Skip to content

Keep EKV available until fixture shutdown persists dirty state - #32

Merged
jeregrine merged 1 commit into
mainfrom
test/ekv-clean-shutdown
Sep 15, 2026
Merged

jeregrine merged 1 commit into
mainfrom
test/ekv-clean-shutdown

Conversation

@jeregrine

Copy link
Copy Markdown
Member

Problem

The shared-EKV fixture replaces DurableServer's child spec with a bare map, losing its shutdown budget. The scenario can report success and then tear down its peer/quorum before final persistence, producing forced kills and missing-ETS errors during cleanup.

Fix

Preserve the production child spec and drain both lifecycle managers before stopping durable supervisors, while EKV quorum remains available. Assert normal child termination, supervisor shutdown, the final stopped status, and preservation of dirty state before the peer is removed.

Add a focused single-node fixture-shutdown regression and exercise an unsynced update in the shared-EKV scenario so pre-existing stored state cannot satisfy the final-persistence assertion.

@jeregrine
jeregrine force-pushed the test/ekv-clean-shutdown branch from ec7ea90 to 45ed2c2 Compare September 15, 2026 14:41
@jeregrine
jeregrine merged commit 5a3877e into main Sep 15, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant