Docs: PuppetDB 1 » Spec » Querying Metrics


PuppetDB 1 » Spec » Querying Metrics

Querying PuppetDB metrics is accomplished by making an HTTP request to paths under the /metrics REST endpoint.

Listing available metrics

Request format

To get a list of all available metric names:

  • Request /metrics/mbeans.
  • Use a GET request.
  • Provide an Accept header containing application/json.

Response format

Responses return a JSON Object mapping a string to a string:

  • The key is the name of a valid MBean
  • The value is a URI to use for requesting that MBean’s attributes

Retrieving a specific metric

Request format

To get the attributes of a particular metric:

  • Request /metrics/mbean/<name>, where <name> is something that was returned in the list of available metrics specified above.
  • Use a GET request.
  • Provide an Accept header containing application/json.

Response format

Responses return a JSON Object mapping strings to (strings/numbers/booleans).

Useful metrics

Population metrics

  • com.puppetlabs.puppetdb.query.population:type=default,name=num-nodes: The number of nodes in your population.
  • com.puppetlabs.puppetdb.query.population:type=default,name=num-resources: The number of resources in your population.
  • com.puppetlabs.puppetdb.query.population:type=default,name=avg-resources-per-node: The average number of resources per node in your population.
  • com.puppetlabs.puppetdb.query.population:type=default,name=pct-resource-dupes: The percentage of resources that exist on more than one node.

Database metrics

  • com.jolbox.bonecp:type=BoneCP: Database connection pool metrics. How long it takes to get a free connection, average execution times, number of free connections, etc.

Command-processing metrics

Each of the following metrics is available for each command supported in PuppetDB. In the below list of metrics, <name> should be substituted with a command specifier. Example <name>s you can use include:

  • global: Aggregate stats for all commands
  • replace catalog.1: Stats for catalog replacement
  • replace facts.1: Stats for facts replacement
  • deactivate node.1: Stats for node deactivation

Other than global, all command specifiers are of the form <command>.<version>. As we version commands, you’ll be able to get statistics for each version independently.

Metrics available for each command:

  • com.puppetlabs.puppetdb.command:type=<name>,name=discarded: stats about commands we’ve discarded (we’ve retried them as many times as we can, to no avail)
  • com.puppetlabs.puppetdb.command:type=<name>,name=fatal: stats about commands we failed to process.
  • com.puppetlabs.puppetdb.command:type=<name>,name=processed: stats about commands we’ve successfully processed
  • com.puppetlabs.puppetdb.command:type=<name>,name=processing-time: stats about how long it takes to process commands
  • com.puppetlabs.puppetdb.command:type=<name>,name=retried: stats about commands that have been submitted for retry (due to transient errors)

HTTP metrics

Each of the following metrics is available for each HTTP endpoint. In the below list of metrics, <name> should be substituted with a REST endpoint name. Example <name>s you can use include:

  • commands: Stats relating to the command processing REST endpoint. The PuppetDB terminus in Puppet talks to this endpoint to submit new catalogs, facts, etc.
  • metrics: Stats relating to the metrics REST endpoint. This is the endpoint you’re reading about right now!
  • facts: Stats relating to fact querying. This is the endpoint used by the puppetmaster for inventory service queries.
  • resources: Stats relating to resource querying. This is the endpoint used when collecting exported resources.

In addition to customizing <name>, the following metrics are available for each HTTP status code (<status code>). For example, you can see the stats for all 200 responses for the resources endpoint. This allows you to see, per endpoint and per response, independent counters and statistics.

  • com.puppetlabs.puppetdb.http.server:type=<name>,name=service-time: stats about how long it takes to service all HTTP requests to this endpoint
  • com.puppetlabs.puppetdb.http.server:type=<name>,name=<status code>: stats about how often we’re returning this response code

Storage metrics

Metrics involving the PuppetDB storage subsystem all begin with the com.puppetlabs.puppetdb.scf.storage:type=default,name= prefix. There are a number of metrics concerned with individual storage operations (storing resources, storing edges, etc.). Metrics of particular note include:

  • com.puppetlabs.puppetdb.scf.storage:type=default,name=duplicate-pct: the percentage of catalogs that PuppetDB determines to be duplicates of existing catalogs.
  • com.puppetlabs.puppetdb.scf.storage:type=default,name=gc-time: state about how long it takes to do storage compaction

JVM Metrics

  • java.lang:type=Memory: memory usage statistics
  • java.lang:type=Threading: stats about JVM threads

MQ Metrics

  • org.apache.activemq:BrokerName=localhost,Type=Queue,Destination=com.puppetlabs.puppetdb.commands: stats about the command processing queue: queue depth, how long messages remain in the queue, etc.

Example

Using curl from localhost:

curl -H "Accept: application/json" 'http://localhost:8080/metrics/mbean/java.lang:type=Memory'

↑ Back to top