Steward.Telemetry (Steward v0.1.0)

Copy Markdown View Source

Telemetry events emitted by Steward.

Reconciliation

  • [:steward, :reconciliation, :start | :stop | :exception] - a reconciliation run, which reconciles all due resources in batches. Metadata: :oban, :steward. :stop carries :batch_count and :resource_count measurements. An :exception event means a whole batch failed (e.g. a resource's controller no longer exists, or a database error).

  • [:steward, :reconciliation, :resource, :start | :stop | :exception] - a single resource, i.e. its Steward.Controller.load/1 and its workers' Steward.Worker.spec/3 calls. Metadata: :steward, :controller, :resource. An :exception event means the resource's reconciliation failed and was rescheduled after the :reconciliation_error_backoff; the error is also stored on the resource.

Lifeline

  • [:steward, :lifeline, :rescue_all_stuck, :start | :stop | :exception] - a lifeline run, which reschedules resources stuck in reconciliation for longer than the :lifeline_reconciliation_timeout (see Steward.Plugin). Metadata: :oban, :steward. The :stop event adds the :rescued_count measurement.

Pruner

  • [:steward, :pruner, :prune_all_expired, :start | :stop | :exception] - a pruner run, which deletes the jobs of resources cancelled more than :prune_max_age ago and marks them pruned (see Steward.Plugin). Metadata: :oban, :steward. The :stop event adds the :pruned_count measurement.