diff options
Diffstat (limited to 'monitor')
| -rw-r--r-- | monitor/dashboard/automated_dashboard_datasource.py | 5 | ||||
| -rw-r--r-- | monitor/dashboard/jump_server_status.json (renamed from monitor/dashboard/prometheus_system_rev1.json) | 1152 | ||||
| -rw-r--r-- | monitor/dashboard/prototype_prometheus_dashboard.json | 2028 | ||||
| -rw-r--r-- | monitor/dashboard/stats_overview.json (renamed from monitor/dashboard/prototype_dashboard_collapsible.json) | 1257 | ||||
| -rw-r--r-- | monitor/monitoring.sh | 23 |
5 files changed, 1350 insertions, 3115 deletions
diff --git a/monitor/dashboard/automated_dashboard_datasource.py b/monitor/dashboard/automated_dashboard_datasource.py index 48f9710c..abfeca2b 100644 --- a/monitor/dashboard/automated_dashboard_datasource.py +++ b/monitor/dashboard/automated_dashboard_datasource.py @@ -65,7 +65,6 @@ def post(url, data): ip_address = socket.gethostbyname(socket.gethostname()) _create_data_source(ip_address, 3000) -# _create_dashboard(ip_address, 3000, -# MONITOR_DASHBOARD + 'prometheus_system_rev1.json') +_create_dashboard(ip_address, 3000, MONITOR_DASHBOARD + 'stats_overview.json') _create_dashboard(ip_address, 3000, - MONITOR_DASHBOARD + 'prototype_dashboard_collapsible.json') + MONITOR_DASHBOARD + 'jump_server_status.json') diff --git a/monitor/dashboard/prometheus_system_rev1.json b/monitor/dashboard/jump_server_status.json index 78d88e42..837b20e4 100644 --- a/monitor/dashboard/prometheus_system_rev1.json +++ b/monitor/dashboard/jump_server_status.json @@ -1,58 +1,40 @@ { - "__inputs": [ - { - "name": "DS_PROMETHEUS", - "label": "Prometheus", - "description": "", - "type": "datasource", - "pluginId": "prometheus", - "pluginName": "Prometheus" - } - ], - "__requires": [ - { - "type": "panel", - "id": "singlestat", - "name": "Singlestat", - "version": "" - }, - { - "type": "panel", - "id": "graph", - "name": "Graph", - "version": "" - }, - { - "type": "grafana", - "id": "grafana", - "name": "Grafana", - "version": "3.1.0" - }, - { - "type": "datasource", - "id": "prometheus", - "name": "Prometheus", - "version": "1.0.0" - } - ], - "id": null, - "title": "Prometheus system", - "tags": [ - "system", - "prometheus", - "online", - "home" - ], - "style": "dark", - "timezone": "browser", + "annotations": { + "list": [ + { + "datasource": "automated-ds", + "enable": false, + "expr": "ALERTS{alias=\"$host\", alertstate=\"firing\"}", + "iconColor": "rgb(252, 5, 0)", + "name": "Alert", + "tagKeys": "severity", + "textFormat": "{{ alias }} : {{alertstate}}", + "titleFormat": "{{ alertname }}" + }, + { + "datasource": "automated-ds", + "enable": true, + "expr": "ALERTS{alias=\"$host\",alertstate=\"pending\"}", + "iconColor": "rgb(228, 242, 9)", + "name": "Warning", + "tagKeys": "severity", + "textFormat": "{{ alias }} : {{ alertstate }}", + "titleFormat": "{{ alertname }}" + } + ] + }, + "description": "Load, CPU, RAM, network, process ...", "editable": true, + "gnetId": 159, + "graphTooltip": 1, "hideControls": false, - "sharedCrosshair": true, + "id": null, + "links": [], + "refresh": false, "rows": [ { "collapse": false, - "editable": true, - "height": "250px", + "height": "101", "panels": [ { "cacheTimeout": null, @@ -75,7 +57,7 @@ "thresholdLabels": false, "thresholdMarkers": true }, - "height": "50px", + "height": "", "id": 19, "interval": null, "links": [], @@ -111,12 +93,14 @@ "lineColor": "rgb(31, 120, 193)", "show": false }, + "tableColumn": "", "targets": [ { "calculatedInterval": "10m", "datasourceErrors": {}, "errors": {}, "expr": "node_time{alias=\"$host\"} - node_boot_time{alias=\"$host\"}", + "format": "time_series", "interval": "5m", "intervalFactor": 1, "legendFormat": "", @@ -154,10 +138,9 @@ "thresholdLabels": false, "thresholdMarkers": true }, - "height": "55px", + "height": "", "id": 25, "interval": null, - "isNew": true, "links": [], "mappingType": 1, "mappingTypes": [ @@ -191,6 +174,7 @@ "lineColor": "rgb(31, 120, 193)", "show": false }, + "tableColumn": "", "targets": [ { "expr": "count(node_cpu{mode=\"user\", alias=\"$host\"})", @@ -234,10 +218,9 @@ "thresholdLabels": false, "thresholdMarkers": true }, - "height": "55px", + "height": "", "id": 26, "interval": null, - "isNew": true, "links": [], "mappingType": 1, "mappingTypes": [ @@ -271,15 +254,17 @@ "lineColor": "rgb(31, 120, 193)", "show": false }, + "tableColumn": "", "targets": [ { "expr": "node_memory_MemAvailable{alias=\"$host\"}", + "format": "time_series", "interval": "", "intervalFactor": 1, "legendFormat": "", "metric": "node_memory_MemAvailable", "refId": "A", - "step": 30 + "step": 1 } ], "thresholds": "", @@ -316,7 +301,7 @@ "thresholdLabels": false, "thresholdMarkers": true }, - "height": "50px", + "height": "", "id": 9, "interval": null, "links": [], @@ -350,21 +335,23 @@ "fillColor": "rgba(31, 118, 189, 0.18)", "full": true, "lineColor": "rgb(31, 120, 193)", - "show": true + "show": false }, + "tableColumn": "", "targets": [ { "calculatedInterval": "10m", "datasourceErrors": {}, "errors": {}, "expr": "(node_memory_MemAvailable{alias=\"$host\"} or (node_memory_MemFree{alias=\"$host\"} + node_memory_Buffers{alias=\"$host\"} + node_memory_Cached{alias=\"$host\"})) / node_memory_MemTotal{alias=\"$host\"} * 100", - "interval": "5m", + "format": "time_series", + "interval": "", "intervalFactor": 1, "legendFormat": "", "metric": "node_mem", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%20%2F%20node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20*%20100%22%2C%22range_input%22%3A%2243201s%22%2C%22end_input%22%3A%222015-9-15%2013%3A54%22%2C%22step_input%22%3A%22%22%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 300 + "step": 1 } ], "thresholds": "90,95", @@ -376,106 +363,42 @@ "valueName": "current" }, { - "aliasColors": {}, - "bars": false, - "datasource": "automated-ds", - "decimals": 2, - "editable": true, - "error": false, - "fill": 6, - "grid": { - "threshold1": null, - "threshold1Color": "rgba(216, 200, 27, 0.27)", - "threshold2": null, - "threshold2Color": "rgba(234, 112, 112, 0.22)", - "thresholdLine": false - }, - "height": "260px", - "id": 2, - "legend": { - "alignAsTable": true, - "avg": true, - "current": false, - "hideEmpty": false, - "max": true, - "min": true, - "rightSide": true, - "show": true, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 2, + "headings": true, + "id": 32, + "limit": 10, "links": [], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], + "query": "", + "recent": true, + "search": false, "span": 12, - "stack": true, - "steppedLine": false, - "targets": [ - { - "calculatedInterval": "2s", - "datasourceErrors": {}, - "errors": {}, - "expr": "sum(rate(node_cpu{alias=\"$host\"}[$interval])) by (mode) * 100 / count_scalar(node_cpu{mode=\"user\", alias=\"$host\"}) or sum(irate(node_cpu{alias=\"$host\"}[5m])) by (mode) * 100 / count_scalar(node_cpu{mode=\"user\", alias=\"$host\"})", - "intervalFactor": 1, - "legendFormat": "{{ mode }}", - "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22sum(rate(node_cpu%7Balias%3D%5C%22%24host%5C%22%7D%5B%24interval%5D))%20by%20(mode)%20*%20100%22%2C%22range_input%22%3A%223600s%22%2C%22end_input%22%3A%222015-10-22%2015%3A27%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "A", - "step": 2 - } - ], - "timeFrom": null, - "timeShift": null, - "title": "CPU Usage", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "show": true - }, - "yaxes": [ - { - "format": "percent", - "label": "", - "logBase": 1, - "max": 100, - "min": 0, - "show": true - }, - { - "format": "short", - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - }, + "starred": true, + "tags": [], + "title": "Status Quick Access", + "type": "dashlist" + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "System Stats", + "titleSize": "h5" + }, + { + "collapse": false, + "height": 299, + "panels": [ { "aliasColors": {}, "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, "fill": 2, - "grid": { - "threshold1": null, - "threshold1Color": "rgb(241, 156, 15)", - "threshold2": null, - "threshold2Color": "rgb(246, 17, 17)", - "thresholdLine": true - }, + "grid": {}, "id": 18, "legend": { "alignAsTable": true, @@ -512,6 +435,7 @@ "color": "#E5AC0E" } ], + "spaceLength": 10, "span": 12, "stack": false, "steppedLine": false, @@ -526,7 +450,7 @@ "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_load1%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%223601s%22%2C%22end_input%22%3A%222015-10-22%2015%3A27%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Afalse%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 2, + "step": 1, "target": "" }, { @@ -539,7 +463,7 @@ "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_load5%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%223600s%22%2C%22end_input%22%3A%222015-10-22%2015%3A27%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Afalse%2C%22tab%22%3A0%7D%5D", "refId": "B", - "step": 2, + "step": 1, "target": "" }, { @@ -552,10 +476,11 @@ "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_load15%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%223600s%22%2C%22end_input%22%3A%222015-10-22%2015%3A27%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Afalse%2C%22tab%22%3A0%7D%5D", "refId": "C", - "step": 2, + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, "title": "Load Average", @@ -568,7 +493,11 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { @@ -587,33 +516,20 @@ "show": true } ] - } - ], - "showTitle": false, - "title": "System Stats" - }, - { - "collapse": false, - "editable": true, - "height": "300px", - "panels": [ + }, { "aliasColors": {}, "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, "fill": 6, - "grid": { - "threshold1": null, - "threshold1Color": "rgba(216, 200, 27, 0.27)", - "threshold2": null, - "threshold2Color": "rgba(234, 112, 112, 0.22)", - "thresholdLine": false - }, + "grid": {}, "height": "", - "id": 6, + "id": 2, "legend": { "alignAsTable": true, "avg": true, @@ -621,7 +537,7 @@ "hideEmpty": false, "max": true, "min": true, - "rightSide": false, + "rightSide": true, "show": true, "total": false, "values": true @@ -634,23 +550,9 @@ "pointradius": 5, "points": false, "renderer": "flot", - "seriesOverrides": [ - { - "alias": "Used", - "color": "#0A437C" - }, - { - "alias": "Available", - "color": "#5195CE" - }, - { - "alias": "Total", - "color": "#052B51", - "legend": false, - "stack": false - } - ], - "span": 6, + "seriesOverrides": [], + "spaceLength": 10, + "span": 12, "stack": true, "steppedLine": false, "targets": [ @@ -658,67 +560,45 @@ "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "node_memory_MemTotal{alias=\"$host\"}", - "intervalFactor": 1, - "legendFormat": "Total", - "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "C", - "step": 5, - "target": "" - }, - { - "calculatedInterval": "2s", - "datasourceErrors": {}, - "errors": {}, - "expr": "node_memory_MemTotal{alias=\"$host\"} - (node_memory_MemAvailable{alias=\"$host\"} or (node_memory_MemFree{alias=\"$host\"} + node_memory_Buffers{alias=\"$host\"} + node_memory_Cached{alias=\"$host\"}))", + "expr": "sum(rate(node_cpu{alias=\"$host\"}[$interval])) by (mode) * 100 / count_scalar(node_cpu{mode=\"user\", alias=\"$host\"}) or sum(irate(node_cpu{alias=\"$host\"}[5m])) by (mode) * 100 / count_scalar(node_cpu{mode=\"user\", alias=\"$host\"})", + "format": "time_series", "intervalFactor": 1, - "legendFormat": "Used", + "legendFormat": "{{ mode }}", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22sum(rate(node_cpu%7Balias%3D%5C%22%24host%5C%22%7D%5B%24interval%5D))%20by%20(mode)%20*%20100%22%2C%22range_input%22%3A%223600s%22%2C%22end_input%22%3A%222015-10-22%2015%3A27%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 5, - "target": "" - }, - { - "calculatedInterval": "2s", - "datasourceErrors": {}, - "errors": {}, - "expr": "node_memory_MemAvailable{alias=\"$host\"} or (node_memory_MemFree{alias=\"$host\"} + node_memory_Buffers{alias=\"$host\"} + node_memory_Cached{alias=\"$host\"})", - "intervalFactor": 1, - "legendFormat": "Available", - "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "B", - "step": 5, - "target": "" + "step": 1 } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Memory", + "title": "CPU Usage", "tooltip": { "msResolution": false, "shared": true, "sort": 0, "value_type": "individual" }, - "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { - "format": "bytes", + "format": "percent", "label": "", "logBase": 1, - "max": null, + "max": 100, "min": 0, "show": true }, { - "format": "bytes", + "format": "short", "logBase": 1, "max": null, "min": 0, @@ -728,21 +608,16 @@ }, { "aliasColors": {}, - "bars": false, + "bars": true, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, - "fill": 6, - "grid": { - "threshold1": null, - "threshold1Color": "rgba(216, 200, 27, 0.27)", - "threshold2": null, - "threshold2Color": "rgba(234, 112, 112, 0.22)", - "thresholdLine": false - }, - "height": "", - "id": 29, + "fill": 2, + "grid": {}, + "id": 20, "legend": { "alignAsTable": true, "avg": true, @@ -755,7 +630,7 @@ "total": false, "values": true }, - "lines": true, + "lines": false, "linewidth": 2, "links": [], "nullPointMode": "null", @@ -763,67 +638,52 @@ "pointradius": 5, "points": false, "renderer": "flot", - "seriesOverrides": [], - "span": 6, + "seriesOverrides": [ + { + "alias": "Processes blocked waiting for I/O to complete", + "color": "#E24D42" + }, + { + "alias": "Processes in runnable state", + "color": "#6ED0E0" + } + ], + "spaceLength": 10, + "span": 5, "stack": true, "steppedLine": false, "targets": [ { - "calculatedInterval": "2s", + "calculatedInterval": "2m", "datasourceErrors": {}, "errors": {}, - "expr": "node_memory_MemTotal{alias=\"$host\"} - (node_memory_MemFree{alias=\"$host\"} + node_memory_Buffers{alias=\"$host\"} + node_memory_Cached{alias=\"$host\"})", + "expr": "node_procs_running{alias=\"$host\"}", "intervalFactor": 1, - "legendFormat": "Used", + "legendFormat": "Processes in runnable state", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_procs_running%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%2243200s%22%2C%22end_input%22%3A%222015-9-18%2013%3A46%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 5, + "step": 1, "target": "" }, { - "calculatedInterval": "2s", + "calculatedInterval": "2m", "datasourceErrors": {}, "errors": {}, - "expr": "node_memory_MemFree{alias=\"$host\"}", + "expr": "node_procs_blocked{alias=\"$host\"}", "intervalFactor": 1, - "legendFormat": "Free", + "legendFormat": "Processes blocked waiting for I/O to complete", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_procs_blocked%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%2243200s%22%2C%22end_input%22%3A%222015-9-18%2013%3A46%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "B", - "step": 5, - "target": "" - }, - { - "calculatedInterval": "2s", - "datasourceErrors": {}, - "errors": {}, - "expr": "node_memory_Buffers{alias=\"$host\"}", - "intervalFactor": 1, - "legendFormat": "Buffers", - "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "D", - "step": 5, - "target": "" - }, - { - "calculatedInterval": "2s", - "datasourceErrors": {}, - "errors": {}, - "expr": "node_memory_Cached{alias=\"$host\"}", - "intervalFactor": 1, - "legendFormat": "Cached", - "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "E", - "step": 5, + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Memory Distribution", + "title": "Processes", "tooltip": { "msResolution": false, "shared": true, @@ -833,11 +693,15 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { - "format": "bytes", + "format": "none", "label": "", "logBase": 1, "max": null, @@ -845,7 +709,7 @@ "show": true }, { - "format": "bytes", + "format": "none", "logBase": 1, "max": null, "min": 0, @@ -856,18 +720,14 @@ { "aliasColors": {}, "bars": true, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, "fill": 2, - "grid": { - "threshold1": null, - "threshold1Color": "rgb(241, 156, 15)", - "threshold2": null, - "threshold2Color": "rgb(246, 17, 17)", - "thresholdLine": true - }, + "grid": {}, "id": 24, "legend": { "alignAsTable": true, @@ -895,7 +755,8 @@ "color": "#EF843C" } ], - "span": 6, + "spaceLength": 10, + "span": 3, "stack": false, "steppedLine": false, "targets": [ @@ -904,15 +765,17 @@ "datasourceErrors": {}, "errors": {}, "expr": "rate(node_forks{alias=\"$host\"}[$interval]) or irate(node_forks{alias=\"$host\"}[5m])", + "format": "time_series", "intervalFactor": 1, "legendFormat": "Forks", "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_procs_running%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%2243200s%22%2C%22end_input%22%3A%222015-9-18%2013%3A46%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 5, + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, "title": "Forks", @@ -925,7 +788,11 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { @@ -947,20 +814,16 @@ }, { "aliasColors": {}, - "bars": true, + "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, "fill": 2, - "grid": { - "threshold1": null, - "threshold1Color": "rgb(241, 156, 15)", - "threshold2": null, - "threshold2Color": "rgb(246, 17, 17)", - "thresholdLine": true - }, - "id": 20, + "grid": {}, + "id": 28, "legend": { "alignAsTable": true, "avg": true, @@ -973,7 +836,7 @@ "total": false, "values": true }, - "lines": false, + "lines": true, "linewidth": 2, "links": [], "nullPointMode": "null", @@ -983,50 +846,35 @@ "renderer": "flot", "seriesOverrides": [ { - "alias": "Processes blocked waiting for I/O to complete", - "color": "#E24D42" - }, - { - "alias": "Processes in runnable state", - "color": "#6ED0E0" + "alias": "Interrupts", + "color": "#D683CE" } ], - "span": 6, - "stack": true, + "spaceLength": 10, + "span": 4, + "stack": false, "steppedLine": false, "targets": [ { "calculatedInterval": "2m", "datasourceErrors": {}, "errors": {}, - "expr": "node_procs_running{alias=\"$host\"}", + "expr": "rate(node_intr{alias=\"$host\"}[$interval]) or irate(node_intr{alias=\"$host\"}[5m])", "intervalFactor": 1, - "legendFormat": "Processes in runnable state", + "legendFormat": "Interrupts", "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_procs_running%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%2243200s%22%2C%22end_input%22%3A%222015-9-18%2013%3A46%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 5, - "target": "" - }, - { - "calculatedInterval": "2m", - "datasourceErrors": {}, - "errors": {}, - "expr": "node_procs_blocked{alias=\"$host\"}", - "intervalFactor": 1, - "legendFormat": "Processes blocked waiting for I/O to complete", - "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_procs_blocked%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%2243200s%22%2C%22end_input%22%3A%222015-9-18%2013%3A46%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "B", - "step": 5, + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Processes", + "title": "Interrupts", "tooltip": { - "msResolution": false, + "msResolution": true, "shared": true, "sort": 0, "value_type": "individual" @@ -1034,7 +882,11 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { @@ -1053,23 +905,31 @@ "show": true } ] - }, + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "CPU Stats", + "titleSize": "h5" + }, + { + "collapse": false, + "height": 291, + "panels": [ { "aliasColors": {}, "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, "fill": 2, - "grid": { - "threshold1": null, - "threshold1Color": "rgb(241, 156, 15)", - "threshold2": null, - "threshold2Color": "rgb(246, 17, 17)", - "thresholdLine": true - }, - "id": 27, + "grid": {}, + "id": 31, "legend": { "alignAsTable": true, "avg": true, @@ -1091,27 +951,42 @@ "points": false, "renderer": "flot", "seriesOverrides": [], - "span": 6, + "spaceLength": 10, + "span": 12, "stack": false, "steppedLine": false, "targets": [ { - "calculatedInterval": "2m", + "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "rate(node_context_switches{alias=\"$host\"}[$interval]) or irate(node_context_switches{alias=\"$host\"}[5m])", + "expr": "rate(node_vmstat_pgpgin{alias=\"$host\"}[$interval]) * 1024 or irate(node_vmstat_pgpgin{alias=\"$host\"}[5m]) * 1024", "intervalFactor": 1, - "legendFormat": "Context Switches", + "legendFormat": "Page In", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_procs_running%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%2243200s%22%2C%22end_input%22%3A%222015-9-18%2013%3A46%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 5, + "step": 1, + "target": "" + }, + { + "calculatedInterval": "2s", + "datasourceErrors": {}, + "errors": {}, + "expr": "rate(node_vmstat_pgpgout{alias=\"$host\"}[$interval]) * 1024 or irate(node_vmstat_pgpgout{alias=\"$host\"}[5m]) * 1024", + "intervalFactor": 1, + "legendFormat": "Page Out", + "metric": "", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "B", + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Context Switches", + "title": "I/O Activity", "tooltip": { "msResolution": false, "shared": true, @@ -1121,11 +996,15 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { - "format": "none", + "format": "Bps", "label": "", "logBase": 1, "max": null, @@ -1133,7 +1012,7 @@ "show": true }, { - "format": "none", + "format": "bytes", "logBase": 1, "max": null, "min": 0, @@ -1144,19 +1023,16 @@ { "aliasColors": {}, "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, - "fill": 2, - "grid": { - "threshold1": null, - "threshold1Color": "rgb(241, 156, 15)", - "threshold2": null, - "threshold2Color": "rgb(246, 17, 17)", - "thresholdLine": true - }, - "id": 28, + "fill": 6, + "grid": {}, + "height": "", + "id": 6, "legend": { "alignAsTable": true, "avg": true, @@ -1179,33 +1055,71 @@ "renderer": "flot", "seriesOverrides": [ { - "alias": "Interrupts", - "color": "#D683CE" + "alias": "Used", + "color": "#0A437C" + }, + { + "alias": "Available", + "color": "#5195CE" + }, + { + "alias": "Total", + "color": "#052B51", + "legend": false, + "stack": false } ], + "spaceLength": 10, "span": 6, - "stack": false, + "stack": true, "steppedLine": false, "targets": [ { - "calculatedInterval": "2m", + "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "rate(node_intr{alias=\"$host\"}[$interval]) or irate(node_intr{alias=\"$host\"}[5m])", + "expr": "node_memory_MemTotal{alias=\"$host\"}", "intervalFactor": 1, - "legendFormat": "Interrupts", + "legendFormat": "Total", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_procs_running%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%2243200s%22%2C%22end_input%22%3A%222015-9-18%2013%3A46%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "C", + "step": 1, + "target": "" + }, + { + "calculatedInterval": "2s", + "datasourceErrors": {}, + "errors": {}, + "expr": "node_memory_MemTotal{alias=\"$host\"} - (node_memory_MemAvailable{alias=\"$host\"} or (node_memory_MemFree{alias=\"$host\"} + node_memory_Buffers{alias=\"$host\"} + node_memory_Cached{alias=\"$host\"}))", + "intervalFactor": 1, + "legendFormat": "Used", + "metric": "", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 5, + "step": 1, + "target": "" + }, + { + "calculatedInterval": "2s", + "datasourceErrors": {}, + "errors": {}, + "expr": "node_memory_MemAvailable{alias=\"$host\"} or (node_memory_MemFree{alias=\"$host\"} + node_memory_Buffers{alias=\"$host\"} + node_memory_Cached{alias=\"$host\"})", + "intervalFactor": 1, + "legendFormat": "Available", + "metric": "", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "B", + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Interrupts", + "title": "Memory", "tooltip": { - "msResolution": true, + "msResolution": false, "shared": true, "sort": 0, "value_type": "individual" @@ -1213,11 +1127,15 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { - "format": "none", + "format": "bytes", "label": "", "logBase": 1, "max": null, @@ -1225,7 +1143,7 @@ "show": true }, { - "format": "none", + "format": "bytes", "logBase": 1, "max": null, "min": 0, @@ -1236,19 +1154,16 @@ { "aliasColors": {}, "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, "fill": 6, - "grid": { - "threshold1": null, - "threshold1Color": "rgba(216, 200, 27, 0.27)", - "threshold2": null, - "threshold2Color": "rgba(234, 112, 112, 0.22)", - "thresholdLine": false - }, - "id": 21, + "grid": {}, + "height": "", + "id": 29, "legend": { "alignAsTable": true, "avg": true, @@ -1270,6 +1185,7 @@ "points": false, "renderer": "flot", "seriesOverrides": [], + "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, @@ -1278,32 +1194,59 @@ "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "rate(node_network_receive_bytes{alias=\"$host\", device!=\"lo\"}[$interval]) or irate(node_network_receive_bytes{alias=\"$host\", device!=\"lo\"}[5m])", + "expr": "node_memory_MemTotal{alias=\"$host\"} - (node_memory_MemFree{alias=\"$host\"} + node_memory_Buffers{alias=\"$host\"} + node_memory_Cached{alias=\"$host\"})", "intervalFactor": 1, - "legendFormat": "Inbound: {{ device }}", + "legendFormat": "Used", + "metric": "", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "A", + "step": 1, + "target": "" + }, + { + "calculatedInterval": "2s", + "datasourceErrors": {}, + "errors": {}, + "expr": "node_memory_MemFree{alias=\"$host\"}", + "intervalFactor": 1, + "legendFormat": "Free", "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "B", - "step": 5, + "step": 1, "target": "" }, { "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "rate(node_network_transmit_bytes{alias=\"$host\", device!=\"lo\"}[$interval]) or irate(node_network_transmit_bytes{alias=\"$host\", device!=\"lo\"}[5m])", + "expr": "node_memory_Buffers{alias=\"$host\"}", "intervalFactor": 1, - "legendFormat": "Outbound: {{ device }}", + "legendFormat": "Buffers", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "A", - "step": 5, + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "D", + "step": 1, + "target": "" + }, + { + "calculatedInterval": "2s", + "datasourceErrors": {}, + "errors": {}, + "expr": "node_memory_Cached{alias=\"$host\"}", + "intervalFactor": 1, + "legendFormat": "Cached", + "metric": "", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "E", + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Network Traffic", + "title": "Memory Distribution", "tooltip": { "msResolution": false, "shared": true, @@ -1313,11 +1256,15 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { - "format": "Bps", + "format": "bytes", "label": "", "logBase": 1, "max": null, @@ -1335,20 +1282,16 @@ }, { "aliasColors": {}, - "bars": true, + "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, "fill": 6, - "grid": { - "threshold1": null, - "threshold1Color": "rgba(216, 200, 27, 0.27)", - "threshold2": null, - "threshold2Color": "rgba(234, 112, 112, 0.22)", - "thresholdLine": false - }, - "id": 22, + "grid": {}, + "id": 23, "legend": { "alignAsTable": true, "avg": true, @@ -1358,12 +1301,10 @@ "min": true, "rightSide": false, "show": true, - "sort": "min", - "sortDesc": true, "total": false, "values": true }, - "lines": false, + "lines": true, "linewidth": 2, "links": [], "nullPointMode": "null", @@ -1371,7 +1312,17 @@ "pointradius": 5, "points": false, "renderer": "flot", - "seriesOverrides": [], + "seriesOverrides": [ + { + "alias": "Used", + "color": "#584477" + }, + { + "alias": "Free", + "color": "#AEA2E0" + } + ], + "spaceLength": 10, "span": 6, "stack": true, "steppedLine": false, @@ -1380,34 +1331,33 @@ "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "sum(increase(node_network_receive_bytes{alias=\"$host\", device!=\"lo\"}[1h]))", - "interval": "1h", + "expr": "node_memory_SwapTotal{alias=\"$host\"} - node_memory_SwapFree{alias=\"$host\"}", "intervalFactor": 1, - "legendFormat": "Received", + "legendFormat": "Used", "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 3600, + "step": 1, "target": "" }, { "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "sum(increase(node_network_transmit_bytes{alias=\"$host\", device!=\"lo\"}[1h]))", - "interval": "1h", + "expr": "node_memory_SwapFree{alias=\"$host\"}", "intervalFactor": 1, - "legendFormat": "Sent", + "legendFormat": "Free", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "B", - "step": 3600, + "step": 1, "target": "" } ], - "timeFrom": "24h", + "thresholds": [], + "timeFrom": null, "timeShift": null, - "title": "Network Utilization Hourly", + "title": "Swap", "tooltip": { "msResolution": false, "shared": true, @@ -1417,7 +1367,11 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { @@ -1440,19 +1394,15 @@ { "aliasColors": {}, "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, - "fill": 6, - "grid": { - "threshold1": null, - "threshold1Color": "rgba(216, 200, 27, 0.27)", - "threshold2": null, - "threshold2Color": "rgba(234, 112, 112, 0.22)", - "thresholdLine": false - }, - "id": 23, + "fill": 2, + "grid": {}, + "id": 30, "legend": { "alignAsTable": true, "avg": true, @@ -1473,50 +1423,43 @@ "pointradius": 5, "points": false, "renderer": "flot", - "seriesOverrides": [ - { - "alias": "Used", - "color": "#584477" - }, - { - "alias": "Free", - "color": "#AEA2E0" - } - ], + "seriesOverrides": [], + "spaceLength": 10, "span": 6, - "stack": true, + "stack": false, "steppedLine": false, "targets": [ { "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "node_memory_SwapTotal{alias=\"$host\"} - node_memory_SwapFree{alias=\"$host\"}", + "expr": "rate(node_vmstat_pswpin{alias=\"$host\"}[$interval]) * 4096 or irate(node_vmstat_pswpin{alias=\"$host\"}[5m]) * 4096", "intervalFactor": 1, - "legendFormat": "Used", + "legendFormat": "Swap In", "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 5, + "step": 1, "target": "" }, { "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "node_memory_SwapFree{alias=\"$host\"}", + "expr": "rate(node_vmstat_pswpout{alias=\"$host\"}[$interval]) * 4096 or irate(node_vmstat_pswpout{alias=\"$host\"}[5m]) * 4096", "intervalFactor": 1, - "legendFormat": "Free", + "legendFormat": "Swap Out", "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "B", - "step": 5, + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Swap", + "title": "Swap Activity", "tooltip": { "msResolution": false, "shared": true, @@ -1526,11 +1469,15 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { - "format": "bytes", + "format": "Bps", "label": "", "logBase": 1, "max": null, @@ -1545,23 +1492,31 @@ "show": true } ] - }, + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Storage Stats", + "titleSize": "h5" + }, + { + "collapse": false, + "height": 283, + "panels": [ { "aliasColors": {}, "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, - "fill": 2, - "grid": { - "threshold1": null, - "threshold1Color": "rgba(216, 200, 27, 0.27)", - "threshold2": null, - "threshold2Color": "rgba(234, 112, 112, 0.22)", - "thresholdLine": false - }, - "id": 30, + "fill": 6, + "grid": {}, + "id": 21, "legend": { "alignAsTable": true, "avg": true, @@ -1569,7 +1524,7 @@ "hideEmpty": false, "max": true, "min": true, - "rightSide": false, + "rightSide": true, "show": true, "total": false, "values": true @@ -1583,40 +1538,44 @@ "points": false, "renderer": "flot", "seriesOverrides": [], - "span": 6, - "stack": false, + "spaceLength": 10, + "span": 12, + "stack": true, "steppedLine": false, "targets": [ { "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "rate(node_vmstat_pswpin{alias=\"$host\"}[$interval]) * 4096 or irate(node_vmstat_pswpin{alias=\"$host\"}[5m]) * 4096", + "expr": "rate(node_network_receive_bytes{alias=\"$host\", device!=\"lo\"}[$interval]) or irate(node_network_receive_bytes{alias=\"$host\", device!=\"lo\"}[5m])", + "format": "time_series", "intervalFactor": 1, - "legendFormat": "Swap In", + "legendFormat": "Inbound: {{ device }}", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "A", - "step": 5, + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "B", + "step": 1, "target": "" }, { "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "rate(node_vmstat_pswpout{alias=\"$host\"}[$interval]) * 4096 or irate(node_vmstat_pswpout{alias=\"$host\"}[5m]) * 4096", + "expr": "rate(node_network_transmit_bytes{alias=\"$host\", device!=\"lo\"}[$interval]) or irate(node_network_transmit_bytes{alias=\"$host\", device!=\"lo\"}[5m])", + "format": "time_series", "intervalFactor": 1, - "legendFormat": "Swap Out", + "legendFormat": "Outbound: {{ device }}", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", - "refId": "B", - "step": 5, + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "A", + "step": 1, "target": "" } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Swap Activity", + "title": "Network Traffic", "tooltip": { "msResolution": false, "shared": true, @@ -1626,7 +1585,11 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { @@ -1649,19 +1612,15 @@ { "aliasColors": {}, "bars": false, + "dashLength": 10, + "dashes": false, "datasource": "automated-ds", "decimals": 2, "editable": true, "error": false, "fill": 2, - "grid": { - "threshold1": null, - "threshold1Color": "rgba(216, 200, 27, 0.27)", - "threshold2": null, - "threshold2Color": "rgba(234, 112, 112, 0.22)", - "thresholdLine": false - }, - "id": 31, + "grid": {}, + "id": 27, "legend": { "alignAsTable": true, "avg": true, @@ -1683,40 +1642,136 @@ "points": false, "renderer": "flot", "seriesOverrides": [], + "spaceLength": 10, "span": 6, "stack": false, "steppedLine": false, "targets": [ { + "calculatedInterval": "2m", + "datasourceErrors": {}, + "errors": {}, + "expr": "rate(node_context_switches{alias=\"$host\"}[$interval]) or irate(node_context_switches{alias=\"$host\"}[5m])", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Context Switches", + "metric": "", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_procs_running%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%2243200s%22%2C%22end_input%22%3A%222015-9-18%2013%3A46%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "refId": "A", + "step": 1, + "target": "" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Context Switches", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "transparent": false, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "none", + "label": "", + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "none", + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": {}, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "automated-ds", + "decimals": 2, + "editable": true, + "error": false, + "fill": 6, + "grid": {}, + "id": 22, + "legend": { + "alignAsTable": true, + "avg": true, + "current": false, + "hideEmpty": false, + "max": true, + "min": true, + "rightSide": false, + "show": true, + "sort": "min", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 2, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "rate(node_vmstat_pgpgin{alias=\"$host\"}[$interval]) * 1024 or irate(node_vmstat_pgpgin{alias=\"$host\"}[5m]) * 1024", + "expr": "sum(increase(node_network_receive_bytes{alias=\"$host\", device!=\"lo\"}[1h]))", + "interval": "1h", "intervalFactor": 1, - "legendFormat": "Page In", + "legendFormat": "Received", "metric": "", "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "A", - "step": 5, + "step": 3600, "target": "" }, { "calculatedInterval": "2s", "datasourceErrors": {}, "errors": {}, - "expr": "rate(node_vmstat_pgpgout{alias=\"$host\"}[$interval]) * 1024 or irate(node_vmstat_pgpgout{alias=\"$host\"}[5m]) * 1024", + "expr": "sum(increase(node_network_transmit_bytes{alias=\"$host\", device!=\"lo\"}[1h]))", + "interval": "1h", "intervalFactor": 1, - "legendFormat": "Page Out", + "legendFormat": "Sent", "metric": "", - "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", + "prometheusLink": "/api/datasources/proxy/1/graph#%5B%7B%22expr%22%3A%22node_memory_MemTotal%7Balias%3D%5C%22%24host%5C%22%7D%20-%20(node_memory_MemFree%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Buffers%7Balias%3D%5C%22%24host%5C%22%7D%20%2B%20node_memory_Cached%7Balias%3D%5C%22%24host%5C%22%7D)%22%2C%22range_input%22%3A%22900s%22%2C%22end_input%22%3A%222015-10-22%2015%3A25%22%2C%22step_input%22%3A%22%22%2C%22stacked%22%3Atrue%2C%22tab%22%3A0%7D%5D", "refId": "B", - "step": 5, + "step": 3600, "target": "" } ], - "timeFrom": null, + "thresholds": [], + "timeFrom": "24h", "timeShift": null, - "title": "I/O Activity", + "title": "Network Utilization Hourly", "tooltip": { "msResolution": false, "shared": true, @@ -1726,11 +1781,15 @@ "transparent": false, "type": "graph", "xaxis": { - "show": true + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] }, "yaxes": [ { - "format": "Bps", + "format": "bytes", "label": "", "logBase": 1, "max": null, @@ -1747,44 +1806,17 @@ ] } ], - "title": "New row" + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Network Stats", + "titleSize": "h5" } ], - "time": { - "from": "now-1h", - "to": "now" - }, - "timepicker": { - "collapse": false, - "enable": true, - "notice": false, - "now": true, - "refresh_intervals": [ - "5s", - "10s", - "30s", - "1m", - "5m", - "15m", - "30m", - "1h", - "2h", - "1d" - ], - "status": "Stable", - "time_options": [ - "5m", - "15m", - "1h", - "6h", - "12h", - "24h", - "2d", - "7d", - "30d" - ], - "type": "timepicker" - }, + "schemaVersion": 14, + "style": "dark", + "tags": [], "templating": { "list": [ { @@ -1793,7 +1825,6 @@ "auto_count": 200, "auto_min": "1s", "current": { - "selected": true, "text": "5s", "value": "5s" }, @@ -1847,12 +1878,17 @@ } ], "query": "1s,5s,1m,5m,1h,6h,1d", - "refresh": 0, + "refresh": 2, "type": "interval" }, { "allFormat": "glob", - "current": {}, + "allValue": null, + "current": { + "isNone": true, + "text": "None", + "value": "" + }, "datasource": "automated-ds", "hide": 0, "includeAll": false, @@ -1865,41 +1901,51 @@ "refresh": 1, "refresh_on_load": false, "regex": "", + "sort": 0, "tagValuesQuery": "alias", + "tags": [], "tagsQuery": "up", "type": "query", "useTags": false } ] }, - "annotations": { - "list": [ - { - "datasource": "automated-ds", - "enable": false, - "expr": "ALERTS{alias=\"$host\", alertstate=\"firing\"}", - "iconColor": "rgb(252, 5, 0)", - "name": "Alert", - "tagKeys": "severity", - "textFormat": "{{ alias }} : {{alertstate}}", - "titleFormat": "{{ alertname }}" - }, - { - "datasource": "automated-ds", - "enable": true, - "expr": "ALERTS{alias=\"$host\",alertstate=\"pending\"}", - "iconColor": "rgb(228, 242, 9)", - "name": "Warning", - "tagKeys": "severity", - "textFormat": "{{ alias }} : {{ alertstate }}", - "titleFormat": "{{ alertname }}" - } - ] + "time": { + "from": "now-30m", + "to": "now" }, - "refresh": "30s", - "schemaVersion": 12, - "version": 18, - "links": [], - "gnetId": 159, - "description": "Prometheus for system metrics. \r\nLoad, CPU, RAM, network, process ... " + "timepicker": { + "collapse": false, + "enable": true, + "notice": false, + "now": true, + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "status": "Stable", + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ], + "type": "timepicker" + }, + "timezone": "browser", + "title": "Jump Server Status", + "version": 4 }
\ No newline at end of file diff --git a/monitor/dashboard/prototype_prometheus_dashboard.json b/monitor/dashboard/prototype_prometheus_dashboard.json deleted file mode 100644 index 36c77c39..00000000 --- a/monitor/dashboard/prototype_prometheus_dashboard.json +++ /dev/null @@ -1,2028 +0,0 @@ -{ - "annotations": { - "list": [] - }, - "description": "A simple overview of the most important Docker host and container metrics. (cAdvisor/Prometheus)", - "editable": true, - "gnetId": 893, - "graphTooltip": 1, - "hideControls": false, - "id": 6, - "links": [], - "refresh": "15m", - "rows": [ - { - "collapse": false, - "height": 212, - "panels": [ - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "rgba(245, 54, 54, 0.9)", - "rgba(237, 129, 40, 0.89)", - "rgba(50, 172, 45, 0.97)" - ], - "datasource": "automated-ds", - "decimals": 0, - "editable": true, - "error": false, - "format": "s", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "height": "", - "id": 24, - "interval": null, - "links": [], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "30%", - "prefix": "", - "prefixFontSize": "20%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "time() - node_boot_time{instance=~\"$server:.*\"}", - "hide": false, - "intervalFactor": 2, - "legendFormat": "", - "refId": "A", - "step": 1800 - } - ], - "thresholds": "", - "title": "Uptime", - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "current" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "rgba(245, 54, 54, 0.9)", - "rgba(237, 129, 40, 0.89)", - "rgba(50, 172, 45, 0.97)" - ], - "datasource": "automated-ds", - "editable": true, - "error": false, - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "id": 31, - "interval": null, - "links": [], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "count(rate(container_last_seen{name=~\".+\"}[$interval]))", - "intervalFactor": 2, - "refId": "A", - "step": 1800 - } - ], - "thresholds": "", - "title": "Containers", - "type": "singlestat", - "valueFontSize": "120%", - "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": "automated-ds", - "decimals": 1, - "editable": true, - "error": false, - "format": "percentunit", - "gauge": { - "maxValue": 1, - "minValue": 0, - "show": true, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "id": 26, - "interval": null, - "links": [], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "min((node_filesystem_size{fstype=~\"xfs|ext4\",instance=~\"$server:.*\"} - node_filesystem_free{fstype=~\"xfs|ext4\",instance=~\"$server:.*\"} )/ node_filesystem_size{fstype=~\"xfs|ext4\",instance=~\"$server:.*\"})", - "hide": false, - "intervalFactor": 2, - "refId": "A", - "step": 1800 - } - ], - "thresholds": "0.75, 0.90", - "title": "Disk space", - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "current" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "rgba(50, 172, 45, 0.97)", - "rgba(237, 129, 40, 0.89)", - "rgba(245, 54, 54, 0.9)" - ], - "datasource": "automated-ds", - "decimals": 0, - "editable": true, - "error": false, - "format": "percent", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": true, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "id": 25, - "interval": null, - "links": [], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "((node_memory_MemTotal{instance=~\"$server:.*\"} - node_memory_MemAvailable{instance=~\"$server:.*\"}) / node_memory_MemTotal{instance=~\"$server:.*\"}) * 100", - "intervalFactor": 2, - "refId": "A", - "step": 1800 - } - ], - "thresholds": "70, 90", - "title": "Memory", - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "current" - }, - { - "content": "<img src=\"http://prometheus.io/assets/prometheus_logo_grey.svg\" alt=\"Prometheus logo\" style=\"height: 40px;\">\n<span style=\"font-family: 'Open Sans', 'Helvetica Neue', Helvetica; font-size: 25px;vertical-align: text-top;color: #bbbfc2;margin-left: 10px;\">Prometheus</span>\n\n<p style=\"margin-top: 10px;\">You're using Prometheus, an open-source systems monitoring and alerting toolkit originally built at SoundCloud. For more information, check out the <a href=\"http://www.grafana.org/\">Grafana</a> and <a href=\"http://prometheus.io/\">Prometheus</a> projects.</p>", - "editable": true, - "error": false, - "id": 9, - "links": [], - "mode": "html", - "span": 4, - "style": {}, - "title": "", - "transparent": true, - "type": "text" - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6" - }, - { - "collapse": false, - "height": 202, - "panels": [ - { - "aliasColors": { - "SENT": "#BF1B00" - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 1, - "grid": {}, - "id": 19, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 1, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 2, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(container_network_receive_bytes_total{id=\"/\"}[$interval])) by (id)", - "intervalFactor": 2, - "legendFormat": "RECEIVED", - "refId": "A", - "step": 600 - }, - { - "expr": "- sum(rate(container_network_transmit_bytes_total{id=\"/\"}[$interval])) by (id)", - "hide": false, - "intervalFactor": 2, - "legendFormat": "SENT", - "refId": "B", - "step": 600 - } - ], - "thresholds": [], - "timeFrom": null, - "timeShift": null, - "title": "Network Traffic", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "transparent": false, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": false, - "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 - } - ] - }, - { - "aliasColors": { - "{id=\"/\",instance=\"cadvisor:8080\",job=\"prometheus\"}": "#BA43A9" - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 1, - "grid": {}, - "id": 5, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 2, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(container_cpu_system_seconds_total[1m]))", - "hide": true, - "intervalFactor": 2, - "legendFormat": "a", - "refId": "B", - "step": 120 - }, - { - "expr": "sum(rate(container_cpu_system_seconds_total{name=~\".+\"}[1m]))", - "hide": true, - "interval": "", - "intervalFactor": 2, - "legendFormat": "nur container", - "refId": "F", - "step": 10 - }, - { - "expr": "sum(rate(container_cpu_system_seconds_total{id=\"/\"}[1m]))", - "hide": true, - "interval": "", - "intervalFactor": 2, - "legendFormat": "nur docker host", - "metric": "", - "refId": "A", - "step": 20 - }, - { - "expr": "sum(rate(process_cpu_seconds_total[$interval])) * 100", - "hide": false, - "interval": "", - "intervalFactor": 2, - "legendFormat": "host", - "metric": "", - "refId": "C", - "step": 600 - }, - { - "expr": "sum(rate(container_cpu_system_seconds_total{name=~\".+\"}[1m])) + sum(rate(container_cpu_system_seconds_total{id=\"/\"}[1m])) + sum(rate(process_cpu_seconds_total[1m]))", - "hide": true, - "intervalFactor": 2, - "legendFormat": "", - "refId": "D", - "step": 120 - } - ], - "thresholds": [], - "timeFrom": null, - "timeShift": null, - "title": "CPU Usage", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": false, - "values": [] - }, - "yaxes": [ - { - "format": "percent", - "label": "", - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "alert": { - "conditions": [ - { - "evaluator": { - "params": [ - 1.25 - ], - "type": "gt" - }, - "query": { - "params": [ - "A", - "5m", - "now" - ] - }, - "reducer": { - "params": [], - "type": "avg" - }, - "type": "query" - } - ], - "executionErrorState": "alerting", - "frequency": "60s", - "handler": 1, - "name": "Panel Title alert", - "noDataState": "keep_state", - "notifications": [ - { - "id": 1 - } - ] - }, - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "decimals": 0, - "editable": true, - "error": false, - "fill": 1, - "id": 28, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [], - "nullPointMode": "connected", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 2, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node_load1{instance=~\"$server:.*\"} / count by(job, instance)(count by(job, instance, cpu)(node_cpu{instance=~\"$server:.*\"}))", - "intervalFactor": 2, - "refId": "A", - "step": 600 - } - ], - "thresholds": [ - { - "colorMode": "critical", - "fill": true, - "line": true, - "op": "gt", - "value": 1.25 - } - ], - "timeFrom": null, - "timeShift": null, - "title": "Load", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": false, - "values": [] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": "1.50", - "min": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "alert": { - "conditions": [ - { - "evaluator": { - "params": [ - 850000000000 - ], - "type": "gt" - }, - "query": { - "params": [ - "A", - "5m", - "now" - ] - }, - "reducer": { - "params": [], - "type": "avg" - }, - "type": "query" - } - ], - "executionErrorState": "alerting", - "frequency": "60s", - "handler": 1, - "name": "Free/Used Disk Space alert", - "noDataState": "keep_state", - "notifications": [ - { - "id": 1 - } - ] - }, - "aliasColors": { - "Belegete Festplatte": "#BF1B00", - "Free Disk Space": "#7EB26D", - "Used Disk Space": "#7EB26D", - "{}": "#BF1B00" - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 1, - "grid": {}, - "id": 13, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - { - "alias": "Used Disk Space", - "yaxis": 1 - } - ], - "spaceLength": 10, - "span": 2, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "node_filesystem_size{fstype=\"aufs\"} - node_filesystem_free{fstype=\"aufs\"}", - "hide": false, - "intervalFactor": 2, - "legendFormat": "Used Disk Space", - "refId": "A", - "step": 600 - } - ], - "thresholds": [ - { - "colorMode": "critical", - "fill": true, - "line": true, - "op": "gt", - "value": 850000000000 - } - ], - "timeFrom": null, - "timeShift": null, - "title": "Used Disk Space", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": false, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "label": "", - "logBase": 1, - "max": 1000000000000, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "content": "#### Samples Ingested\nThis graph displays the count of samples ingested by the Prometheus server, as measured over the last 5 minutes, per time series in the range vector. When troubleshooting an issue on IRC or Github, this is often the first stat requested by the Prometheus team. ", - "editable": true, - "error": false, - "id": 9, - "links": [], - "mode": "markdown", - "span": 4, - "style": {}, - "title": "", - "transparent": true, - "type": "text" - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "New row", - "titleSize": "h6" - }, - { - "collapse": false, - "height": 238, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 5, - "grid": {}, - "id": 1, - "legend": { - "alignAsTable": true, - "avg": false, - "current": false, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 8, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(container_cpu_usage_seconds_total{name=~\".+\"}[$interval])) by (name) * 100", - "hide": false, - "interval": "", - "intervalFactor": 2, - "legendFormat": "{{name}}", - "metric": "", - "refId": "F", - "step": 240 - } - ], - "thresholds": [], - "timeFrom": null, - "timeShift": null, - "title": "CPU Usage per Container", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "percent", - "label": "", - "logBase": 1, - "max": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "aliasColors": { - "IN on /sda": "#7EB26D", - "OUT on /sda": "#890F02" - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 1, - "grid": {}, - "id": 3, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 2, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "-sum(rate(node_disk_bytes_read[$interval])) by (device)", - "hide": false, - "intervalFactor": 2, - "legendFormat": "OUT on /{{device}}", - "metric": "node_disk_bytes_read", - "refId": "A", - "step": 600 - }, - { - "expr": "sum(rate(node_disk_bytes_written[$interval])) by (device)", - "intervalFactor": 2, - "legendFormat": "IN on /{{device}}", - "metric": "", - "refId": "B", - "step": 600 - } - ], - "thresholds": [], - "timeFrom": null, - "timeShift": null, - "title": "Disk I/O", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": false, - "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 - } - ] - }, - { - "aliasColors": { - "Available Memory": "#7EB26D", - "Unavailable Memory": "#7EB26D" - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 1, - "grid": {}, - "id": 38, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 2, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "container_memory_rss{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "D", - "step": 20 - }, - { - "expr": "sum(container_memory_rss{name=~\".+\"})", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "A", - "step": 20 - }, - { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "B", - "step": 20 - }, - { - "expr": "container_memory_rss{id=\"/\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "C", - "step": 20 - }, - { - "expr": "sum(container_memory_rss)", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "E", - "step": 20 - }, - { - "expr": "node_memory_Buffers", - "hide": true, - "intervalFactor": 2, - "legendFormat": "node_memory_Dirty", - "refId": "N", - "step": 30 - }, - { - "expr": "node_memory_MemFree", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "F", - "step": 20 - }, - { - "expr": "node_memory_MemAvailable", - "hide": true, - "intervalFactor": 2, - "legendFormat": "Available Memory", - "refId": "H", - "step": 20 - }, - { - "expr": "node_memory_MemTotal - node_memory_MemAvailable", - "hide": false, - "intervalFactor": 2, - "legendFormat": "Unavailable Memory", - "refId": "G", - "step": 600 - }, - { - "expr": "node_memory_Inactive", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "I", - "step": 30 - }, - { - "expr": "node_memory_KernelStack", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "J", - "step": 30 - }, - { - "expr": "node_memory_Active", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "K", - "step": 30 - }, - { - "expr": "node_memory_MemTotal - (node_memory_Active + node_memory_MemFree + node_memory_Inactive)", - "hide": true, - "intervalFactor": 2, - "legendFormat": "Unknown", - "refId": "L", - "step": 40 - }, - { - "expr": "node_memory_MemFree + node_memory_Inactive ", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "M", - "step": 30 - }, - { - "expr": "container_memory_rss{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "O", - "step": 30 - }, - { - "expr": "node_memory_Inactive + node_memory_MemFree + node_memory_MemAvailable", - "hide": true, - "intervalFactor": 2, - "legendFormat": "", - "refId": "P", - "step": 40 - } - ], - "thresholds": [ - { - "colorMode": "critical", - "fill": true, - "line": true, - "op": "gt", - "value": 10000000000 - } - ], - "timeFrom": null, - "timeShift": null, - "title": "Available Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": false, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "label": "", - "logBase": 1, - "max": 16000000000, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Row", - "titleSize": "h6" - }, - { - "collapse": false, - "height": 251, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 1, - "grid": {}, - "id": 8, - "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 as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(container_network_receive_bytes_total{name=~\".+\"}[$interval])) by (name)", - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "A", - "step": 240 - }, - { - "expr": "- rate(container_network_transmit_bytes_total{name=~\".+\"}[$interval])", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "B", - "step": 10 - } - ], - "thresholds": [], - "timeFrom": null, - "timeShift": null, - "title": "Received Network Traffic per Container", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "transparent": false, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "Bps", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - } - ] - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 1, - "grid": {}, - "id": 9, - "legend": { - "alignAsTable": true, - "avg": false, - "current": false, - "hideEmpty": false, - "hideZero": false, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(container_network_transmit_bytes_total{name=~\".+\"}[$interval])) by (name)", - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "A", - "step": 240 - }, - { - "expr": "rate(container_network_transmit_bytes_total{id=\"/\"}[$interval])", - "hide": true, - "intervalFactor": 2, - "legendFormat": "", - "refId": "B", - "step": 10 - } - ], - "thresholds": [], - "timeFrom": null, - "timeShift": null, - "title": "Sent Network Traffic per Container", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "transparent": false, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "Bps", - "label": "", - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "short", - "label": "", - "logBase": 10, - "max": 8, - "min": 0, - "show": false - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "New row", - "titleSize": "h6" - }, - { - "collapse": false, - "height": 250, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 3, - "grid": {}, - "id": 10, - "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 as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 6, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "sum(container_memory_rss{name=~\".+\"}) by (name)", - "hide": false, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "A", - "step": 240 - }, - { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "B", - "step": 240 - } - ], - "thresholds": [], - "timeFrom": null, - "timeShift": null, - "title": "Memory Usage per Container", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "label": "", - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - } - ] - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 3, - "grid": {}, - "id": 34, - "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 as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "span": 6, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "sum(container_memory_swap{name=~\".+\"}) by (name)", - "hide": false, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "A", - "step": 240 - }, - { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "B", - "step": 240 - } - ], - "thresholds": [], - "timeFrom": null, - "timeShift": null, - "title": "Memory Swap per Container", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "label": "", - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6" - }, - { - "collapse": false, - "height": 97, - "panels": [ - { - "columns": [ - { - "text": "Current", - "value": "current" - } - ], - "editable": true, - "error": false, - "fontSize": "100%", - "id": 37, - "links": [], - "pageSize": null, - "scroll": true, - "showHeader": true, - "sort": { - "col": 0, - "desc": true - }, - "span": 4, - "styles": [ - { - "colorMode": null, - "colors": [ - "rgba(245, 54, 54, 0.9)", - "rgba(237, 129, 40, 0.89)", - "rgba(50, 172, 45, 0.97)" - ], - "decimals": 2, - "pattern": "/.*/", - "thresholds": [ - "10000000", - " 25000000" - ], - "type": "number", - "unit": "decbytes" - } - ], - "targets": [ - { - "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"} - container_memory_usage_bytes{name=~\".+\"}) by (name) ", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "metric": "", - "refId": "A", - "step": 240 - }, - { - "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"}) by (name) ", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "B", - "step": 240 - }, - { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": false, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "C", - "step": 240 - } - ], - "title": "Usage memory", - "transform": "timeseries_aggregations", - "type": "table" - }, - { - "columns": [ - { - "text": "Current", - "value": "current" - } - ], - "editable": true, - "error": false, - "fontSize": "100%", - "id": 35, - "links": [], - "pageSize": null, - "scroll": true, - "showHeader": true, - "sort": { - "col": 1, - "desc": true - }, - "span": 4, - "styles": [ - { - "colorMode": "cell", - "colors": [ - "rgba(50, 172, 45, 0.97)", - "rgba(237, 129, 40, 0.89)", - "rgba(245, 54, 54, 0.9)" - ], - "decimals": 2, - "pattern": "/.*/", - "thresholds": [ - "80", - "90" - ], - "type": "number", - "unit": "percent" - } - ], - "targets": [ - { - "expr": "sum(100 - ((container_spec_memory_limit_bytes{name=~\".+\"} - container_memory_usage_bytes{name=~\".+\"}) * 100 / container_spec_memory_limit_bytes{name=~\".+\"}) ) by (name) ", - "intervalFactor": 2, - "legendFormat": "{{name}}", - "metric": "", - "refId": "A", - "step": 240 - }, - { - "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"}) by (name) ", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "B", - "step": 240 - }, - { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "C", - "step": 240 - } - ], - "title": "Remaining memory", - "transform": "timeseries_aggregations", - "type": "table" - }, - { - "columns": [ - { - "text": "Current", - "value": "current" - } - ], - "editable": true, - "error": false, - "fontSize": "100%", - "id": 36, - "links": [], - "pageSize": null, - "scroll": true, - "showHeader": true, - "sort": { - "col": 0, - "desc": true - }, - "span": 4, - "styles": [ - { - "colorMode": null, - "colors": [ - "rgba(245, 54, 54, 0.9)", - "rgba(237, 129, 40, 0.89)", - "rgba(50, 172, 45, 0.97)" - ], - "decimals": 2, - "pattern": "/.*/", - "thresholds": [ - "10000000", - " 25000000" - ], - "type": "number", - "unit": "decbytes" - } - ], - "targets": [ - { - "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"} - container_memory_usage_bytes{name=~\".+\"}) by (name) ", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "metric": "", - "refId": "A", - "step": 240 - }, - { - "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"}) by (name) ", - "hide": false, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "B", - "step": 240 - }, - { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "C", - "step": 240 - } - ], - "title": "Limit memory", - "transform": "timeseries_aggregations", - "type": "table" - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6" - } - ], - "schemaVersion": 14, - "style": "dark", - "tags": [], - "templating": { - "list": [ - { - "allValue": ".+", - "current": { - "text": "All", - "value": "$__all" - }, - "datasource": "automated-ds", - "hide": 0, - "includeAll": true, - "label": "Container Group", - "multi": true, - "name": "containergroup", - "options": [], - "query": "label_values(container_group)", - "refresh": 1, - "regex": "", - "sort": 0, - "tagValuesQuery": null, - "tags": [], - "tagsQuery": null, - "type": "query", - "useTags": false - }, - { - "auto": true, - "auto_count": 50, - "auto_min": "50s", - "current": { - "text": "auto", - "value": "$__auto_interval" - }, - "datasource": null, - "hide": 0, - "includeAll": false, - "label": "Interval", - "multi": false, - "name": "interval", - "options": [ - { - "selected": true, - "text": "auto", - "value": "$__auto_interval" - }, - { - "selected": false, - "text": "30s", - "value": "30s" - }, - { - "selected": false, - "text": "1m", - "value": "1m" - }, - { - "selected": false, - "text": "2m", - "value": "2m" - }, - { - "selected": false, - "text": "3m", - "value": "3m" - }, - { - "selected": false, - "text": "5m", - "value": "5m" - }, - { - "selected": false, - "text": "7m", - "value": "7m" - }, - { - "selected": false, - "text": "10m", - "value": "10m" - }, - { - "selected": false, - "text": "30m", - "value": "30m" - }, - { - "selected": false, - "text": "1h", - "value": "1h" - }, - { - "selected": false, - "text": "6h", - "value": "6h" - }, - { - "selected": false, - "text": "12h", - "value": "12h" - }, - { - "selected": false, - "text": "1d", - "value": "1d" - }, - { - "selected": false, - "text": "7d", - "value": "7d" - }, - { - "selected": false, - "text": "14d", - "value": "14d" - }, - { - "selected": false, - "text": "30d", - "value": "30d" - } - ], - "query": "30s,1m,2m,3m,5m,7m,10m,30m,1h,6h,12h,1d,7d,14d,30d", - "refresh": 2, - "type": "interval" - }, - { - "allValue": null, - "current": { - "text": "192.168.121.2", - "value": "192.168.121.2" - }, - "datasource": "automated-ds", - "hide": 0, - "includeAll": false, - "label": "Node", - "multi": true, - "name": "server", - "options": [], - "query": "label_values(node_boot_time, instance)", - "refresh": 1, - "regex": "/([^:]+):.*/", - "sort": 0, - "tagValuesQuery": null, - "tags": [], - "tagsQuery": null, - "type": "query", - "useTags": false - } - ] - }, - "time": { - "from": "now-24h", - "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": "Monitoring docker with Prometheus", - "version": 6 -} diff --git a/monitor/dashboard/prototype_dashboard_collapsible.json b/monitor/dashboard/stats_overview.json index 58882361..440bb5a2 100644 --- a/monitor/dashboard/prototype_dashboard_collapsible.json +++ b/monitor/dashboard/stats_overview.json @@ -2,40 +2,41 @@ "annotations": { "list": [] }, - "description": "A simple overview of the most important Docker host and container metrics. (Barometer/Prometheus)", + "description": "A monitoring dashboard for OPNFV long duration test", "editable": true, "gnetId": 893, "graphTooltip": 1, "hideControls": false, "id": null, "links": [], - "refresh": "15m", + "refresh": false, "rows": [ { "collapse": false, - "height": 193, + "height": -295, "panels": [ { - "content": "<img src=\"http://prometheus.io/assets/prometheus_logo_grey.svg\" alt=\"Prometheus logo\" style=\"height: 40px;\">\n<span style=\"font-family: 'Open Sans', 'Helvetica Neue', Helvetica; font-size: 25px;vertical-align: text-top;color: #bbbfc2;margin-left: 10px;\">Prometheus</span>\n\n<p style=\"margin-top: 10px;\">You're using Prometheus, an open-source systems monitoring and alerting toolkit originally built at SoundCloud. For more information, check out the <a href=\"http://www.grafana.org/\">Grafana</a> and <a href=\"http://prometheus.io/\">Prometheus</a> projects.</p>", + "content": "<img src=\"https://wiki.opnfv.org/download/attachments/13212037/soak_monitoring.png\" alt=\"Monitoring Soak Test logo\" style=\"height: 66px;\">\n\n<p style=\"margin-top: 10px;\">\nYou're using the monitoring suite developed by OPNFV Bottlenecks team for OPNFV long duration test.\nThe monitoring suite consists of a monitoring dashboard - \n<a href=\"http://www.grafana.org/\">Grafana</a> ,\na monitoring data source -\n<a href=\"http://prometheus.io/\">Prometheus</a> and\nseveral monitoring tools -\n<a href=\"https://wiki.opnfv.org/display/fastpath/Barometer+Home/\">Barometer</a>,\n<a href=\"https://github.com/google/cadvisor/\">Cadvisor</a> and\n<a href=\"https://github.com/prometheus/node_exporter/\">Node</a>.\nThe monitoring suite is evolving along with OPNFV releases and enriched features will be included based on practical needs.\n</p>", "editable": true, "error": false, + "height": "100", "id": 91, "links": [], "mode": "html", - "span": 6, + "span": 7, "style": {}, "title": "", "transparent": true, "type": "text" }, { - "content": "#### Samples Ingested\nThis graph displays the count of samples ingested by the Prometheus server, as measured over the last 5 minutes, per time series in the range vector. When troubleshooting an issue on IRC or Github, this is often the first stat requested by the Prometheus team. ", + "content": "### OPNFV Long Duration Test\nThis is a collaborative initiative launched by OPNFV test working group (Testperf) for long duration stability test (soak test) in NFV realm.\nSeveral testing projects in OPNFV has been involved, i.e.,\n[Bottlenecks - system limit testing](https://wiki.opnfv.org/display/bottlenecks),\n[Functest - functional testing](https://wiki.opnfv.org/display/functest),\n[NFVbench -NFVI benchmark testing](https://wiki.opnfv.org/display/bottlenecks),\n[Storperf - storage testing](https://wiki.opnfv.org/display/bottlenecks),\n[VSperf - data path testing](https://wiki.opnfv.org/display/bottlenecks) and\n[Yardstick - performance testing](https://wiki.opnfv.org/display/bottlenecks).\nFor more detail, please refer to the [OPNFV LDT WIKI](https://wiki.opnfv.org/display/testing/Long+Duration+Testing).", "editable": true, "error": false, "id": 92, "links": [], "mode": "markdown", - "span": 6, + "span": 5, "style": {}, "title": "", "transparent": true, @@ -46,12 +47,12 @@ "repeatIteration": null, "repeatRowId": null, "showTitle": false, - "title": "Prometheus Information", - "titleSize": "h6" + "title": "Dashboard Information", + "titleSize": "h5" }, { - "collapse": true, - "height": 239, + "collapse": false, + "height": 177, "panels": [ { "cacheTimeout": null, @@ -103,7 +104,7 @@ "to": "null" } ], - "span": 3, + "span": 2, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, @@ -113,12 +114,13 @@ "tableColumn": "", "targets": [ { - "expr": "time() - node_boot_time{instance=~\"$server:.*\"}", + "expr": "time() - node_boot_time", + "format": "time_series", "hide": false, "intervalFactor": 2, "legendFormat": "", "refId": "A", - "step": 1800 + "step": 40 } ], "thresholds": "", @@ -182,7 +184,7 @@ "to": "null" } ], - "span": 3, + "span": 1, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, @@ -193,9 +195,10 @@ "targets": [ { "expr": "count(rate(container_last_seen{name=~\".+\"}[$interval]))", + "format": "time_series", "intervalFactor": 2, "refId": "A", - "step": 1800 + "step": 40 } ], "thresholds": "", @@ -260,7 +263,7 @@ "to": "null" } ], - "span": 3, + "span": 2, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, @@ -270,11 +273,12 @@ "tableColumn": "", "targets": [ { - "expr": "min((node_filesystem_size{fstype=~\"xfs|ext4\",instance=~\"$server:.*\"} - node_filesystem_free{fstype=~\"xfs|ext4\",instance=~\"$server:.*\"} )/ node_filesystem_size{fstype=~\"xfs|ext4\",instance=~\"$server:.*\"})", + "expr": "min((node_filesystem_size{fstype=~\"xfs|ext4\"} - node_filesystem_free{fstype=~\"xfs|ext4\"} )/ node_filesystem_size{fstype=~\"xfs|ext4\"})", + "format": "time_series", "hide": false, "intervalFactor": 2, "refId": "A", - "step": 1800 + "step": 40 } ], "thresholds": "0.75, 0.90", @@ -339,7 +343,7 @@ "to": "null" } ], - "span": 3, + "span": 2, "sparkline": { "fillColor": "rgba(31, 118, 189, 0.18)", "full": false, @@ -349,10 +353,11 @@ "tableColumn": "", "targets": [ { - "expr": "((node_memory_MemTotal{instance=~\"$server:.*\"} - node_memory_MemAvailable{instance=~\"$server:.*\"}) / node_memory_MemTotal{instance=~\"$server:.*\"}) * 100", + "expr": "((node_memory_MemTotal - node_memory_MemAvailable) / node_memory_MemTotal) * 100", + "format": "time_series", "intervalFactor": 2, "refId": "A", - "step": 1800 + "step": 40 } ], "thresholds": "70, 90", @@ -367,22 +372,10 @@ } ], "valueName": "current" - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Container, Disk, Memory Statistics", - "titleSize": "h6" - }, - { - "collapse": true, - "height": 218, - "panels": [ + }, { "aliasColors": { - "SENT": "#BF1B00" + "{id=\"/\",instance=\"cadvisor:8080\",job=\"prometheus\"}": "#BA43A9" }, "bars": false, "dashLength": 10, @@ -392,7 +385,7 @@ "error": false, "fill": 1, "grid": {}, - "id": 19, + "id": 5, "legend": { "avg": false, "current": false, @@ -407,42 +400,76 @@ "links": [], "nullPointMode": "null as zero", "percentage": false, - "pointradius": 1, + "pointradius": 5, "points": false, "renderer": "flot", "seriesOverrides": [], "spaceLength": 10, "span": 3, - "stack": false, + "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(rate(container_network_receive_bytes_total{id=\"/\"}[$interval])) by (id)", + "expr": "sum(rate(container_cpu_system_seconds_total[1m]))", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "a", + "refId": "B", + "step": 120 + }, + { + "expr": "sum(rate(container_cpu_system_seconds_total{name=~\".+\"}[1m]))", + "format": "time_series", + "hide": true, + "interval": "", "intervalFactor": 2, - "legendFormat": "RECEIVED", + "legendFormat": "nur container", + "refId": "F", + "step": 4 + }, + { + "expr": "sum(rate(container_cpu_system_seconds_total{id=\"/\"}[1m]))", + "format": "time_series", + "hide": true, + "interval": "", + "intervalFactor": 2, + "legendFormat": "nur docker host", + "metric": "", "refId": "A", - "step": 600 + "step": 4 }, { - "expr": "- sum(rate(container_network_transmit_bytes_total{id=\"/\"}[$interval])) by (id)", + "expr": "sum(rate(process_cpu_seconds_total[$interval])) * 100", + "format": "time_series", "hide": false, + "interval": "", "intervalFactor": 2, - "legendFormat": "SENT", - "refId": "B", - "step": 600 + "legendFormat": "", + "metric": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum(rate(container_cpu_system_seconds_total{name=~\".+\"}[1m])) + sum(rate(container_cpu_system_seconds_total{id=\"/\"}[1m])) + sum(rate(process_cpu_seconds_total[1m]))", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 4 } ], "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Network Traffic", + "title": "CPU Usage", "tooltip": { "msResolution": true, "shared": true, "sort": 0, "value_type": "cumulative" }, - "transparent": false, "type": "graph", "xaxis": { "buckets": null, @@ -453,8 +480,8 @@ }, "yaxes": [ { - "format": "bytes", - "label": null, + "format": "percent", + "label": "", "logBase": 1, "max": null, "min": null, @@ -472,7 +499,8 @@ }, { "aliasColors": { - "{id=\"/\",instance=\"cadvisor:8080\",job=\"prometheus\"}": "#BA43A9" + "Available Memory": "#7EB26D", + "Unavailable Memory": "#7EB26D" }, "bars": false, "dashLength": 10, @@ -482,7 +510,7 @@ "error": false, "fill": 1, "grid": {}, - "id": 5, + "id": 38, "legend": { "avg": false, "current": false, @@ -502,65 +530,172 @@ "renderer": "flot", "seriesOverrides": [], "spaceLength": 10, - "span": 3, + "span": 2, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(rate(container_cpu_system_seconds_total[1m]))", + "expr": "container_memory_rss{name=~\".+\"}", + "format": "time_series", "hide": true, "intervalFactor": 2, - "legendFormat": "a", + "legendFormat": "{{__name__}}", + "refId": "D", + "step": 20 + }, + { + "expr": "sum(container_memory_rss{name=~\".+\"})", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{__name__}}", + "refId": "A", + "step": 20 + }, + { + "expr": "container_memory_usage_bytes{name=~\".+\"}", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{name}}", "refId": "B", - "step": 120 + "step": 20 }, { - "expr": "sum(rate(container_cpu_system_seconds_total{name=~\".+\"}[1m]))", + "expr": "container_memory_rss{id=\"/\"}", + "format": "time_series", "hide": true, - "interval": "", "intervalFactor": 2, - "legendFormat": "nur container", - "refId": "F", - "step": 10 + "legendFormat": "{{__name__}}", + "refId": "C", + "step": 20 }, { - "expr": "sum(rate(container_cpu_system_seconds_total{id=\"/\"}[1m]))", + "expr": "sum(container_memory_rss)", + "format": "time_series", "hide": true, - "interval": "", "intervalFactor": 2, - "legendFormat": "nur docker host", - "metric": "", - "refId": "A", + "legendFormat": "{{__name__}}", + "refId": "E", "step": 20 }, { - "expr": "sum(rate(process_cpu_seconds_total[$interval])) * 100", + "expr": "node_memory_Buffers", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "node_memory_Dirty", + "refId": "N", + "step": 30 + }, + { + "expr": "node_memory_MemFree", + "format": "time_series", "hide": false, - "interval": "", "intervalFactor": 2, - "legendFormat": "host", - "metric": "", - "refId": "C", - "step": 600 + "legendFormat": "Free Memory", + "refId": "F", + "step": 20 }, { - "expr": "sum(rate(container_cpu_system_seconds_total{name=~\".+\"}[1m])) + sum(rate(container_cpu_system_seconds_total{id=\"/\"}[1m])) + sum(rate(process_cpu_seconds_total[1m]))", + "expr": "node_memory_MemAvailable", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "Available Memory", + "refId": "H", + "step": 20 + }, + { + "expr": "node_memory_MemTotal - node_memory_MemAvailable", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "Unavailable Memory", + "refId": "G", + "step": 20 + }, + { + "expr": "node_memory_Inactive", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{__name__}}", + "refId": "I", + "step": 2 + }, + { + "expr": "node_memory_KernelStack", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{__name__}}", + "refId": "J", + "step": 30 + }, + { + "expr": "node_memory_Active", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{__name__}}", + "refId": "K", + "step": 30 + }, + { + "expr": "node_memory_MemTotal - (node_memory_Active + node_memory_MemFree + node_memory_Inactive)", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "Unknown", + "refId": "L", + "step": 40 + }, + { + "expr": "node_memory_MemFree + node_memory_Inactive ", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{__name__}}", + "refId": "M", + "step": 30 + }, + { + "expr": "container_memory_rss{name=~\".+\"}", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{__name__}}", + "refId": "O", + "step": 30 + }, + { + "expr": "node_memory_Inactive + node_memory_MemFree + node_memory_MemAvailable", + "format": "time_series", "hide": true, "intervalFactor": 2, "legendFormat": "", - "refId": "D", - "step": 120 + "refId": "P", + "step": 40 + } + ], + "thresholds": [ + { + "colorMode": "critical", + "fill": true, + "line": true, + "op": "gt", + "value": 10000000000 } ], - "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "CPU Usage", + "title": "Available Memory", "tooltip": { "msResolution": true, "shared": true, "sort": 0, - "value_type": "cumulative" + "value_type": "individual" }, "type": "graph", "xaxis": { @@ -572,11 +707,11 @@ }, "yaxes": [ { - "format": "percent", + "format": "bytes", "label": "", "logBase": 1, "max": null, - "min": null, + "min": 0, "show": true }, { @@ -590,6 +725,32 @@ ] }, { + "headings": true, + "id": 94, + "limit": 10, + "links": [], + "query": "", + "recent": true, + "search": false, + "span": 12, + "starred": true, + "tags": [], + "title": "Status Quick Access", + "type": "dashlist" + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Nodes Overview - CPU, Network, Storage, Container Count", + "titleSize": "h5" + }, + { + "collapse": true, + "height": 265, + "panels": [ + { "alert": { "conditions": [ { @@ -635,10 +796,12 @@ "fill": 1, "id": 28, "legend": { + "alignAsTable": false, "avg": false, "current": false, "max": false, "min": false, + "rightSide": false, "show": false, "total": false, "values": false @@ -653,15 +816,26 @@ "renderer": "flot", "seriesOverrides": [], "spaceLength": 10, - "span": 3, + "span": 4, "stack": false, "steppedLine": false, "targets": [ { - "expr": "node_load1{instance=~\"$server:.*\"} / count by(job, instance)(count by(job, instance, cpu)(node_cpu{instance=~\"$server:.*\"}))", + "expr": "node_load1 / count by(job, instance)(count by(job, instance, cpu)(node_cpu))", + "format": "time_series", + "hide": false, "intervalFactor": 2, + "legendFormat": "", "refId": "A", - "step": 600 + "step": 10 + }, + { + "expr": "count by(exported_instance, job)(collectd_load_0{exported_instance=~\"host.\"}) / (count by(job, exported_instance)(count by(exported_instance, job, cpu)(collectd_cpu{exported_instance=~\"host.