Workflows remain stuck in the Running state due to an author pod deadlock

This article explains how to identify and mitigate an author pod deadlock in Adobe Experience Manager (AEM) as a Cloud Service when workflow instances remain stuck in the Running state after the Start step.

Description description

Environment

Adobe Experience Manager (AEM) as a Cloud Service

Issue/Symptoms

  • Workflow instances complete the Start step but remain in the Running state indefinitely.
  • Workflow jobs do not progress because processing related to the Sling Event Dispatcher and Oak observation on an affected author pod is blocked.
  • In observed cases, no corresponding application error is visible in the logs while the issue is occurring.
  • Restarting or replacing the affected author pod restores workflow and job processing.
  • Example workflow instance paths may resemble /var/workflow/instances/<server>/<date>/<workflow-instance>.

Cause

A deadlock affecting Sling Event Dispatcher and Oak observation on an author pod can prevent workflow and job events from being delivered, leaving workflow instances stuck in the Running state until the affected pod is restarted or replaced.

Resolution resolution

To fix workflows that remain stuck in the Running state because of an author pod deadlock, try this workaround:

  1. Verify that affected workflow instances remain in the Running state after the Start step and confirm that related jobs are not progressing.
  2. Determine whether the behavior is isolated to a specific author pod by comparing workflow activity across author instances and reviewing available indicators of stalled event or observation processing.
  3. Collect thread dumps from the affected author pod during the incident when possible so the deadlock condition can be validated.
  4. Restart or replace the affected author pod to restore workflow and job processing.
  5. After the restart, monitor workflow execution and job progression to confirm that newly started and previously stalled items begin processing again.
recommendation-more-help
experience-cloud-kcs-help-kbarticles