Queues, retry and stale recovery

Queue states

A queue item moves through states with owner and lease information. A worker claims the job, sends heartbeats and completes with success, retry or failure. This prevents simultaneous processing by two workers.

Requeue

Use Requeue only when the cause is temporary or has been corrected, such as restored network access or a fixed destination. First verify that the original attempt did not succeed without reporting, to avoid duplicate delivery.

Mark Failed

For a permanent error after sufficient attempts, mark the item failed with a clear reason. This stops endless retry and preserves operator evidence.

Recover Stale

Recover stale releases jobs with an expired lease or heartbeat. Do not use it while a worker is still performing a long operation. Check service state, process and last heartbeat, then recover only genuinely stuck items.