Verify Nextcloud Collabora session stickiness and cron reliability #11

Open
opened 2026-07-06 00:12:26 +00:00 by claude-bot · 0 comments
Collaborator

Two untested risk notes flagged during the Nextcloud migration, not yet confirmed one way or the other since cutover (2026-07-04):

  • Collabora session stickiness: port 9980 is published via the routing mesh with round-robin in front of it -- a single WOPI editing session could in theory bounce between manager nodes if Collabora doesn't hold state across requests. Watch for document-editing weirdness; pin to one node if it surfaces.
  • Cron container reliability: Swarm doesn't enforce depends_on ordering/health, so if db/redis aren't up yet when the cron container starts, background jobs could silently fail to run. Check docker service logs nextcloud_cron periodically rather than assuming it's fine.

See ccnet-nextcloud-swarm/README.md 'Known gaps' section.

Two untested risk notes flagged during the Nextcloud migration, not yet confirmed one way or the other since cutover (2026-07-04): - **Collabora session stickiness**: port 9980 is published via the routing mesh with round-robin in front of it -- a single WOPI editing session could in theory bounce between manager nodes if Collabora doesn't hold state across requests. Watch for document-editing weirdness; pin to one node if it surfaces. - **Cron container reliability**: Swarm doesn't enforce depends_on ordering/health, so if db/redis aren't up yet when the cron container starts, background jobs could silently fail to run. Check `docker service logs nextcloud_cron` periodically rather than assuming it's fine. See ccnet-nextcloud-swarm/README.md 'Known gaps' section.
Sign in to join this conversation.
No description provided.