When: Thursday 21st 06:15 AM
- Prerequisites **: https://wikitech.wikimedia.org/wiki/MariaDB/Primary_switchover
- Team calendar invite
Affected wikis::
The x1 cluster is used by MediaWiki at WMF for databases that are "global" or "cross-wiki" in nature, and are typically associated with a MediaWiki extension
Affected features:
BounceHandler Cognate ContentTranslation Echo Flow GrowthExperiments ReadingLists UrlShortener WikimediaEditorTasks
Checklist:
NEW primary: db1103
OLD primary: db1120
- Check configuration differences between new and old primary:
sudo pt-config-diff --defaults-file /root/.my.cnf h=db1120.eqiad.wmnet h=db1103.eqiad.wmnet
Failover prep:
- Silence alerts on all hosts:
sudo cookbook sre.hosts.downtime --hours 1 -r "Primary switchover x1 T313398" 'A:db-section-x1'
- Set NEW primary with weight 0 (and depool it from API or vslow/dump groups if it is present).
sudo dbctl instance db1103 set-weight 0 sudo dbctl config commit -m "Set db1103 with weight 0 T313398"
- Topology changes, move all replicas under NEW primary
sudo db-switchover --timeout=25 --only-slave-move db1120 db1103
- Disable puppet on both nodes
sudo cumin 'db1120* or db1103*' 'disable-puppet "primary switchover T313398"'
- Merge gerrit puppet change to promote NEW primary: https://gerrit.wikimedia.org/r/c/operations/puppet/+/815830/
Failover:
- Log the failover:
!log Starting x1 eqiad failover from db1120 to db1103 - T313398
- Set section read-only:
sudo dbctl --scope eqiad section x1 ro "Maintenance until 06:15 UTC - T313398" sudo dbctl config commit -m "Set x1 eqiad as read-only for maintenance - T313398"
- Check x1 is indeed read-only
- Switch primaries:
db-mysql db1120 -e "set global read_only=1" sudo db-switchover --skip-slave-move db1120 db1103 echo "===== db1120 (OLD)"; sudo db-mysql db1120 -e 'show slave status\G' echo "===== db1103 (NEW)"; sudo db-mysql db1103 -e 'show slave status\G' db-mysql db1103 -e "set global read_only=0"
- Promote NEW primary in dbctl, and remove read-only
sudo dbctl --scope eqiad section x1 set-master db1103 sudo dbctl config commit -m "Promote db1103 to x1 primary and set section read-write T313398"
- Restart puppet on both hosts:
sudo cumin 'db1120* or db1103*' 'run-puppet-agent -e "primary switchover T313398"'
Clean up tasks:
- Clean up heartbeat table(s).
sudo db-mysql db1103 heartbeat -e "delete from heartbeat where file like 'db1120%';"
- change events for query killer:
events_coredb_master.sql on the new primary db1103 events_coredb_slave.sql on the new slave db1120
- Update DNS: https://gerrit.wikimedia.org/r/c/operations/dns/+/815831/
- Update candidate primary dbctl and orchestrator notes
sudo dbctl instance db1120 set-candidate-master --section x1 true sudo dbctl instance db1103 set-candidate-master --section x1 false (dborch1001): sudo orchestrator-client -c untag -i db1103 --tag name=candidate (dborch1001): sudo orchestrator-client -c tag -i db1120 --tag name=candidate
- Check zarcillo was updated
- db-switchover should do this. If it fails, do it manually: https://phabricator.wikimedia.org/P13956
sudo db-mysql db1115 zarcillo -e "select * from masters where section = 'x1';"
- (If needed): Depool db1120 for maintenance.
sudo dbctl instance db1120 depool sudo dbctl config commit -m "Depool db1120 T313398"
- Change db1120 weight to mimic the previous weight db1103:
sudo dbctl instance db1120 edit
- Apply outstanding schema changes to db1120 (if any)
- Update/resolve this ticket.