Skip to content

test(goclusterfs): keep a crashed node's locks through the test replay - #91

Merged
mattmattox merged 2 commits into
mainfrom
claude/crash-test-adopt-locks
Nov 13, 2026
Merged

mattmattox merged 2 commits into
mainfrom
claude/crash-test-adopt-locks

Conversation

@mattmattox

Copy link
Copy Markdown
Contributor

Test-only. remountAndReplay replayed a crashed node's journal as a new node, but didn't take over the crashed node's lock claims first. So the replay and the test after it waited 15s for those claims to go stale. It now adopts them for the replay and releases them after, as mount does for a dead node.

Each test node keeps its slot in the last byte of its UUID, so the helper finds the crashed node's slot from the owner recorded in its journal.

On .68, all 86 goclusterfs tests pass, and the package run drops from 132s to 85s. The remaining slow tests wait on stuck locks on purpose (30s), plus one fsck test at 7s.

Under #90 a claim never goes stale once an owner view is installed. Test lock managers don't install one, so these tests didn't hang under #90 either.

🤖 Generated with Claude Code

mattmattox and others added 2 commits November 13, 2026 12:09
remountAndReplay replayed the crashed node's journal as a new node without
taking over the crashed node's lock claims, so its replay and the tests
after it waited 15s for those claims to go stale. It now adopts them for
the replay and releases them after, as mount does for a dead node.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@mattmattox
mattmattox merged commit e8983f7 into main Nov 13, 2026
13 checks passed
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