\"})))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 } ], "thresholds": [ @@ -675,7 +849,7 @@ ], "timeFrom": null, "timeShift": null, - "title": "Load", + "title": "CPU Loads on Nodes", "tooltip": { "msResolution": false, "shared": true, @@ -695,7 +869,7 @@ "format": "percentunit", "label": null, "logBase": 1, - "max": "1.50", + "max": "1", "min": null, "show": true }, @@ -710,45 +884,8 @@ ] }, { - "alert": { - "conditions": [ - { - "evaluator": { - "params": [ - 850000000000 - ], - "type": "gt" - }, - "query": { - "params": [ - "A", - "5m", - "now" - ] - }, - "reducer": { - "params": [], - "type": "avg" - }, - "type": "query" - } - ], - "executionErrorState": "alerting", - "frequency": "60s", - "handler": 1, - "name": "Free/Used Disk Space alert", - "noDataState": "keep_state", - "notifications": [ - { - "id": 1 - } - ] - }, "aliasColors": { - "Belegete Festplatte": "#BF1B00", - "Free Disk Space": "#7EB26D", - "Used Disk Space": "#7EB26D", - "{}": "#BF1B00" + "SENT": "#BF1B00" }, "bars": false, "dashLength": 10, @@ -758,7 +895,7 @@ "error": false, "fill": 1, "grid": {}, - "id": 13, + "id": 19, "legend": { "avg": false, "current": false, @@ -773,46 +910,145 @@ "links": [], "nullPointMode": "null as zero", "percentage": false, - "pointradius": 5, + "pointradius": 1, "points": false, "renderer": "flot", - "seriesOverrides": [ + "seriesOverrides": [], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ { - "alias": "Used Disk Space", - "yaxis": 1 + "expr": "sum(rate(container_network_transmit_bytes_total[$interval])) by (instance) - sum(rate(container_network_receive_bytes_total[$interval])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "(SENT - RECEIVED) {{instance}}", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum(rate(container_network_transmit_bytes_total[$interval])) by (instance)", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "SENT - {{instance}}", + "refId": "B", + "step": 10 + }, + { + "expr": "sum(rate(container_network_receive_bytes_total[$interval])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "RECEIVED - {{instance}}", + "refId": "C", + "step": 10 } ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Network Loads on Nodes", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "transparent": false, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": false, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + "IN on /sda": "#7EB26D", + "OUT on /sda": "#890F02" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "automated-ds", + "editable": true, + "error": false, + "fill": 1, + "grid": {}, + "id": 3, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], "spaceLength": 10, - "span": 3, - "stack": true, + "span": 4, + "stack": false, "steppedLine": false, "targets": [ { - "expr": "node_filesystem_size{fstype=\"aufs\"} - node_filesystem_free{fstype=\"aufs\"}", + "expr": "-sum(rate(node_disk_bytes_read[$interval])) by (device)", + "format": "time_series", "hide": false, "intervalFactor": 2, - "legendFormat": "Used Disk Space", + "legendFormat": "OUT on /{{device}}", + "metric": "node_disk_bytes_read", "refId": "A", - "step": 600 - } - ], - "thresholds": [ + "step": 10 + }, { - "colorMode": "critical", - "fill": true, - "line": true, - "op": "gt", - "value": 850000000000 + "expr": "sum(rate(node_disk_bytes_written[$interval])) by (device)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "IN on /{{device}}", + "metric": "", + "refId": "B", + "step": 10 } ], + "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Used Disk Space", + "title": "Disk I/O", "tooltip": { "msResolution": true, "shared": true, "sort": 0, - "value_type": "individual" + "value_type": "cumulative" }, "type": "graph", "xaxis": { @@ -824,11 +1060,11 @@ }, "yaxes": [ { - "format": "bytes", - "label": "", + "format": "Bps", + "label": null, "logBase": 1, - "max": 1000000000000, - "min": 0, + "max": null, + "min": null, "show": true }, { @@ -845,13 +1081,13 @@ "repeat": null, "repeatIteration": null, "repeatRowId": null, - "showTitle": false, - "title": "Network, CPU, Load Statistics", - "titleSize": "h6" + "showTitle": true, + "title": "Loads on Nodes - CPU, Network, Disk Loads Records", + "titleSize": "h5" }, { "collapse": true, - "height": 249, + "height": 281, "panels": [ { "aliasColors": {}, @@ -861,9 +1097,9 @@ "datasource": "automated-ds", "editable": true, "error": false, - "fill": 5, + "fill": 1, "grid": {}, - "id": 1, + "id": 8, "legend": { "alignAsTable": true, "avg": false, @@ -876,7 +1112,7 @@ "values": false }, "lines": true, - "linewidth": 1, + "linewidth": 2, "links": [], "nullPointMode": "null as zero", "percentage": false, @@ -885,31 +1121,37 @@ "renderer": "flot", "seriesOverrides": [], "spaceLength": 10, - "span": 8, - "stack": true, + "span": 6, + "stack": false, "steppedLine": false, "targets": [ { - "expr": "sum(rate(container_cpu_usage_seconds_total{name=~\".+\"}[$interval])) by (name) * 100", - "hide": false, - "interval": "", + "expr": "sum(rate(container_network_receive_bytes_total{name=~\".+\"}[$interval])) by (name)", "intervalFactor": 2, "legendFormat": "{{name}}", - "metric": "", - "refId": "F", - "step": 240 + "refId": "A", + "step": 4 + }, + { + "expr": "- rate(container_network_transmit_bytes_total{name=~\".+\"}[$interval])", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{name}}", + "refId": "B", + "step": 10 } ], "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "CPU Usage per Container", + "title": "Received Network Traffic per Container", "tooltip": { "msResolution": true, "shared": true, "sort": 0, - "value_type": "individual" + "value_type": "cumulative" }, + "transparent": false, "type": "graph", "xaxis": { "buckets": null, @@ -920,10 +1162,11 @@ }, "yaxes": [ { - "format": "percent", - "label": "", + "format": "Bps", + "label": null, "logBase": 1, "max": null, + "min": null, "show": true }, { @@ -932,15 +1175,12 @@ "logBase": 1, "max": null, "min": null, - "show": false + "show": true } ] }, { - "aliasColors": { - "IN on /sda": "#7EB26D", - "OUT on /sda": "#890F02" - }, + "aliasColors": {}, "bars": false, "dashLength": 10, "dashes": false, @@ -949,18 +1189,22 @@ "error": false, "fill": 1, "grid": {}, - "id": 3, + "id": 9, "legend": { + "alignAsTable": true, "avg": false, "current": false, + "hideEmpty": false, + "hideZero": false, "max": false, "min": false, - "show": false, + "rightSide": true, + "show": true, "total": false, "values": false }, "lines": true, - "linewidth": 1, + "linewidth": 2, "links": [], "nullPointMode": "null as zero", "percentage": false, @@ -969,50 +1213,51 @@ "renderer": "flot", "seriesOverrides": [], "spaceLength": 10, - "span": 2, + "span": 6, "stack": false, "steppedLine": false, "targets": [ { - "expr": "-sum(rate(node_disk_bytes_read[$interval])) by (device)", - "hide": false, + "expr": "sum(rate(container_network_transmit_bytes_total{name=~\".+\"}[$interval])) by (name)", + "format": "time_series", "intervalFactor": 2, - "legendFormat": "OUT on /{{device}}", - "metric": "node_disk_bytes_read", + "legendFormat": "{{name}}", "refId": "A", - "step": 600 + "step": 4 }, { - "expr": "sum(rate(node_disk_bytes_written[$interval])) by (device)", + "expr": "rate(container_network_transmit_bytes_total{id=\"/\"}[$interval])", + "format": "time_series", + "hide": true, "intervalFactor": 2, - "legendFormat": "IN on /{{device}}", - "metric": "", + "legendFormat": "", "refId": "B", - "step": 600 + "step": 10 } ], "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Disk I/O", + "title": "Sent Network Traffic per Container", "tooltip": { "msResolution": true, "shared": true, "sort": 0, "value_type": "cumulative" }, + "transparent": false, "type": "graph", "xaxis": { "buckets": null, "mode": "time", "name": null, - "show": false, + "show": true, "values": [] }, "yaxes": [ { "format": "Bps", - "label": null, + "label": "", "logBase": 1, "max": null, "min": null, @@ -1020,18 +1265,54 @@ }, { "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, + "label": "", + "logBase": 10, + "max": 8, + "min": 0, "show": false } ] }, { + "alert": { + "conditions": [ + { + "evaluator": { + "params": [ + 850000000000 + ], + "type": "gt" + }, + "query": { + "params": [ + "A", + "5m", + "now" + ] + }, + "reducer": { + "params": [], + "type": "avg" + }, + "type": "query" + } + ], + "executionErrorState": "alerting", + "frequency": "60s", + "handler": 1, + "name": "Free/Used Disk Space alert", + "noDataState": "keep_state", + "notifications": [ + { + "id": 1 + } + ] + }, "aliasColors": { - "Available Memory": "#7EB26D", - "Unavailable Memory": "#7EB26D" + "Belegete Festplatte": "#BF1B00", + "Free Disk Space": "#7EB26D", + "Used Disk Space": "#7EB26D", + "{}": "#BF1B00" }, "bars": false, "dashLength": 10, @@ -1041,7 +1322,7 @@ "error": false, "fill": 1, "grid": {}, - "id": 38, + "id": 13, "legend": { "avg": false, "current": false, @@ -1059,139 +1340,25 @@ "pointradius": 5, "points": false, "renderer": "flot", - "seriesOverrides": [], + "seriesOverrides": [ + { + "alias": "Used Disk Space", + "yaxis": 1 + } + ], "spaceLength": 10, - "span": 2, + "span": 3, "stack": true, "steppedLine": false, "targets": [ { - "expr": "container_memory_rss{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "D", - "step": 20 - }, - { - "expr": "sum(container_memory_rss{name=~\".+\"})", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "A", - "step": 20 - }, - { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "B", - "step": 20 - }, - { - "expr": "container_memory_rss{id=\"/\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "C", - "step": 20 - }, - { - "expr": "sum(container_memory_rss)", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "E", - "step": 20 - }, - { - "expr": "node_memory_Buffers", - "hide": true, - "intervalFactor": 2, - "legendFormat": "node_memory_Dirty", - "refId": "N", - "step": 30 - }, - { - "expr": "node_memory_MemFree", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "F", - "step": 20 - }, - { - "expr": "node_memory_MemAvailable", - "hide": true, - "intervalFactor": 2, - "legendFormat": "Available Memory", - "refId": "H", - "step": 20 - }, - { - "expr": "node_memory_MemTotal - node_memory_MemAvailable", + "expr": "node_filesystem_size{fstype=\"aufs\"} - node_filesystem_free{fstype=\"aufs\"}", + "format": "time_series", "hide": false, "intervalFactor": 2, - "legendFormat": "Unavailable Memory", - "refId": "G", - "step": 600 - }, - { - "expr": "node_memory_Inactive", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "I", - "step": 30 - }, - { - "expr": "node_memory_KernelStack", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "J", - "step": 30 - }, - { - "expr": "node_memory_Active", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "K", - "step": 30 - }, - { - "expr": "node_memory_MemTotal - (node_memory_Active + node_memory_MemFree + node_memory_Inactive)", - "hide": true, - "intervalFactor": 2, - "legendFormat": "Unknown", - "refId": "L", - "step": 40 - }, - { - "expr": "node_memory_MemFree + node_memory_Inactive ", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "M", - "step": 30 - }, - { - "expr": "container_memory_rss{name=~\".+\"}", - "hide": true, - "intervalFactor": 2, - "legendFormat": "{{__name__}}", - "refId": "O", - "step": 30 - }, - { - "expr": "node_memory_Inactive + node_memory_MemFree + node_memory_MemAvailable", - "hide": true, - "intervalFactor": 2, "legendFormat": "", - "refId": "P", - "step": 40 + "refId": "A", + "step": 10 } ], "thresholds": [ @@ -1200,12 +1367,12 @@ "fill": true, "line": true, "op": "gt", - "value": 10000000000 + "value": 850000000000 } ], "timeFrom": null, "timeShift": null, - "title": "Available Memory", + "title": "Used Disk Space", "tooltip": { "msResolution": true, "shared": true, @@ -1225,7 +1392,7 @@ "format": "bytes", "label": "", "logBase": 1, - "max": 16000000000, + "max": 1000000000000, "min": 0, "show": true }, @@ -1238,19 +1405,7 @@ "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Container-wise CPU usage, Disk I/O, Memory Available", - "titleSize": "h6" - }, - { - "collapse": true, - "height": 251, - "panels": [ + }, { "aliasColors": {}, "bars": false, @@ -1259,9 +1414,9 @@ "datasource": "automated-ds", "editable": true, "error": false, - "fill": 1, + "fill": 5, "grid": {}, - "id": 8, + "id": 1, "legend": { "alignAsTable": true, "avg": false, @@ -1274,7 +1429,7 @@ "values": false }, "lines": true, - "linewidth": 2, + "linewidth": 1, "links": [], "nullPointMode": "null as zero", "percentage": false, @@ -1283,37 +1438,32 @@ "renderer": "flot", "seriesOverrides": [], "spaceLength": 10, - "span": 6, - "stack": false, + "span": 9, + "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(rate(container_network_receive_bytes_total{name=~\".+\"}[$interval])) by (name)", - "intervalFactor": 2, - "legendFormat": "{{name}}", - "refId": "A", - "step": 240 - }, - { - "expr": "- rate(container_network_transmit_bytes_total{name=~\".+\"}[$interval])", - "hide": true, + "expr": "sum(rate(container_cpu_usage_seconds_total{name=~\".+\"}[$interval])) by (name) * 100", + "format": "time_series", + "hide": false, + "interval": "", "intervalFactor": 2, "legendFormat": "{{name}}", - "refId": "B", - "step": 10 + "metric": "", + "refId": "F", + "step": 4 } ], "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Received Network Traffic per Container", + "title": "CPU Usage per Container", "tooltip": { "msResolution": true, "shared": true, "sort": 0, - "value_type": "cumulative" + "value_type": "individual" }, - "transparent": false, "type": "graph", "xaxis": { "buckets": null, @@ -1324,11 +1474,10 @@ }, "yaxes": [ { - "format": "Bps", - "label": null, + "format": "percent", + "label": "", "logBase": 1, "max": null, - "min": null, "show": true }, { @@ -1337,7 +1486,7 @@ "logBase": 1, "max": null, "min": null, - "show": true + "show": false } ] }, @@ -1349,15 +1498,13 @@ "datasource": "automated-ds", "editable": true, "error": false, - "fill": 1, + "fill": 3, "grid": {}, - "id": 9, + "id": 10, "legend": { "alignAsTable": true, "avg": false, "current": false, - "hideEmpty": false, - "hideZero": false, "max": false, "min": false, "rightSide": true, @@ -1376,36 +1523,36 @@ "seriesOverrides": [], "spaceLength": 10, "span": 6, - "stack": false, + "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(rate(container_network_transmit_bytes_total{name=~\".+\"}[$interval])) by (name)", + "expr": "sum(container_memory_rss{name=~\".+\"}) by (name)", + "hide": false, "intervalFactor": 2, "legendFormat": "{{name}}", "refId": "A", - "step": 240 + "step": 4 }, { - "expr": "rate(container_network_transmit_bytes_total{id=\"/\"}[$interval])", + "expr": "container_memory_usage_bytes{name=~\".+\"}", "hide": true, "intervalFactor": 2, - "legendFormat": "", + "legendFormat": "{{name}}", "refId": "B", - "step": 10 + "step": 240 } ], "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Sent Network Traffic per Container", + "title": "Memory Usage per Container", "tooltip": { "msResolution": true, "shared": true, "sort": 0, - "value_type": "cumulative" + "value_type": "individual" }, - "transparent": false, "type": "graph", "xaxis": { "buckets": null, @@ -1416,7 +1563,7 @@ }, "yaxes": [ { - "format": "Bps", + "format": "bytes", "label": "", "logBase": 1, "max": null, @@ -1425,26 +1572,14 @@ }, { "format": "short", - "label": "", - "logBase": 10, - "max": 8, - "min": 0, - "show": false + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Network Traffic", - "titleSize": "h6" - }, - { - "collapse": true, - "height": 250, - "panels": [ + }, { "aliasColors": {}, "bars": false, @@ -1455,7 +1590,7 @@ "error": false, "fill": 3, "grid": {}, - "id": 10, + "id": 34, "legend": { "alignAsTable": true, "avg": false, @@ -1482,15 +1617,17 @@ "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_rss{name=~\".+\"}) by (name)", + "expr": "sum(container_memory_swap{name=~\".+\"}) by (name)", + "format": "time_series", "hide": false, "intervalFactor": 2, "legendFormat": "{{name}}", "refId": "A", - "step": 240 + "step": 4 }, { "expr": "container_memory_usage_bytes{name=~\".+\"}", + "format": "time_series", "hide": true, "intervalFactor": 2, "legendFormat": "{{name}}", @@ -1501,7 +1638,7 @@ "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Memory Usage per Container", + "title": "Memory Swap per Container", "tooltip": { "msResolution": true, "shared": true, @@ -1541,11 +1678,8 @@ "dashLength": 10, "dashes": false, "datasource": "automated-ds", - "editable": true, - "error": false, - "fill": 3, - "grid": {}, - "id": 34, + "fill": 1, + "id": 95, "legend": { "alignAsTable": true, "avg": false, @@ -1558,9 +1692,9 @@ "values": false }, "lines": true, - "linewidth": 2, + "linewidth": 1, "links": [], - "nullPointMode": "null as zero", + "nullPointMode": "null", "percentage": false, "pointradius": 5, "points": false, @@ -1568,32 +1702,40 @@ "seriesOverrides": [], "spaceLength": 10, "span": 6, - "stack": true, + "stack": false, "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_swap{name=~\".+\"}) by (name)", - "hide": false, + "expr": "count by(device)(container_fs_io_current)", + "format": "time_series", + "hide": true, "intervalFactor": 2, - "legendFormat": "{{name}}", "refId": "A", - "step": 240 + "step": 120 }, { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": true, + "expr": "container_fs_read_seconds_total{name=~\".+\"}", + "format": "time_series", + "hide": false, "intervalFactor": 2, - "legendFormat": "{{name}}", + "legendFormat": "Read - {{name}}", "refId": "B", - "step": 240 + "step": 4 + }, + { + "expr": "-container_fs_write_seconds_total{name=~\".+\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Write - {{name}}", + "refId": "C", + "step": 4 } ], "thresholds": [], "timeFrom": null, "timeShift": null, - "title": "Memory Swap per Container", + "title": "Container IO Seconds Total", "tooltip": { - "msResolution": true, "shared": true, "sort": 0, "value_type": "individual" @@ -1608,8 +1750,8 @@ }, "yaxes": [ { - "format": "bytes", - "label": "", + "format": "short", + "label": null, "logBase": 1, "max": null, "min": null, @@ -1624,166 +1766,205 @@ "show": true } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Container Memory Statistics", - "titleSize": "h6" - }, - { - "collapse": true, - "height": 97, - "panels": [ + }, { - "columns": [ - { - "text": "Current", - "value": "current" - } - ], - "editable": true, - "error": false, - "fontSize": "100%", - "id": 37, - "links": [], - "pageSize": null, - "scroll": true, - "showHeader": true, - "sort": { - "col": 0, - "desc": true + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "automated-ds", + "fill": 1, + "id": 96, + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false }, - "span": 4, - "styles": [ - { - "colorMode": null, - "colors": [ - "rgba(245, 54, 54, 0.9)", - "rgba(237, 129, 40, 0.89)", - "rgba(50, 172, 45, 0.97)" - ], - "decimals": 2, - "pattern": "/.*/", - "thresholds": [ - "10000000", - " 25000000" - ], - "type": "number", - "unit": "decbytes" - } - ], + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, "targets": [ { - "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"} - container_memory_usage_bytes{name=~\".+\"}) by (name) ", - "hide": true, + "expr": "container_fs_io_current{name=~\".+\"}", + "format": "time_series", + "hide": false, + "interval": "", "intervalFactor": 2, "legendFormat": "{{name}}", - "metric": "", "refId": "A", - "step": 240 + "step": 4 }, { - "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"}) by (name) ", + "expr": "container_fs_read_seconds_total{name=~\".+\"}", + "format": "time_series", "hide": true, "intervalFactor": 2, - "legendFormat": "{{name}}", + "legendFormat": "Read - {{name}}", "refId": "B", "step": 240 }, { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": false, + "expr": "container_fs_write_seconds_total{name=~\".+\"}", + "format": "time_series", + "hide": true, "intervalFactor": 2, - "legendFormat": "{{name}}", + "legendFormat": "Write - {{name}}", "refId": "C", "step": 240 } ], - "title": "Usage memory", - "transform": "timeseries_aggregations", - "type": "table" - }, - { - "columns": [ + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Container IO Current", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ { - "text": "Current", - "value": "current" + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true } - ], + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Containers on Nodes - CPU, Traffic, Memory Usage", + "titleSize": "h5" + }, + { + "collapse": true, + "height": null, + "panels": [ + { + "columns": [], + "datasource": "automated-ds", "editable": true, "error": false, + "filterNull": false, "fontSize": "100%", - "id": 35, + "height": "400", + "hideTimeOverride": false, + "id": 93, "links": [], - "pageSize": null, + "pageSize": 100, + "repeat": null, "scroll": true, "showHeader": true, "sort": { - "col": 1, - "desc": true + "col": 0, + "desc": false }, - "span": 4, + "span": 9, "styles": [ { - "colorMode": "cell", + "colorMode": null, "colors": [ - "rgba(50, 172, 45, 0.97)", + "rgba(245, 54, 54, 0.9)", "rgba(237, 129, 40, 0.89)", - "rgba(245, 54, 54, 0.9)" + "rgba(50, 172, 45, 0.97)" ], "decimals": 2, "pattern": "/.*/", - "thresholds": [ - "80", - "90" - ], + "thresholds": [], "type": "number", - "unit": "percent" + "unit": "short" } ], "targets": [ { - "expr": "sum(100 - ((container_spec_memory_limit_bytes{name=~\".+\"} - container_memory_usage_bytes{name=~\".+\"}) * 100 / container_spec_memory_limit_bytes{name=~\".+\"}) ) by (name) ", + "expr": "cadvisor_version_info", + "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{name}}", - "metric": "", + "legendFormat": "{{instance}} - cAdvisor Version: {{cadvisorVersion}} - Docker Version: {{dockerVersion}} - OS Version: {{osVersion}} - Host Kernel Version: {{kernelVersion}}", "refId": "A", - "step": 240 + "step": 4 }, { - "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"}) by (name) ", - "hide": true, + "expr": "prometheus_build_info", + "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{name}}", + "legendFormat": "{{instance}} - Prometheus Version: {{version}} - Go Version: {{goversion}} - Prometheus Branch: {{branch}}", "refId": "B", - "step": 240 + "step": 4 }, { - "expr": "container_memory_usage_bytes{name=~\".+\"}", - "hide": true, + "expr": "node_exporter_build_info", + "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{name}}", + "legendFormat": "{{instance}} - Node-Exporter Version: {{version}} - Go Version: {{goversion}} - Node Exporter Branch: {{branch}}", "refId": "C", - "step": 240 + "step": 4 + }, + { + "expr": "collectd_exporter_build_info", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "{{instance}} - Collectd Exporter Version: {{version}} - Go Version: {{goversion}} - Collectd Exporter Branch: {{branch}} ", + "refId": "G", + "step": 4 } ], - "title": "Remaining memory", + "timeFrom": null, + "timeShift": null, + "title": "Versions", "transform": "timeseries_aggregations", "type": "table" }, { "columns": [ { - "text": "Current", - "value": "current" + "text": "Avg", + "value": "avg" } ], + "datasource": "automated-ds", "editable": true, "error": false, "fontSize": "100%", + "height": "400", "id": 36, "links": [], "pageSize": null, @@ -1793,7 +1974,7 @@ "col": 0, "desc": true }, - "span": 4, + "span": 3, "styles": [ { "colorMode": null, @@ -1815,6 +1996,7 @@ "targets": [ { "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"} - container_memory_usage_bytes{name=~\".+\"}) by (name) ", + "format": "time_series", "hide": true, "intervalFactor": 2, "legendFormat": "{{name}}", @@ -1824,14 +2006,16 @@ }, { "expr": "sum(container_spec_memory_limit_bytes{name=~\".+\"}) by (name) ", + "format": "time_series", "hide": false, "intervalFactor": 2, "legendFormat": "{{name}}", "refId": "B", - "step": 240 + "step": 10 }, { "expr": "container_memory_usage_bytes{name=~\".+\"}", + "format": "time_series", "hide": true, "intervalFactor": 2, "legendFormat": "{{name}}", @@ -1847,9 +2031,42 @@ "repeat": null, "repeatIteration": null, "repeatRowId": null, - "showTitle": false, - "title": "Memory Usage", - "titleSize": "h6" + "showTitle": true, + "title": "Versions - Host OS, Docker, APP, Build Information", + "titleSize": "h5" + }, + { + "collapse": true, + "height": 250, + "panels": [], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "OVS Bridges on Nodes - TODO", + "titleSize": "h5" + }, + { + "collapse": true, + "height": 250, + "panels": [], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "VMs on Nodes - TODO", + "titleSize": "h5" + }, + { + "collapse": true, + "height": 250, + "panels": [], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Traffic Generators Status - TODO", + "titleSize": "h5" } ], "schemaVersion": 14, @@ -1983,8 +2200,8 @@ { "allValue": null, "current": { - "text": "192.168.114.2", - "value": "192.168.114.2" + "text": null, + "value": null }, "datasource": "automated-ds", "hide": 0, @@ -2006,7 +2223,7 @@ ] }, "time": { - "from": "now-24h", + "from": "now-30m", "to": "now" }, "timepicker": { @@ -2035,6 +2252,6 @@ ] }, "timezone": "browser", - "title": "Collapsibe System Monitor using Prometheus", - "version": 3 -} + "title": "Stats Overview", + "version": 2 +}
\ No newline at end of file diff --git a/monitor/monitoring.sh b/monitor/monitoring.sh index b42126b7..524958a6 100644 --- a/monitor/monitoring.sh +++ b/monitor/monitoring.sh @@ -15,16 +15,16 @@ DISPATCH="/home/opnfv/bottlenecks/monitor/dispatch" # INSTALL GRAFANA + PROMETHEUS + CADVISOR + BAROMETER on the JUMPERSERVER # # Node-Exporter -# sudo docker run --name bottlenecks-node-exporter \ -# -d -p 9100:9100 \ -# -v "/proc:/host/proc:ro" \ -# -v "/sys:/host/sys:ro" \ -# -v "/:/rootfs:ro" \ -# --net="host" \ -# quay.io/prometheus/node-exporter:v0.14.0 \ -# -collector.procfs /host/proc \ -# -collector.sysfs /host/sys \ -# -collector.filesystem.ignored-mount-points "^/(sys|proc|dev|host|etc)($|/)" +sudo docker run --name bottlenecks-node-exporter \ + -d -p 9100:9100 \ + -v "/proc:/host/proc:ro" \ + -v "/sys:/host/sys:ro" \ + -v "/:/rootfs:ro" \ + --net="host" \ + quay.io/prometheus/node-exporter:v0.14.0 \ + -collector.procfs /host/proc \ + -collector.sysfs /host/sys \ + -collector.filesystem.ignored-mount-points "^/(sys|proc|dev|host|etc)($|/)" # # Collectd # # Configure IP Address in collectd server configuration @@ -58,6 +58,7 @@ sudo docker run --name bottlenecks-grafana \ # Automate Prometheus Datasource and Grafana Dashboard creation set -e +sleep 5 python dashboard/automated_dashboard_datasource.py set +e @@ -76,7 +77,7 @@ sudo docker run \ set -e # Barometer # Configure IP Address in barometer server configuration -sleep 10 +sleep 5 python ${DISPATCH}/server_ip_configure.py ${MONITOR_CONFIG}/barometer_server.conf set +e |