unkin/jellyfin-ha
4
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity

fix: make plugin sync safe for concurrent replica starts #12

Merged
benvin merged 1 commits from benvin/fix-plugin-sync-race into main 2026-09-13 18:57:11 +10:00
Conversation 0 Commits 1 Files Changed 1
+48 -9
unkin-agent commented 2026-09-13 18:50:11 +10:00
Member
Copy Link
Copy Source

Replicas share /config over RWX CephFS. The plugin sync unconditionally rm -rf'd then cp -a'd each plugin dir on every start, so two replicas starting together raced: one hit a permission error deleting a dir the other was writing, or crashed later with UnauthorizedAccessException loading a half-copied plugin. Reproduced 3/3 with two simultaneous starts.

  • skip the sync once the correct version is already installed
  • stage a new/changed version privately per replica, then atomic-rename into place
  • treat a lost rename race as success, not an error
  • best-effort remove stale differently-versioned copies so they don't shadow the current one
  • no locking: a stuck lock from a dead replica would be worse than the race
Replicas share /config over RWX CephFS. The plugin sync unconditionally rm -rf'd then cp -a'd each plugin dir on every start, so two replicas starting together raced: one hit a permission error deleting a dir the other was writing, or crashed later with UnauthorizedAccessException loading a half-copied plugin. Reproduced 3/3 with two simultaneous starts. - skip the sync once the correct version is already installed - stage a new/changed version privately per replica, then atomic-rename into place - treat a lost rename race as success, not an error - best-effort remove stale differently-versioned copies so they don't shadow the current one - no locking: a stuck lock from a dead replica would be worse than the race
unkin-agent added 1 commit 2026-09-13 18:50:11 +10:00
fix: make plugin sync safe for concurrent replica starts
ci/woodpecker/pr/build Pipeline was successful
Details
e0a195179b 
Two replicas starting together on the shared /config volume race the
unconditional rm+cp, so one dies with a permission error or a plugin
UnauthorizedAccessException. Skip the sync once the correct version
is already present, stage new/changed versions and move them into
place with an atomic rename, and treat a lost rename race or an
already-removed stale version as success rather than failure.
benvin merged commit e103b57e97 into main 2026-09-13 18:57:11 +10:00
benvin deleted branch benvin/fix-plugin-sync-race 2026-09-13 18:57:11 +10:00
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
Clear labels
bug
Something is not working
 duplicate
This issue or pull request already exists
 enhancement
New feature
 help wanted
Need some help
 invalid
Something is wrong
 question
More information is needed
 wontfix
This won't be fixed
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
benvin (BenVincent) unkin-agent (Unkin Agent) unkinben (Ben Vincent)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: unkin/jellyfin-ha#12
Delete Branch "benvin/fix-plugin-sync-race"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?