Merge pull request #28 from tcpcloud/stacklight
Stacklight
diff --git a/metadata/service/support.yml b/metadata/service/support.yml
index 5f2e84d..cb2f13a 100644
--- a/metadata/service/support.yml
+++ b/metadata/service/support.yml
@@ -4,7 +4,7 @@
collectd:
enabled: true
heka:
- enabled: false
+ enabled: true
sensu:
enabled: true
sphinx:
@@ -12,4 +12,6 @@
iptables:
enabled: true
config:
- enabled: true
\ No newline at end of file
+ enabled: true
+ grafana:
+ enabled: true
diff --git a/rabbitmq/files/collectd_rabbitmq.conf b/rabbitmq/files/collectd_rabbitmq.conf
deleted file mode 100644
index 70aa668..0000000
--- a/rabbitmq/files/collectd_rabbitmq.conf
+++ /dev/null
@@ -1,24 +0,0 @@
-LoadPlugin python
-<Plugin python>
-
- ModulePath "/usr/lib/collectd-plugins/"
- LogTraces true
- Interactive false
- Import rabbitmq
-
- <Module rabbitmq>
-
- Username "guest"
- Password "guest"
- Realm "RabbitMQ Management"
- Host "localhost"
- Port "15672"
-
- <Ignore "queue">
- Regex "amq-gen-.*"
- Regex "tmp-.*"
- </Ignore>
-
- </Module>
-
-</Plugin>
\ No newline at end of file
diff --git a/rabbitmq/files/collectd_rabbitmq_info.conf b/rabbitmq/files/collectd_rabbitmq_info.conf
new file mode 100644
index 0000000..fae857f
--- /dev/null
+++ b/rabbitmq/files/collectd_rabbitmq_info.conf
@@ -0,0 +1,8 @@
+Import "rabbitmq_info"
+
+<Module "rabbitmq_info">
+ Host "{{ plugin.host }}"
+ Port "{{ plugin.port }}"
+ Username "{{ plugin.username }}"
+ Password "{{ plugin.password }}"
+</Module>
diff --git a/rabbitmq/files/grafana_dashboards/rabbitmq_influxdb.json b/rabbitmq/files/grafana_dashboards/rabbitmq_influxdb.json
new file mode 100644
index 0000000..1f49bb7
--- /dev/null
+++ b/rabbitmq/files/grafana_dashboards/rabbitmq_influxdb.json
@@ -0,0 +1,1261 @@
+{
+ "annotations": {
+ "list": [
+ {
+ "datasource": "lma",
+ "enable": true,
+ "iconColor": "#C0C6BE",
+ "iconSize": 13,
+ "lineColor": "rgba(255, 96, 96, 0.592157)",
+ "name": "Status",
+ "query": "select title,tags,text from annotations where $timeFilter and cluster = 'rabbitmq'",
+ "showLine": true,
+ "tagsColumn": "tags",
+ "textColumn": "text",
+ "titleColumn": "title"
+ }
+ ]
+ },
+ "editable": true,
+ "hideControls": false,
+ "id": null,
+ "links": [],
+ "originalTitle": "RabbitMQ",
+ "refresh": "1m",
+ "rows": [
+ {
+ "collapse": false,
+ "editable": true,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": true,
+ "colorValue": false,
+ "colors": [
+ "rgba(71, 212, 59, 0.4)",
+ "rgba(241, 181, 37, 0.73)",
+ "rgba(225, 40, 40, 0.59)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 34,
+ "interval": "> 60s",
+ "links": [],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "column": "value",
+ "condition": "",
+ "dsType": "influxdb",
+ "fill": "",
+ "function": "last",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "groupByTags": [],
+ "groupby_field": "",
+ "interval": "",
+ "measurement": "cluster_status",
+ "policy": "default",
+ "query": "SELECT last(\"value\") FROM \"cluster_status\" WHERE \"environment_label\" =~ /^$environment$/ AND \"cluster_name\" = 'rabbitmq' AND $timeFilter GROUP BY time($interval) fill(null)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=",
+ "value": "$environment"
+ },
+ {
+ "key": "cluster_name",
+ "operator": "=",
+ "value": "rabbitmq"
+ }
+ ]
+ }
+ ],
+ "thresholds": "1,3",
+ "title": "",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "no data",
+ "value": "null"
+ },
+ {
+ "op": "=",
+ "text": "OKAY",
+ "value": "0"
+ },
+ {
+ "op": "=",
+ "text": "WARN",
+ "value": "1"
+ },
+ {
+ "op": "=",
+ "text": "UNKN",
+ "value": "2"
+ },
+ {
+ "op": "=",
+ "text": "CRIT",
+ "value": "3"
+ },
+ {
+ "op": "=",
+ "text": "DOWN",
+ "value": "4"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 27,
+ "interval": "> 60s",
+ "links": [],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "alias": "Running nodes",
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "last",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "groupByTags": [],
+ "measurement": "rabbitmq_running_nodes",
+ "policy": "default",
+ "query": "SELECT last(\"value\") FROM \"rabbitmq_running_nodes\" WHERE \"hostname\" =~ /$server/ AND $timeFilter GROUP BY time($interval) fill(null)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/$server/"
+ },
+ {
+ "condition": "AND",
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Running nodes",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 31,
+ "interval": "> 60s",
+ "links": [],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "alias": "connections",
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "last",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "groupByTags": [],
+ "measurement": "rabbitmq_connections",
+ "policy": "default",
+ "query": "SELECT last(\"value\") FROM \"rabbitmq_connections\" WHERE \"hostname\" =~ /$server/ AND $timeFilter GROUP BY time($interval) fill(null)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/$server/"
+ },
+ {
+ "condition": "AND",
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Connections",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 30,
+ "interval": "> 60s",
+ "links": [],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "alias": "queues",
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "last",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "groupByTags": [],
+ "measurement": "rabbitmq_queues",
+ "policy": "default",
+ "query": "SELECT last(\"value\") FROM \"rabbitmq_queues\" WHERE \"hostname\" =~ /$server/ AND $timeFilter GROUP BY time($interval) fill(null)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/$server/"
+ },
+ {
+ "condition": "AND",
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Queues",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "title": ""
+ },
+ {
+ "collapse": false,
+ "editable": true,
+ "height": "250px",
+ "panels": [
+ {
+ "content": "",
+ "editable": true,
+ "error": false,
+ "id": 35,
+ "links": [],
+ "mode": "markdown",
+ "span": 3,
+ "style": {},
+ "title": "",
+ "type": "text"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(71, 212, 59, 0.4)",
+ "rgba(245, 150, 40, 0.73)",
+ "rgba(225, 40, 40, 0.59)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 32,
+ "interval": "> 60s",
+ "links": [],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "last",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "groupByTags": [],
+ "measurement": "rabbitmq_consumers",
+ "policy": "default",
+ "query": "SELECT last(\"value\") FROM \"rabbitmq_consumers\" WHERE \"hostname\" =~ /$server/ AND $timeFilter GROUP BY time($interval) fill(null)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/$server/"
+ },
+ {
+ "condition": "AND",
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Consumers",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 39,
+ "interval": "> 60s",
+ "links": [],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "last",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "groupByTags": [],
+ "measurement": "rabbitmq_channels",
+ "policy": "default",
+ "query": "SELECT last(\"value\") FROM \"rabbitmq_exchanges\" WHERE \"hostname\" =~ /$server/ AND $timeFilter GROUP BY time($interval) fill(null)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/$server/"
+ },
+ {
+ "condition": "AND",
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Channels",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 33,
+ "interval": "> 60s",
+ "links": [],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "last",
+ "groupBy": [
+ {
+ "params": [
+ "$interval"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "null"
+ ],
+ "type": "fill"
+ }
+ ],
+ "groupByTags": [],
+ "measurement": "rabbitmq_exchanges",
+ "policy": "default",
+ "query": "SELECT last(\"value\") FROM \"rabbitmq_exchanges\" WHERE \"hostname\" =~ /$server/ AND $timeFilter GROUP BY time($interval) fill(null)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "last"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "hostname",
+ "operator": "=~",
+ "value": "/$server/"
+ },
+ {
+ "condition": "AND",
+ "key": "environment_label",
+ "operator": "=~",
+ "value": "/^$environment$/"
+ }
+ ]
+ }
+ ],
+ "thresholds": "",
+ "title": "Exchanges",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "fill": 0,
+ "grid": {
+ "threshold1": null,
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2": null,
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 29,
+ "interval": "> 60s",
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "alias": "$tag_hostname",
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "mean",
+ "groupBy": [
+ {
+ "interval": "auto",
+ "params": [
+ "auto"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "hostname"
+ ],
+ "type": "tag"
+ },
+ {
+ "params": [
+ "previous"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "rabbitmq_messages",
+ "policy": "default",
+ "query": "SELECT mean(\"value\") FROM \"rabbitmq_messages\" WHERE \"environment_label\" =~ /^$environment$/ AND $timeFilter GROUP BY time($interval), \"hostname\" fill(previous)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=",
+ "value": "$environment"
+ }
+ ]
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Outstanding Messages",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "show": true
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "per second",
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "fill": 0,
+ "grid": {
+ "threshold1": null,
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2": null,
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 25,
+ "interval": "> 60s",
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "alias": "$tag_hostname",
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "mean",
+ "groupBy": [
+ {
+ "interval": "auto",
+ "params": [
+ "auto"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "hostname"
+ ],
+ "type": "tag"
+ },
+ {
+ "params": [
+ "previous"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "rabbitmq_used_memory",
+ "policy": "default",
+ "query": "SELECT mean(\"value\") FROM \"rabbitmq_used_memory\" WHERE \"environment_label\" =~ /^$environment$/ AND $timeFilter GROUP BY time($interval), \"hostname\" fill(previous)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=",
+ "value": "$environment"
+ }
+ ]
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Used memory",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "show": true
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "fill": 0,
+ "grid": {
+ "threshold1": null,
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2": null,
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 38,
+ "interval": "> 60s",
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "alias": "$tag_hostname",
+ "column": "value",
+ "dsType": "influxdb",
+ "function": "mean",
+ "groupBy": [
+ {
+ "interval": "auto",
+ "params": [
+ "auto"
+ ],
+ "type": "time"
+ },
+ {
+ "params": [
+ "hostname"
+ ],
+ "type": "tag"
+ },
+ {
+ "params": [
+ "previous"
+ ],
+ "type": "fill"
+ }
+ ],
+ "measurement": "rabbitmq_disk_free",
+ "policy": "default",
+ "query": "SELECT mean(\"value\") FROM \"rabbitmq_disk_free\" WHERE \"environment_label\" =~ /^$environment$/ AND $timeFilter GROUP BY time($interval), \"hostname\" fill(previous)",
+ "rawQuery": false,
+ "refId": "A",
+ "resultFormat": "time_series",
+ "select": [
+ [
+ {
+ "params": [
+ "value"
+ ],
+ "type": "field"
+ },
+ {
+ "params": [],
+ "type": "mean"
+ }
+ ]
+ ],
+ "tags": [
+ {
+ "key": "environment_label",
+ "operator": "=",
+ "value": "$environment"
+ }
+ ]
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Free disk",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "show": true
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "RabbitMQ"
+ }
+ ],
+ "schemaVersion": 12,
+ "sharedCrosshair": true,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "enable": true,
+ "list": [
+ {
+ "allFormat": "regex values",
+ "current": {},
+ "datasource": null,
+ "hide": 0,
+ "includeAll": false,
+ "name": "environment",
+ "options": [],
+ "query": "show tag values from cpu_idle with key = environment_label",
+ "refresh": 1,
+ "refresh_on_load": true,
+ "regex": "",
+ "type": "query"
+ },
+ {
+ "allFormat": "regex values",
+ "current": {},
+ "datasource": null,
+ "hide": 0,
+ "includeAll": true,
+ "name": "server",
+ "options": [],
+ "query": "show tag values from rabbitmq_consumers with key = hostname where environment_label =~ /^$environment$/ ",
+ "refresh": 1,
+ "refresh_on_load": true,
+ "regex": "",
+ "type": "query"
+ }
+ ]
+ },
+ "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"
+ },
+ "timezone": "browser",
+ "title": "RabbitMQ",
+ "version": 2
+}
diff --git a/rabbitmq/meta/collectd.yml b/rabbitmq/meta/collectd.yml
index f52a4fa..eefe46d 100644
--- a/rabbitmq/meta/collectd.yml
+++ b/rabbitmq/meta/collectd.yml
@@ -1,14 +1,13 @@
-plugin:
- # rabbitmq_server_rabbitmq:
- # plugin: rabbitmq
- # interval: 60
- #template: rabbitmq/files/collectd_rabbitmq.conf
+{%- from "rabbitmq/map.jinja" import server with context %}
+local_plugin:
+ rabbitmq_info:
+ plugin: python
+ template: rabbitmq/files/collectd_rabbitmq_info.conf
+ host: {{ server.management.bind.address }}
+ port: {{ server.management.bind.port }}
+ username: {{ server.admin.name }}
+ password: {{ server.admin.password }}
collectd_processes:
- plugin: processes
- interval: 60
- template: collectd/files/collectd_processes.conf
process:
rabbitmq:
match: '^[-\.\/a-z0-9]+beam(\.smp)?\ .*/var/lib/rabbitmq.*kernel.*mnesia.*'
- collectdmon:
- match: collectdmon
\ No newline at end of file
diff --git a/rabbitmq/meta/grafana.yml b/rabbitmq/meta/grafana.yml
new file mode 100644
index 0000000..1ccead2
--- /dev/null
+++ b/rabbitmq/meta/grafana.yml
@@ -0,0 +1,4 @@
+dashboard:
+ rabbitmq:
+ format: json
+ template: rabbitmq/files/grafana_dashboards/rabbitmq_influxdb.json
diff --git a/rabbitmq/meta/heka.yml b/rabbitmq/meta/heka.yml
index a33f56c..6d82aed 100644
--- a/rabbitmq/meta/heka.yml
+++ b/rabbitmq/meta/heka.yml
@@ -1,15 +1,151 @@
-{%- if pillar.rabbitmq.server is defined %}
-input:
- logstreamer_rabbitmq:
- type: "LogstreamerInput"
- log_directory: "/var/log/rabbitmq"
- file_match: 'rabbit@.+\.log\.?(?P<Index>\d+)?(.gz)?'
- priority: ["^Index"]
- decoder: "rabbitmq"
- oldest_duration: "168h"
-decoder:
- rabbitmq:
- type: "SandboxDecoder"
- filename: "lua_modules/decoders/rabbitmq.lua"
- module_directory: "/usr/share/heka/lua_modules;/usr/share/heka/lua_modules/common"
-{% endif %}
+log_collector:
+ decoder:
+ rabbitmq_log:
+ engine: sandbox
+ module_file: /usr/share/lma_collector/decoders/rabbitmq.lua
+ module_dir: /usr/share/lma_collector/common;/usr/share/heka/lua_modules
+ adjust_timezone: true
+ input:
+ rabbitmq_log:
+ engine: logstreamer
+ log_directory: "/var/log/rabbitmq"
+ file_match: 'rabbit@(?P<Node>.+)\.log$'
+ differentiator: ["rabbitmq.", "Node"]
+ decoder: "rabbitmq_log_decoder"
+ splitter: "rabbitmq_log_splitter"
+ splitter:
+ rabbitmq_log:
+ engine: regex
+ delimiter: '\n\n(=[^=]+====)'
+ delimiter_eol: false
+metric_collector:
+ trigger:
+ rabbitmq_disk_limit_critical:
+ description: 'RabbitMQ has reached the free disk threshold. All producers are blocked.'
+ severity: 'critical'
+ no_data_policy: 'okay'
+ rules:
+ - metric: rabbitmq_remaining_disk
+ relational_operator: '<='
+ threshold: 0
+ window: 20
+ periods: 0
+ function: min
+ rabbitmq_disk_limit_warning:
+ description: 'RabbitMQ is getting close to the free disk threshold.'
+ severity: 'warning'
+ no_data_policy: 'okay'
+ rules:
+ - metric: rabbitmq_remaining_disk
+ relational_operator: '<='
+ threshold: 104857600 # 100MB
+ window: 20
+ periods: 0
+ function: min
+ rabbitmq_memory_limit_critical:
+ description: 'RabbitMQ has reached the memory threshold. All producers are blocked.'
+ severity: 'critical'
+ no_data_policy: 'okay'
+ rules:
+ - metric: rabbitmq_remaining_memory
+ relational_operator: '<='
+ threshold: 0
+ window: 20
+ periods: 0
+ function: min
+ rabbitmq_memory_limit_warning:
+ description: 'RabbitMQ is getting close to the memory threshold.'
+ severity: warning
+ no_data_policy: 'okay'
+ rules:
+ - metric: rabbitmq_remaining_memory
+ relational_operator: '<='
+ threshold: 104857600 # 100MB
+ window: 20
+ periods: 0
+ function: min
+ rabbitmq_queue_warning:
+ description: 'The number of outstanding messages is too high.'
+ severity: warning
+ no_data_policy: 'okay'
+ rules:
+ - metric: rabbitmq_messages
+ relational_operator: '>='
+ threshold: 200
+ window: 120
+ periods: 0
+ function: avg
+ rabbitmq_check:
+ description: 'RabbitMQ cannot be checked'
+ severity: down
+ rules:
+ - metric: rabbitmq_check
+ relational_operator: '=='
+ threshold: 0
+ window: 60
+ periods: 0
+ function: last
+ alarm:
+ rabbitmq_server_disk:
+ alerting: enabled
+ triggers:
+ - rabbitmq_disk_limit_warning
+ - rabbitmq_disk_limit_critical
+ dimension:
+ service: rabbitmq-cluster
+ rabbitmq_server_memory:
+ alerting: enabled
+ triggers:
+ - rabbitmq_memory_limit_warning
+ - rabbitmq_memory_limit_critical
+ dimension:
+ service: rabbitmq-cluster
+ rabbitmq_server_queue:
+ alerting: enabled
+ triggers:
+ - rabbitmq_queue_warning
+ dimension:
+ service: rabbitmq-cluster
+ rabbitmq_check:
+ alerting: enabled
+ triggers:
+ - rabbitmq_check
+aggregator:
+ alarm_cluster:
+ rabbitmq_cluster:
+ alerting: enabled
+ policy: highest_severity
+ match:
+ service: rabbitmq-cluster
+ members:
+ - rabbitmq_server_disk
+ - rabbitmq_server_memory
+ - rabbitmq_server_queue
+ dimension:
+ service: rabbitmq
+ nagios_host: 01-service-clusters
+ rabbitmq_service:
+ # A check failure on a single node doesn't mean that the whole cluster
+ # is down, this is why a 'hostname' group_by and 'majority_of_members'
+ # policy are used here
+ policy: majority_of_members
+ alerting: enabled
+ group_by: hostname
+ match:
+ member: rabbitmq_check
+ members:
+ - rabbitmq_check
+ dimension:
+ service: rabbitmq
+ nagios_host: 01-service-clusters
+ rabbitmq:
+ policy: highest_severity
+ alerting: enabled_with_notification
+ match:
+ service: rabbitmq
+ members:
+ - rabbitmq_cluster
+ - rabbitmq_service
+ dimension:
+ cluster_name: rabbitmq
+ nagios_host: 00-top-clusters