See SURFconext Incident 2026-08-25. Short summary:
- database restore in unrelated database schema cause global (write) lock in Galera
- this cause consent-writes to block
- after a while. all database connections are busy (waiting of the lock for writes)
- then also simply authentications (requiring only reads) fail.
Can this be improved? For example by decreasing lock-timeouts?
See SURFconext Incident 2026-08-25. Short summary:
Can this be improved? For example by decreasing lock-timeouts?