Add StackLight monitoring
This commit enables support for heka, collectd and grafana.
Depends-On: Ie6859fade90db3fe528ae8078c79c8be4c3ebcd7
Change-Id: Ic1c7373e6a7985873671d366dac132ac923f3431
diff --git a/docker/files/collectd_docker.conf b/docker/files/collectd_docker.conf
new file mode 100644
index 0000000..b3fafc8
--- /dev/null
+++ b/docker/files/collectd_docker.conf
@@ -0,0 +1,6 @@
+Import 'collectd_docker_info'
+
+<Module collectd_docker_info>
+ Interval "60"
+ Timeout "5"
+</Module>
diff --git a/docker/files/grafana_dashboards/docker_influxdb.json b/docker/files/grafana_dashboards/docker_influxdb.json
new file mode 100644
index 0000000..95c3736
--- /dev/null
+++ b/docker/files/grafana_dashboards/docker_influxdb.json
@@ -0,0 +1,1470 @@
+{
+ "annotations": {
+ "list": [
+ {
+ "datasource": "lma",
+ "enable": true,
+ "iconColor": "rgba(255, 96, 96, 1)",
+ "limit": 100,
+ "name": "Status",
+ "query": "select title,tags,text from annotations where $timeFilter and cluster = 'docker'",
+ "tagsColumn": "tags",
+ "textColumn": "text",
+ "titleColumn": "title",
+ "type": "alert"
+ }
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [],
+ "refresh": "1m",
+ "rows": [
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": true,
+ "colorValue": false,
+ "colors": [
+ "rgba(71, 212, 59, 0.4)",
+ "rgba(241, 181, 37, 0.73)",
+ "rgba(225, 40, 40, 0.59)"
+ ],
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 2,
+ "interval": "> 60s",
+ "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
+ },
+ "targets": [
+ {
+ "column": "value",
+ "condition": "",
+ "dsType": "influxdb",
+ "fill": "",
+ "function": "last",
+ "groupBy": [],
+ "groupby_field": "",
+ "interval": "",
+ "measurement": "cluster_status",
+ "policy": "default",
+ "query": "SELECT last(\"value\") FROM \"cluster_status\" WHERE \"cluster_name\" = 'docker' AND $timeFilter",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "cluster_name",
+ "operator": "=",
+ "value": "docker"
+ }
+ ]
+ }
+ ],
+ "thresholds": "1,3",
+ "title": "",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "no data",
+ "value": "null"
+ },
+ {
+ "op": "=",
+ "text": "OKAY",
+ "value": "0"
+ },
+ {
+ "op": "=",
+ "text": "WARN",
+ "value": "1"
+ },
+ {
+ "op": "=",
+ "text": "UNKW",
+ "value": "2"
+ },
+ {
+ "op": "=",
+ "text": "CRIT",
+ "value": "3"
+ },
+ {
+ "op": "=",
+ "text": "DOWN",
+ "value": "4"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 3,
+ "interval": ">60s",
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "docker_containers_total",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Total Containers",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "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)"
+ ],
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 7,
+ "interval": ">60s",
+ "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": true
+ },
+ "targets": [
+ {
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "docker_images",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Total Images",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": 250,
+ "panels": [
+ {
+ "content": "<br />\n<br />\n<br />\n<br />\n\n\n<h1>Containers status</h1>",
+ "id": 12,
+ "links": [],
+ "mode": "html",
+ "span": 2,
+ "title": "",
+ "type": "text"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 4,
+ "interval": ">60s",
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "docker_containers",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ },
+ {
+ "condition": "AND",
+ "key": "status",
+ "operator": "=",
+ "value": "running"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Running",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "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)"
+ ],
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 6,
+ "interval": ">60s",
+ "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": true
+ },
+ "targets": [
+ {
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "docker_containers",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ },
+ {
+ "condition": "AND",
+ "key": "status",
+ "operator": "=",
+ "value": "paused"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Paused",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 5,
+ "interval": ">60s",
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "docker_containers",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ },
+ {
+ "condition": "AND",
+ "key": "status",
+ "operator": "=",
+ "value": "stopped"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Stopped",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": 250,
+ "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)"
+ ],
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 8,
+ "interval": ">60s",
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "process_threads",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "service",
+ "operator": "=",
+ "value": "docker"
+ },
+ {
+ "condition": "AND",
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Threads",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "fill": 1,
+ "id": 9,
+ "interval": ">60s",
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "span": 3,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "alias": "user",
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "process_cputime_user",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "service",
+ "operator": "=",
+ "value": "docker"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ }
+ ]
+ },
+ {
+ "alias": "user",
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "process_cputime_syst",
+ "policy": "default",
+ "refId": "B",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "service",
+ "operator": "=",
+ "value": "docker"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "fill": 1,
+ "id": 10,
+ "interval": ">60s",
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "span": 3,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "alias": "rss",
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "process_memory_rss",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "service",
+ "operator": "=",
+ "value": "docker"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Resident Set Size",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "decbytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "fill": 1,
+ "id": 11,
+ "interval": ">60s",
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "span": 3,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "alias": "read",
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "process_disk_bytes_read",
+ "policy": "default",
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "service",
+ "operator": "=",
+ "value": "docker"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ }
+ ]
+ },
+ {
+ "alias": "write",
+ "dsType": "influxdb",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "process_disk_bytes_write",
+ "policy": "default",
+ "refId": "B",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ },
+ {
+ "condition": "AND",
+ "key": "service",
+ "operator": "=",
+ "value": "docker"
+ },
+ {
+ "condition": "AND",
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/^$server$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk IO",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "decbytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "System usage $server",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": true,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {},
+ "hide": 0,
+ "includeAll": false,
+ "label": null,
+ "multi": false,
+ "name": "environment",
+ "options": [],
+ "query": "show tag values from cpu_idle with key = environment_label",
+ "refresh": 1,
+ "refresh_on_load": true,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {},
+ "hide": 0,
+ "includeAll": false,
+ "label": null,
+ "multi": false,
+ "name": "server",
+ "options": [],
+ "query": "show tag values from process_threads with key=\"hostname\" where service = 'docker' and environment_label =~ /^$environment$/",
+ "refresh": 1,
+ "refresh_on_load": true,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Docker",
+ "version": 4
+}
diff --git a/docker/meta/collectd.yml b/docker/meta/collectd.yml
new file mode 100644
index 0000000..26b75d4
--- /dev/null
+++ b/docker/meta/collectd.yml
@@ -0,0 +1,12 @@
+{% from "docker/map.jinja" import host with context %}
+
+{%- if host.enabled %}
+local_plugin:
+ collectd_docker:
+ plugin: python
+ template: 'docker/files/collectd_docker.conf'
+ collectd_processes:
+ process:
+ docker:
+ match: dockerd
+{%- endif %}
diff --git a/docker/meta/grafana.yml b/docker/meta/grafana.yml
new file mode 100644
index 0000000..2f8efd9
--- /dev/null
+++ b/docker/meta/grafana.yml
@@ -0,0 +1,19 @@
+dashboard:
+ docker:
+ format: json
+ template: docker/files/grafana_dashboards/docker_influxdb.json
+ main:
+ row:
+ docker-data-plane:
+ title: Docker
+ panel:
+ docker:
+ title: Docker
+ links:
+ - dashboard: Docker
+ title: Docker
+ type: dashboard
+ target:
+ cluster_status:
+ rawQuery: true
+ query: SELECT last(value) FROM cluster_status WHERE cluster_name = 'docker' AND environment_label = '$environment' AND $timeFilter GROUP BY time($interval) fill(null)
diff --git a/docker/meta/heka.yml b/docker/meta/heka.yml
new file mode 100644
index 0000000..43fcd79
--- /dev/null
+++ b/docker/meta/heka.yml
@@ -0,0 +1,37 @@
+{% from "docker/map.jinja" import host with context %}
+
+{%- if host.enabled %}
+metric_collector:
+ trigger:
+ docker-local-check:
+ description: 'Docker is locally down'
+ severity: down
+ rules:
+ - metric: docker_check
+ relational_operator: '=='
+ threshold: 0
+ window: 90
+ periods: 0
+ function: last
+ alarm:
+ docker-check:
+ triggers:
+ - docker-local-check
+ dimension:
+ service: docker-check
+
+aggregator:
+ alarm_cluster:
+ docker:
+ policy: availability_of_members
+ alerting: enabled_with_notification
+ group_by: hostname
+ match:
+ service: docker-check
+ members:
+ - docker-check
+ dimension:
+ cluster_name: docker
+ nagios_host: 00-top-clusters
+
+{%- endif %}
diff --git a/metadata/service/support.yml b/metadata/service/support.yml
index a77821b..7cf7003 100644
--- a/metadata/service/support.yml
+++ b/metadata/service/support.yml
@@ -2,9 +2,11 @@
docker:
_support:
collectd:
- enabled: false
+ enabled: true
heka:
- enabled: false
+ enabled: true
+ grafana:
+ enabled: true
sensu:
enabled: false
sphinx: