Skip to content

improve observability of provisionerd metrics #12474

@johnstcn

Description

@johnstcn

We currently define the following tags for provisioner daemon metrics:

  • coderd_provisionerd_jobs_current: provisioner (terraform)
  • coderd_provisionerd_num_daemons: no labels, could be useful to have by provisioner in future
  • coderd_provisionerd_job_timings_seconds: provisioner (terraform), status (success|failed)

This does not allow answering questions such as:

  • How long do workspace builds take, on average?
  • Which template takes the longest to execute?
  • Which provisioners take longer to build workspaces?

To answer questions like the above, it would be useful to have the following information:

  • scope: scope of the provisioner job (org|user)
  • type: type of the provisioner job (template_version_import|workspace_build_start|workspace_build_stop)
  • template: in case of workspace builds, what is the template to which the job refers?
  • name: name of the provisioner (defaults to hostname)

cc @kylecarbs

Metadata

Metadata

Assignees

No one assigned

    Labels

    customer-requestedFeatures requested by enterprise customers. Only humans may set this.observabilityIssues related to observability (metrics, dashboards, alerts, opentelemetry)

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions