Monitoring replication slave

Note: this recipe is working with ArangoDB 2.5, you need a collectd curl_json plugin with correct boolean type mapping.

Problem

How to monitor the slave status using the collectd curl_JSON plugin.

Solution

Since arangodb reports the replication status in JSON, integrating it with the collectd curl_JSON plugin should be an easy exercise. However, only very recent versions of collectd will handle boolean flags correctly.

Our test master/slave setup runs with the master listening on tcp://127.0.0.1:8529 and the slave (which we query) listening on tcp://127.0.0.1:8530. They replicate a database by the name testDatabase.

Since replication appliers are active per database and our example doesn’t use the default _system, we need to specify its name in the URL like this: _db/testDatabase.

We need to parse a document from a request like this:

  1. curl --dump - http://localhost:8530/_db/testDatabase/_api/replication/applier-state

If the replication is not running the document will look like that:

  1. {
  2.   "state": {
  3.     "running": false,
  4.     "lastAppliedContinuousTick": null,
  5.     "lastProcessedContinuousTick": null,
  6.     "lastAvailableContinuousTick": null,
  7.     "safeResumeTick": null,
  8.     "progress": {
  9.       "time": "2015-11-02T13:24:07Z",
  10.       "message": "applier shut down",
  11.       "failedConnects": 0
  12.     },
  13.     "totalRequests": 1,
  14.     "totalFailedConnects": 0,
  15.     "totalEvents": 0,
  16.     "totalOperationsExcluded": 0,
  17.     "lastError": {
  18.       "time": "2015-11-02T13:24:07Z",
  19.       "errorMessage": "no start tick",
  20.       "errorNum": 1413
  21.     },
  22.     "time": "2015-11-02T13:31:53Z"
  23.   },
  24.   "server": {
  25.     "version": "2.7.0",
  26.     "serverId": "175584498800385"
  27.   },
  28.   "endpoint": "tcp://127.0.0.1:8529",
  29.   "database": "testDatabase"
  30. }

A running replication will return something like this:

  1. {
  2.   "state": {
  3.     "running": true,
  4.     "lastAppliedContinuousTick": "1150610894145",
  5.     "lastProcessedContinuousTick": "1150610894145",
  6.     "lastAvailableContinuousTick": "1151639153985",
  7.     "safeResumeTick": "1150610894145",
  8.     "progress": {
  9.       "time": "2015-11-02T13:49:56Z",
  10.       "message": "fetching master log from tick 1150610894145",
  11.       "failedConnects": 0
  12.     },
  13.     "totalRequests": 12,
  14.     "totalFailedConnects": 0,
  15.     "totalEvents": 2,
  16.     "totalOperationsExcluded": 0,
  17.     "lastError": {
  18.       "errorNum": 0
  19.     },
  20.     "time": "2015-11-02T13:49:57Z"
  21.   },
  22.   "server": {
  23.     "version": "2.7.0",
  24.     "serverId": "175584498800385"
  25.   },
  26.   "endpoint": "tcp://127.0.0.1:8529",
  27.   "database": "testDatabase"
  28. }

We create a simple collectd configuration in /etc/collectd/collectd.conf.d/slave_testDatabase.conf that matches our API:

  1. TypesDB "/etc/collectd/collectd.conf.d/slavestate_types.db"
  2. <Plugin curl_json>
  3.   # Adjust the URL so collectd can reach your arangod slave instance:
  4.   <URL "http://localhost:8530/_db/testDatabase/_api/replication/applier-state">
  5.    # Set your authentication to that database here:
  6.    # User "foo"
  7.    # Password "bar"
  8.     <Key "state/running">
  9.        Type "boolean"
  10.      </Key>
  11.     <Key "state/totalOperationsExcluded">
  12.        Type "counter"
  13.      </Key>
  14.     <Key "state/totalRequests">
  15.        Type "counter"
  16.      </Key>
  17.     <Key "state/totalFailedConnects">
  18.        Type "counter"
  19.      </Key>
  20.   </URL>
  21. </Plugin>

To get nice metric names, we specify our own types.db file in /etc/collectd/collectd.conf.d/slavestate_types.db:

  1. boolean                     value:ABSOLUTE:0:1

So, basically state/running will give you 0/1 if its (not / ) running through the collectd monitor.