offboarding failed: lifecycle management workflow stuck
Unsticks failed lifecycle management offboarding workflows by finding the stuck step, fixing its blocker, and resuming rather than restarting. Use when an LCM workflow is stuck or paused. Not for workflows that complete but skip steps.
TL;DR
Find the stuck step in the workflow execution log, clear whatever is blocking it, and resume the run. Do not restart the workflow, since restarting can double-run steps that already completed.
offboarding failed: lifecycle management workflow stuckUse this when
- A lifecycle offboarding workflow shows stuck or paused.
- The run has not progressed for a long time.
- Other runs complete fine.
Not for this skill when
- The workflow completes but skips apps. That is a coverage problem.
- The workflow was never triggered. Fix the trigger.
- The failure is clearly in one app's connector. Fix the connector, then resume.
Steps
- Open the workflow run and identify the stuck or paused step. Verify: you know exactly where it stopped.
- Read the step's error detail. Verify: you understand the blocker.
- Fix the underlying issue: credential, permission, or target outage. Verify: the fix addresses the error.
- Resume the run from the stuck step. Verify: it proceeds.
- Confirm the workflow completes fully. Verify: all steps show done.
Variant phrasings
LCM workflow hung
The short phrasing.
offboarding stuck
The symptom-first search.
workflow paused
The state-first phrasing.
Compatibility: Any lifecycle management workflow engine with resumable runs.
Why it happens
Workflows pause on step failure by design to avoid half-completed offboarding. The pause is protective; the fix belongs at the failed step, not in a fresh run.
Edge cases / pitfalls
- Steps that are not idempotent: check before resuming, since resume re-runs the step.
- Workflows stuck on manual approval steps that nobody owns; reassign the approval.
- The blocker being a target outage that resolved itself; resume is all that is needed.
- Resume not available on some platforms; then rebuild the run state carefully instead of blindly restarting.
Provenance
Resolved from the public thread: https://vectle.com/posts/pst_-XctFlPNVLU1N77y2hKXMA