The "db1047/eventlogging_sync processes" icinga alert keeps getting critical (and pages me) since at least early January. This should be investigated and either resolved or the Icinga check needs to be adopted.
Description
Description
Details
Details
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| mariadb: allow up to two eventlogging_sync processes | operations/puppet | production | +1 -1 |
Related Objects
Related Objects
- Mentioned Here
- T124307: Improve eventlogging replication procedure
Event Timeline
Comment Actions
This check should be marked as non-critical, and not sending pages (but alters to chat/web interface/email). For some reason, either this was changed or some configuration is paging all dba checks, instead of only the critical ones.
Comment Actions
Change 285640 had a related patch set uploaded (by Filippo Giunchedi):
mariadb: allow up to two eventlogging_sync processes
Comment Actions
Change 285640 merged by Filippo Giunchedi:
mariadb: allow up to two eventlogging_sync processes
Comment Actions
I think this will not completely fix the issue, as it seems the script may still fail due to some real issue with data corruption charset, but once this is deployed, we could merge it into https://phabricator.wikimedia.org/T133588#2237441 to fix the pending db-specific issue.