Backend plugin for statsd to publish output to the Stackdriver custom metrics API over HTTPS. Note that this plugin does not work with Google Stackdriver at this time.
Install statsd normally. We'll call the root directory of the statsd install $STATSD_HOME
From your $STATSD_HOME directory run $ npm install stackdriver-statsd-backend will install this module into the appropriate place, and the configurations below will reference it as a backend.
For now you can pull stackdriver.js and put it in the backends directory of your statsd and it will be configurable like below.
To set up the Stackdriver backend, you need a Stackdriver account and API key. Everything else is optional. Any of the configurations below can be put into a stackdriverConfig.js and used as a statsd config on startup.
$ bin/statsd stackdriverConfig.js
Please set flushInterval to 1 minute (60000 milliseconds) or more, as that is the highest frequency we support at this time (another good reason to use this statsd plugin).
{
flushInterval: 60000,
backends: [ "stackdriver-statsd-backend" ],
stackdriver: {
apiKey: "YOUR_API_KEY_HERE"
}
}To associate the metrics with a particular instance (such as the one statsd is running on) add the source parameter to your configuration. The custom metrics generated will be associated with that AWS, GCE or Rackspace Cloud instance. For AWS, instance ID is in the form i-00000000.
If statsd is running locally on an AWS EC2 instance, and you wish to associate the metrics with that instance in Stackdriver, you can also set the source parameter to "detect-aws". This will use the local EC2 metadata URL to determine the instance ID where it is running.
If statsd is running locally on an Google Compute instance, and you wish to associate the metrics with that instance in Stackdriver, you can also set the source parameter to "detect-gce". This will use the local GCE metadata URL to determine the instance ID where it is running.
{
flushInterval: 60000,
backends: [ "stackdriver-statsd-backend" ],
stackdriver: {
apiKey: "YOUR_API_KEY_HERE",
source: "AWS Instance ID here"
}
}If you are sending to Stackdriver from an aggregated metrics source (for example into a statsd cluster) the instance can be inferred from a prefix on the metric name rather than a fixed value. If your metric names look like "instanceId--metricName", you would use:
Note that statsd is very restrictive on which characters are allowed in metric names so we've had to default to -- as the separator (Their regex is .replace(/[^a-zA-Z_\-0-9\.]/g, '')).
{
flushInterval: 60000,
backends: [ "stackdriver-statsd-backend" ],
stackdriver: {
apiKey: "YOUR_API_KEY_HERE",
sourceFromPrefix: true,
sourcePrefixSeparator: "--"
}
}To send percentiles with your timer values, use the percentThreshold key in your configuration to set the percentiles you want (statsd defaults to computing the 90th percentile only), and set the sendTimerPercentiles key in the Stackdriver backend config to enable sending them.
{
flushInterval: 60000,
percentThreshold: [95, 99],
backends: [ "stackdriver-statsd-backend" ],
stackdriver: {
apiKey: "YOUR_API_KEY_HERE",
sendTimerPercentiles: true
}
}To output additional logging information, add the debug parameter set to true. It will be more verbose, and can be helpful to tell what exactly is being sent to Stackdriver.
{
flushInterval: 60000,
backends: [ "stackdriver-statsd-backend" ],
stackdriver: {
apiKey: "YOUR_API_KEY_HERE",
debug: "true"
}
}