Puppet Platform reference manual

HTTP client metrics available in Puppet Server 5 allows users to measure how long it takes for Puppet Server to make requests to and receive responses from other services, such as PuppetDB.

Determining metrics IDs

All of these metrics are of the form puppetlabs.<SERVER ID>.http-client.experimental.with-metric-id.<METRIC ID>.full-response.

Note: The <METRIC ID> describes what the metric measures. A metric ID is represented in the status endpoint as an array of strings, and in the metric itself the strings are joined together with periods. For instance, the metric ID of [puppetdb resource search] is puppetdb.resource.search, so the full metric name would be puppetlabs.<server-id>.http-client.experimental.with-metric-id.puppetdb.resource.search.full-response.

You can configure PuppetDB to be a backend for configuration files (through the storeconfigs setting), and you can configure Puppet Server to send reports to an external report processing service. If you configure either of these, then during the course of handling a Puppet agent run, Puppet Server makes several calls to external services to retrieve or store information.

  • During handling of a /puppet/v3/node request, Puppet Server issues:
    • a facts find request to PuppetDB for facts about the node, if they aren’t yet cached (typically the first time it requests facts for the node). Metric ID: [puppetdb facts find].
  • During handling of a /puppet/v3/catalog request, Puppet Server issues several requests:
    • a PuppetDB replace facts request, to replace the facts for the agent in PuppetDB with the facts it received from the agent. Metric ID: [puppetdb, command, replace_facts].
    • a PuppetDB resource search request, to search for resources if exported resources are used. Metric ID: [puppetdb, resource, search].
    • a PuppetDB query request, if the puppetdb_query function is used in Puppet code. Metric ID: [puppetdb, query].
    • a PuppetDB replace catalog request, to replace the catalog for the agent in PuppetDB with the newly compiled catalog. Metric ID: [puppetdb, command, replace_catalog].
  • During handling of a /puppet/v3/report request, Puppet Server issues:
    • a PuppetDB store report request, to store the submitted report. Metric ID: [puppetdb command store_report].
    • a request to the configured reports_url to store the report, if the HTTP report processor is enabled. Metric ID: [puppetdb report http].

Configuring

HTTP client metrics are enabled by default, but can be disabled by setting metrics-enabled to false in the http-client section of puppetserver.conf.

These metrics also depend on the server-id setting in the metrics section of puppetserver.conf. This defaults to localhost, and while localhost can collect metrics, change this setting to something unique to avoid metric naming collisions when exporting metrics to an external tool, such as Graphite.

This data is all available via the status API endpoint, at https://<MASTER HOSTNAME>:8140/status/v1/services/master?level=debug. Puppet Server 5.0 adds a http-client-metrics keyword in the map. If metrics are not enabled, or if Puppet Server has not issued any requests yet, then this array will be empty, like so: "http-client-metrics": [].

In the sample Grafana dashboard, the External HTTP Communications graph visualizes all of these metrics, and the tooltip describes each of them.

Example metrics output

"http-client-metrics": [
  {
    "aggregate": 407,
    "count": 1,
    "mean": 407,
    "metric-id": [
      "puppetdb",
      "facts",
      "find"
    ],
    "metric-name": "puppetlabs.localhost.http-client.experimental.with-metric-id.puppetdb.facts.find.full-response"
  },
  {
    "aggregate": 66,
    "count": 1,
    "mean": 66,
    "metric-id": [
      "puppetdb",
      "command",
      "replace_facts"
    ],
    "metric-name": "puppetlabs.localhost.http-client.experimental.with-metric-id.puppetdb.command.replace_facts.full-response"
  },
  {
    "aggregate": 60,
    "count": 2,
    "mean": 30,
    "metric-id": [
      "puppetdb",
      "resource",
      "search"
    ],
    "metric-name": "puppetlabs.localhost.http-client.experimental.with-metric-id.puppetdb.resource.search.full-response"
  },
  {
    "aggregate": 53,
    "count": 1,
    "mean": 53,
    "metric-id": [
      "puppetdb",
      "query"
    ],
    "metric-name": "puppetlabs.localhost.http-client.experimental.with-metric-id.puppetdb.query.full-response"
  },
  {
    "aggregate": 22,
    "count": 1,
    "mean": 22,
    "metric-id": [
      "puppetdb",
      "command",
      "store_report"
    ],
    "metric-name": "puppetlabs.localhost.http-client.experimental.with-metric-id.puppetdb.command.store_report.full-response"
  },
  {
    "aggregate": 16,
    "count": 1,
    "mean": 16,
    "metric-id": [
      "puppetdb",
      "command",
      "replace_catalog"
    ],
    "metric-name": "puppetlabs.localhost.http-client.experimental.with-metric-id.puppetdb.command.replace_catalog.full-response"
  },
  {
    "aggregate": 2,
    "count": 1,
    "mean": 2,
    "metric-id": [
      "puppet",
      "report",
      "http"
    ],
    "metric-name": "puppetlabs.localhost.http-client.experimental.with-metric-id.puppet.report.http.full-response"
  }
],
Back to top