Steward.Plugin (Steward v0.1.0)

Copy Markdown View Source

Supervisor for Steward's active components.

NOTE: Do not use this directly, use Steward.

Usage

After defining your application's plugin module with use Steward, you can add it to the list of plugins in your Oban configuration:

config :my_app, Oban,
  repo: MyApp.Repo,
  plugins: [MyApp.Steward]

Options

You may pass options to the plugin as usual.

config :my_app, Oban,
  repo: MyApp.Repo,
  plugins: [{MyApp.Steward, reconcile_batch_size: 10}]
  • :reconcile_poll_interval - how often to check for due resources, in milliseconds. Defaults to 5_000. While reconciliation runs fail as a whole (e.g. the database is unavailable), the interval doubles after each failed run, up to 10 minutes, and is reset after the next successful run.
  • :reconcile_batch_size - resources reconciled per poll. Defaults to 10.
  • :reconciliation_error_backoff - how long after a failed reconciliation (i.e. Steward.Controller.load/1 or Steward.Worker.spec/3 raised) a resource is reconciled again, in milliseconds. The error is stored in the resource's :reconciliation_error field. Errors affecting a whole batch of resources back off the whole batch, without storing the error. Defaults to 1 hour.
  • :prune_poll_interval - how often to prune, in milliseconds. Defaults to 60_000.
  • :prune_batch_size - resources pruned per batch. Defaults to 500.
  • :prune_max_age - how long after cancellation a resource's jobs are deleted, in milliseconds, or :infinity to disable pruning. Defaults to 30 days.
  • :lifeline_poll_interval - how often to rescue resources stuck in reconciliation, in milliseconds. Defaults to 60_000.
  • :lifeline_reconciliation_timeout - how long a resource may stay in reconciliation, in milliseconds, before it's considered stuck (e.g. after a node crashed mid-reconciliation) and scheduled again, or :infinity to disable rescuing stuck resources. Defaults to 1 day.