17 +515,16 @@
"tableColumn": "",
"targets": [
{
- "expr": "sum (irate (container_cpu_usage_seconds_total{image!=\"\", container_name!=\"POD\", datacenter=\"$datacenter\", pod_name=\"$pod-0\"}[5m]))",
- "interval": "10s",
+ "expr": "kube_pod_container_status_restarts_total{datacenter=\"$site\",namespace=~\"omec|omec-cbrs\",pod=~\"spgwu.*\"}",
"refId": "A"
}
],
- "thresholds": "",
+ "thresholds": "3,10",
"timeFrom": null,
"timeShift": null,
- "title": "CPU used (5m avg)",
+ "title": "SPGWU Restarts Total",
"type": "singlestat",
- "valueFontSize": "80%",
+ "valueFontSize": "110%",
"valueMaps": [
{
"op": "=",
@@ -797,30 +536,85 @@
},
{
"cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "#299c46",
- "rgba(237, 129, 40, 0.89)",
- "#d44a3a"
+ "datasource": "Prometheus",
+ "description": "",
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 4,
+ "y": 9
+ },
+ "id": 33,
+ "links": [],
+ "options": {
+ "fieldOptions": {
+ "calcs": [
+ "mean"
+ ],
+ "defaults": {
+ "mappings": [],
+ "max": 100,
+ "min": 0,
+ "thresholds": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "override": {},
+ "values": false
+ },
+ "orientation": "horizontal",
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true
+ },
+ "pluginVersion": "6.4.2",
+ "repeatDirection": "v",
+ "targets": [
+ {
+ "expr": "sum(rate(container_cpu_usage_seconds_total{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", image!=\"\", container_name!=\"POD\", pod_name=~\"accelleran-cbrs-cu.*\"}[5m])) / sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod=~\"accelleran-cbrs-cu.*\"}) * 100",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
],
- "datasource": null,
- "decimals": 2,
- "format": "decbytes",
+ "title": "Accelleran-CU CPU Utilization",
+ "type": "gauge"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": null,
+ "description": "",
+ "format": "percent",
"gauge": {
"maxValue": 100,
"minValue": 0,
- "show": false,
+ "show": true,
"thresholdLabels": false,
"thresholdMarkers": true
},
"gridPos": {
- "h": 4,
- "w": 3,
- "x": 6,
- "y": 33
+ "h": 8,
+ "w": 4,
+ "x": 8,
+ "y": 9
},
- "id": 22,
+ "id": 34,
"interval": null,
"links": [],
"mappingType": 1,
@@ -835,7 +629,8 @@
}
],
"maxDataPoints": 100,
- "nullPointMode": "connected",
+ "maxPerRow": 12,
+ "nullPointMode": "null",
"nullText": null,
"options": {},
"postfix": "",
@@ -849,27 +644,27 @@
"to": "null"
}
],
- "repeat": null,
"repeatDirection": "v",
"sparkline": {
"fillColor": "rgba(31, 118, 189, 0.18)",
"full": false,
"lineColor": "rgb(31, 120, 193)",
- "show": false,
- "ymax": null,
- "ymin": null
+ "show": true
},
"tableColumn": "",
"targets": [
{
- "expr": "sum(kube_pod_container_resource_limits_memory_bytes{datacenter=\"$datacenter\", pod=\"$pod-0\"})",
+ "expr": "100 * (sum(container_memory_usage_bytes{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod_name=~\"accelleran-cbrs-cu.*\", container_name!=\"POD\"}) / sum(container_spec_memory_limit_bytes{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod_name=~\"accelleran-cbrs-cu.*\", container_name!=\"POD\"}))",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "10s",
+ "intervalFactor": 1,
"refId": "A"
}
],
- "thresholds": "",
- "timeFrom": null,
- "timeShift": null,
- "title": "Memory limits",
+ "thresholds": "85,95",
+ "title": "Accelleran-CU Memory Utilization",
"type": "singlestat",
"valueFontSize": "80%",
"valueMaps": [
@@ -879,11 +674,113 @@
"value": "null"
}
],
- "valueName": "avg"
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 2,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 12,
+ "y": 9
+ },
+ "id": 35,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 1,
+ "points": false,
+ "renderer": "flot",
+ "repeatDirection": "v",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(container_network_receive_bytes_total{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod_name=~\"accelleran-cbrs-cu.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "rx: {{ interface }}",
+ "refId": "B"
+ },
+ {
+ "expr": "-sum(rate(container_network_transmit_bytes_total{datacenter=\"$site\", namespace=~\"omec|omec-cbrs\", pod_name=~\"accelleran-cbrs-cu.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "tx: {{ interface }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Accelleran-CU Network I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 1,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
},
{
"cacheTimeout": null,
- "colorBackground": false,
+ "colorBackground": true,
+ "colorPrefix": false,
"colorValue": false,
"colors": [
"#299c46",
@@ -891,7 +788,7 @@
"#d44a3a"
],
"datasource": null,
- "format": "decbytes",
+ "format": "none",
"gauge": {
"maxValue": 100,
"minValue": 0,
@@ -901,11 +798,11 @@
},
"gridPos": {
"h": 4,
- "w": 3,
- "x": 9,
- "y": 33
+ "w": 4,
+ "x": 0,
+ "y": 13
},
- "id": 23,
+ "id": 37,
"interval": null,
"links": [],
"mappingType": 1,
@@ -934,7 +831,6 @@
"to": "null"
}
],
- "repeat": null,
"repeatDirection": "v",
"sparkline": {
"fillColor": "rgba(31, 118, 189, 0.18)",
@@ -947,17 +843,16 @@
"tableColumn": "",
"targets": [
{
- "expr": "sum(container_memory_working_set_bytes{container_name!=\"POD\", container_name!=\"\", datacenter=\"$datacenter\", pod_name=\"$pod-0\"})",
- "interval": "10s",
+ "expr": "kube_pod_container_status_restarts_total{datacenter=\"$site\",namespace=~\"omec|omec-cbrs\",pod=~\"accelleran-cbrs-cu.*\"}",
"refId": "A"
}
],
- "thresholds": "",
+ "thresholds": "3,10",
"timeFrom": null,
"timeShift": null,
- "title": "Memory used",
+ "title": "Accelleran-CU Restarts Total",
"type": "singlestat",
- "valueFontSize": "80%",
+ "valueFontSize": "110%",
"valueMaps": [
{
"op": "=",
@@ -968,61 +863,19 @@
"valueName": "avg"
}
],
- "refresh": "5s",
+ "refresh": "1m",
"schemaVersion": 20,
"style": "dark",
"tags": [
- "aether"
+ "aether",
+ "omec"
],
"templating": {
"list": [
{
- "allValue": ".*",
+ "allValue": null,
"current": {
- "text": "spgwu",
- "value": [
- "spgwu"
- ]
- },
- "hide": 0,
- "includeAll": false,
- "label": "Pod",
- "multi": false,
- "name": "pod",
- "options": [
- {
- "selected": true,
- "text": "spgwu",
- "value": "spgwu"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-ric",
- "value": "accelleran-cbrs-ric"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-nats",
- "value": "accelleran-cbrs-nats"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-redis",
- "value": "accelleran-cbrs-redis"
- },
- {
- "selected": false,
- "text": "accelleran-cbrs-ricweb",
- "value": "accelleran-cbrs-ricweb"
- }
- ],
- "query": "spgwu,accelleran-cbrs-ric,accelleran-cbrs-nats,accelleran-cbrs-redis,accelleran-cbrs-ricweb",
- "skipUrlSync": false,
- "type": "custom"
- },
- {
- "allValue": ".*",
- "current": {
+ "tags": [],
"text": "All",
"value": [
"$__all"
@@ -1032,15 +885,15 @@
"definition": "label_values(kube_pod_info, datacenter)",
"hide": 0,
"includeAll": true,
- "label": "Datacenter",
+ "label": "Site",
"multi": true,
- "name": "datacenter",
+ "name": "site",
"options": [],
"query": "label_values(kube_pod_info, datacenter)",
"refresh": 2,
- "regex": "/.*edge/",
+ "regex": "/.*edge.*/",
"skipUrlSync": false,
- "sort": 1,
+ "sort": 0,
"tagValuesQuery": "",
"tags": [],
"tagsQuery": "",
@@ -1050,7 +903,7 @@
]
},
"time": {
- "from": "now-5m",
+ "from": "now-3h",
"to": "now"
},
"timepicker": {
@@ -1080,7 +933,7 @@
]
},
"timezone": "",
- "title": "Aether Edges",
- "uid": "gKU2e27Za",
- "version": 3
-}
\ No newline at end of file
+ "title": "OMEC User Plane Summary",
+ "uid": "jccoF4UWz",
+ "version": 1
+}
diff --git a/nem-monitoring/grafana-dashboards/omec-user-plane.json b/nem-monitoring/grafana-dashboards/omec-user-plane.json
new file mode 100644
index 0000000..02c73c9
--- /dev/null
+++ b/nem-monitoring/grafana-dashboards/omec-user-plane.json
@@ -0,0 +1,2098 @@
+{
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "type": "dashboard"
+ }
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "id": 10,
+ "iteration": 1580642657156,
+ "links": [],
+ "panels": [
+ {
+ "collapsed": false,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 25,
+ "panels": [],
+ "repeat": "pod",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "title": "$namespace:$pod Resource Usage",
+ "type": "row"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": null,
+ "description": "",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 6,
+ "x": 0,
+ "y": 1
+ },
+ "id": 10,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 12,
+ "nullPointMode": "null",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(rate(container_cpu_usage_seconds_total{datacenter=\"$site\", namespace=\"$namespace\", image!=\"\", container_name!=\"POD\", pod_name=~\"$pod.*\"}[5m])) / sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$site\", namespace=\"$namespace\", pod=~\"$pod.*\"}) * 100",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "85,95",
+ "title": "CPU usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": null,
+ "description": "",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 6,
+ "x": 6,
+ "y": 1
+ },
+ "id": 12,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 12,
+ "nullPointMode": "null",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "100 * (sum(container_memory_usage_bytes{datacenter=\"$site\", namespace=\"$namespace\", pod_name=~\"$pod.*\", container_name!=\"POD\"}) / sum(container_spec_memory_limit_bytes{datacenter=\"$site\", namespace=\"$namespace\", pod_name=~\"$pod.*\", container_name!=\"POD\"}))",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A"
+ }
+ ],
+ "thresholds": "85,95",
+ "title": "Memory usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 2,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 12,
+ "y": 1
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 1,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "repeatDirection": "v",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(container_network_receive_bytes_total{datacenter=\"$site\", namespace=\"$namespace\", pod_name=~\"$pod.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "rx: {{pod_name}}| {{ interface }}",
+ "refId": "B"
+ },
+ {
+ "expr": "-sum(rate(container_network_transmit_bytes_total{datacenter=\"$site\", namespace=\"$namespace\", pod_name=~\"$pod.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "tx: {{pod_name}}| {{ interface }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Network I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 1,
+ "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": null,
+ "show": true
+ },
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": null,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 0,
+ "y": 6
+ },
+ "id": 20,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$site\", namespace=\"$namespace\", pod=~\"$pod.*\"})",
+ "interval": "10s",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU limits",
+ "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": null,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 3,
+ "y": 6
+ },
+ "id": 21,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum (rate(container_cpu_usage_seconds_total{datacenter=\"$site\", namespace=\"$namespace\", image!=\"\", container_name!=\"POD\", pod_name=~\"$pod.*\"}[5m]))",
+ "interval": "10s",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU used (5m avg)",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": null,
+ "decimals": 2,
+ "format": "decbytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 6,
+ "y": 6
+ },
+ "id": 22,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{datacenter=\"$site\", namespace=\"$namespace\", pod=~\"$pod.*\"})",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory limits",
+ "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": null,
+ "decimals": null,
+ "format": "decbytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 9,
+ "y": 6
+ },
+ "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,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(container_memory_working_set_bytes{datacenter=\"$site\", namespace=\"$namespace\", container_name!=\"POD\", container_name!=\"\", pod_name=~\"$pod.*\"})",
+ "interval": "10s",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory used",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 3,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 6,
+ "w": 12,
+ "x": 0,
+ "y": 9
+ },
+ "height": "",
+ "id": 27,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",datacenter=\"$site\", namespace=\"$namespace\",pod_name=~\"^$pod-.*\",name=~\"^k8s_.*\",container_name!=\"POD\"}[5m])) by (container_name, pod_name)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "{{ pod_name }} | {{ container_name }}",
+ "metric": "container_cpu",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod_name=\"$pod-0\",name!~\"^k8s_.*\",datacenter=\"$site\"}[5m])) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "container_cpu",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{rkt_container_name!=\"\",datacenter=\"$site\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "container_cpu",
+ "refId": "C",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers CPU usage (5m avg)",
+ "tooltip": {
+ "msResolution": true,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": "cores",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 6,
+ "w": 12,
+ "x": 12,
+ "y": 9
+ },
+ "id": 29,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ }
+ },
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",datacenter=\"$site\", namespace=\"$namespace\",pod_name=~\"^$pod.*\",name=~\"^k8s_.*\",container_name!=\"POD\"}) by (container_name, pod_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "{{ pod_name }} | {{ container_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",name!~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum (container_memory_working_set_bytes{rkt_container_name!=\"\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "C",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers memory usage",
+ "tooltip": {
+ "msResolution": false,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "collapsed": false,
+ "datasource": null,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 15
+ },
+ "id": 30,
+ "panels": [],
+ "repeat": null,
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 25,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "title": "$namespace:$pod Resource Usage",
+ "type": "row"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": null,
+ "description": "",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 6,
+ "x": 0,
+ "y": 16
+ },
+ "id": 31,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 12,
+ "nullPointMode": "null",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 10,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(rate(container_cpu_usage_seconds_total{datacenter=\"$site\", namespace=\"$namespace\", image!=\"\", container_name!=\"POD\", pod_name=~\"$pod.*\"}[5m])) / sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$site\", namespace=\"$namespace\", pod=~\"$pod.*\"}) * 100",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "85,95",
+ "title": "CPU usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": null,
+ "description": "",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 6,
+ "x": 6,
+ "y": 16
+ },
+ "id": 32,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 12,
+ "nullPointMode": "null",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 12,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "100 * (sum(container_memory_usage_bytes{datacenter=\"$site\", namespace=\"$namespace\", pod_name=~\"$pod.*\", container_name!=\"POD\"}) / sum(container_spec_memory_limit_bytes{datacenter=\"$site\", namespace=\"$namespace\", pod_name=~\"$pod.*\", container_name!=\"POD\"}))",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "10s",
+ "intervalFactor": 1,
+ "refId": "A"
+ }
+ ],
+ "thresholds": "85,95",
+ "title": "Memory usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 2,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 12,
+ "y": 16
+ },
+ "id": 33,
+ "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",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 1,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "repeatDirection": "v",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 2,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(container_network_receive_bytes_total{datacenter=\"$site\", namespace=\"$namespace\", pod_name=~\"$pod.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "rx: {{pod_name}}| {{ interface }}",
+ "refId": "B"
+ },
+ {
+ "expr": "-sum(rate(container_network_transmit_bytes_total{datacenter=\"$site\", namespace=\"$namespace\", pod_name=~\"$pod.*\", interface!=\"tunl0\"}[3m]))by (pod_name, interface)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "tx: {{pod_name}}| {{ interface }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Network I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 1,
+ "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": null,
+ "show": true
+ },
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": null,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 0,
+ "y": 21
+ },
+ "id": 34,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 20,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kube_pod_container_resource_limits_cpu_cores{datacenter=\"$site\", namespace=\"$namespace\", pod=~\"$pod.*\"})",
+ "interval": "10s",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU limits",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": null,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 3,
+ "y": 21
+ },
+ "id": 35,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 21,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum (rate(container_cpu_usage_seconds_total{datacenter=\"$site\", namespace=\"$namespace\", image!=\"\", container_name!=\"POD\", pod_name=~\"$pod.*\"}[5m]))",
+ "interval": "10s",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU used (5m avg)",
+ "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": null,
+ "decimals": 2,
+ "format": "decbytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 6,
+ "y": 21
+ },
+ "id": 36,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 22,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kube_pod_container_resource_limits_memory_bytes{datacenter=\"$site\", namespace=\"$namespace\", pod=~\"$pod.*\"})",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory limits",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": null,
+ "decimals": null,
+ "format": "decbytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 9,
+ "y": 21
+ },
+ "id": 37,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "options": {},
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "repeat": null,
+ "repeatDirection": "v",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 23,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false,
+ "ymax": null,
+ "ymin": null
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(container_memory_working_set_bytes{datacenter=\"$site\", namespace=\"$namespace\", container_name!=\"POD\", container_name!=\"\", pod_name=~\"$pod.*\"})",
+ "interval": "10s",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory used",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 3,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 6,
+ "w": 12,
+ "x": 0,
+ "y": 24
+ },
+ "height": "",
+ "id": 38,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 27,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",datacenter=\"$site\", namespace=\"$namespace\",pod_name=~\"^$pod-.*\",name=~\"^k8s_.*\",container_name!=\"POD\"}[5m])) by (container_name, pod_name)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "{{ pod_name }} | {{ container_name }}",
+ "metric": "container_cpu",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod_name=\"$pod-0\",name!~\"^k8s_.*\",datacenter=\"$site\"}[5m])) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "container_cpu",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum (rate (container_cpu_usage_seconds_total{rkt_container_name!=\"\",datacenter=\"$site\"}[5m])) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "container_cpu",
+ "refId": "C",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers CPU usage (5m avg)",
+ "tooltip": {
+ "msResolution": true,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": "cores",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "editable": true,
+ "error": false,
+ "fill": 1,
+ "fillGradient": 0,
+ "grid": {},
+ "gridPos": {
+ "h": 6,
+ "w": 12,
+ "x": 12,
+ "y": 24
+ },
+ "id": 39,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 0,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "dataLinks": []
+ },
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": true,
+ "renderer": "flot",
+ "repeatIteration": 1580642657156,
+ "repeatPanelId": 29,
+ "repeatedByRow": true,
+ "scopedVars": {
+ "pod": {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ },
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",datacenter=\"$site\", namespace=\"$namespace\",pod_name=~\"^$pod.*\",name=~\"^k8s_.*\",container_name!=\"POD\"}) by (container_name, pod_name)",
+ "format": "time_series",
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "{{ pod_name }} | {{ container_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum (container_memory_working_set_bytes{image!=\"\",name!~\"^k8s_.*\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (kubernetes_io_hostname, name, image)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "B",
+ "step": 10
+ },
+ {
+ "expr": "sum (container_memory_working_set_bytes{rkt_container_name!=\"\",datacenter=\"$Site\",kubernetes_io_hostname=~\"^$Node$\",namespace=~\"$Namespace\"}) by (kubernetes_io_hostname, rkt_container_name)",
+ "format": "time_series",
+ "hide": true,
+ "interval": "10s",
+ "intervalFactor": 5,
+ "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}",
+ "metric": "container_memory_usage:sort_desc",
+ "refId": "C",
+ "step": 10
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Containers memory usage",
+ "tooltip": {
+ "msResolution": false,
+ "shared": false,
+ "sort": 2,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "refresh": "1m",
+ "schemaVersion": 20,
+ "style": "dark",
+ "tags": [
+ "aether",
+ "omec"
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "datasource": "Prometheus",
+ "definition": "label_values(kube_pod_info, datacenter)",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Site",
+ "multi": false,
+ "name": "site",
+ "query": "label_values(kube_pod_info, datacenter)",
+ "refresh": 2,
+ "regex": "/.*edge.*/",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "datasource": "Prometheus",
+ "definition": "label_values(kube_pod_info{datacenter=\"$site\"},namespace)",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Namespace",
+ "multi": false,
+ "name": "namespace",
+ "query": "label_values(kube_pod_info{datacenter=\"$site\"},namespace)",
+ "refresh": 2,
+ "regex": "\\b(?:omec|omec-cbrs|omec-b7)\\b",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "tags": [],
+ "text": "All",
+ "value": [
+ "$__all"
+ ]
+ },
+ "hide": 0,
+ "includeAll": true,
+ "label": "Pod",
+ "multi": true,
+ "name": "pod",
+ "options": [
+ {
+ "selected": true,
+ "text": "All",
+ "value": "$__all"
+ },
+ {
+ "selected": false,
+ "text": "spgwu",
+ "value": "spgwu"
+ },
+ {
+ "selected": false,
+ "text": "accelleran",
+ "value": "accelleran"
+ }
+ ],
+ "query": "spgwu,accelleran",
+ "skipUrlSync": false,
+ "type": "custom"
+ }
+ ]
+ },
+ "time": {
+ "from": "now-3h",
+ "to": "now"
+ },
+ "timepicker": {
+ "hidden": false,
+ "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": "OMEC User Plane",
+ "uid": "noPib7UZz",
+ "version": 9
+}
diff --git a/nem-monitoring/templates/grafana-dashboard-configmap.yaml b/nem-monitoring/templates/grafana-dashboard-configmap.yaml
index 1b361cc..8246e6c 100644
--- a/nem-monitoring/templates/grafana-dashboard-configmap.yaml
+++ b/nem-monitoring/templates/grafana-dashboard-configmap.yaml
@@ -52,27 +52,23 @@
omec.json: |
{{ .Files.Get "grafana-dashboards/omec.json" | indent 4 }}
{{- end }}
-{{- if .Values.dashboards.aetherClusters }}
- aether-clusters.json: |
-{{ .Files.Get "grafana-dashboards/aether-clusters.json" | indent 4 }}
-{{- end }}
-{{- if .Values.dashboards.aetherEdges }}
- aether-edges.json: |
-{{ .Files.Get "grafana-dashboards/aether-edges.json" | indent 4 }}
-{{- end }}
-{{- if .Values.dashboards.aetherCentralPods }}
- aether-central-pods.json: |
-{{ .Files.Get "grafana-dashboards/aether-central-pods.json" | indent 4 }}
-{{- end }}
-{{- if .Values.dashboards.aetherEdgePods }}
- aether-edge-pods.json: |
-{{ .Files.Get "grafana-dashboards/aether-edge-pods.json" | indent 4 }}
+{{- if .Values.dashboards.aether }}
+ multi-site-system.json: |
+{{ .Files.Get "grafana-dashboards/multi-site-system.json" | indent 4 }}
+ multi-site-system-summary.json: |
+{{ .Files.Get "grafana-dashboards/multi-site-system-summary.json" | indent 4 }}
+ multi-site-k8s.json: |
+{{ .Files.Get "grafana-dashboards/multi-site-k8s.json" | indent 4 }}
+ omec-user-plane.json: |
+{{ .Files.Get "grafana-dashboards/omec-user-plane.json" | indent 4 }}
+ omec-user-plane-summary.json: |
+{{ .Files.Get "grafana-dashboards/omec-user-plane-summary.json" | indent 4 }}
+ omec-control-plane.json: |
+{{ .Files.Get "grafana-dashboards/omec-control-plane.json" | indent 4 }}
+ omec-imsi-list.json: |
+{{ .Files.Get "grafana-dashboards/omec-imsi-list.json" | indent 4 }}
{{- end }}
{{- if .Values.dashboards.bng }}
bng-dashboard.json: |
{{ .Files.Get "grafana-dashboards/bng-dashboard.json" | indent 4 }}
{{- end }}
-{{- if .Values.dashboards.aetherMMEExporter }}
- aether-mme-exporter.json: |
-{{ .Files.Get "grafana-dashboards/aether-mme-exporter.json" | indent 4 }}
-{{- end }}
diff --git a/nem-monitoring/values.yaml b/nem-monitoring/values.yaml
index b6af9b7..0578f41 100644
--- a/nem-monitoring/values.yaml
+++ b/nem-monitoring/values.yaml
@@ -100,9 +100,5 @@
voltha: true
voltha2: false
omec: false
- aetherClusters: false
- aetherEdges: false
- aetherCentralPods: false
- aetherEdgePods: false
- aetherMMEExporter: false
+ aether: false
bng: false