When: During a pre-defined DBA maintenance windows
Prerequisites: https://wikitech.wikimedia.org/wiki/MariaDB/Primary_switchover
- Team calendar invite
Affected wikis:: https://noc.wikimedia.org/conf/highlight.php?file=dblists/s4.dblist
Checklist:
NEW primary: db1238
OLD primary: db1160
- Check configuration differences between new and old primary:
sudo pt-config-diff --defaults-file /root/.my.cnf h=db1160.eqiad.wmnet h=db1238.eqiad.wmnet
Failover prep:
- Silence alerts on all hosts:
sudo cookbook sre.hosts.downtime --hours 1 -r "Primary switchover s4 T367378" 'A:db-section-s4'
- Set NEW primary with weight 0
sudo dbctl instance db1238 set-weight 0 sudo dbctl config commit -m "Set db1238 with weight 0 T367378"
- Depool NEW from any specific group (API, vslow, dump) if present.
sudo dbctl instance db1238 edit # If some changes were made: sudo dbctl config commit -m "Remove db1238 from API/vslow/dump T367378"
- Topology changes, move all replicas under NEW primary
sudo db-switchover --timeout=25 --only-slave-move db1160 db1238
- Disable puppet on both nodes
sudo cumin 'db1160* or db1238*' 'disable-puppet "primary switchover T367378"'
- Merge gerrit puppet change to promote NEW primary: https://gerrit.wikimedia.org/r/1042595
Failover:
- Log the failover:
!log Starting s4 eqiad failover from db1160 to db1238 - T367378
- Set section read-only:
sudo dbctl --scope eqiad section s4 ro "Maintenance until 06:15 UTC - T367378" sudo dbctl --scope codfw section s4 ro "Maintenance until 06:15 UTC - T367378" sudo dbctl config commit -m "Set s4 eqiad as read-only for maintenance - T367378"
- Check s4 is indeed read-only
- Switch primaries:
sudo db-switchover --skip-slave-move db1160 db1238 echo "===== db1160 (OLD)"; sudo db-mysql db1160 -e 'show slave status\G' echo "===== db1238 (NEW)"; sudo db-mysql db1238 -e 'show slave status\G'
- Promote NEW primary in dbctl, and remove read-only
sudo dbctl --scope eqiad section s4 set-master db1238 sudo dbctl --scope eqiad section s4 rw sudo dbctl --scope codfw section s4 rw sudo dbctl config commit -m "Promote db1238 to s4 primary and set section read-write T367378"
- Restart puppet on both hosts:
sudo cumin 'db1160* or db1238*' 'run-puppet-agent -e "primary switchover T367378"'
Clean up tasks:
- Clean up heartbeat table(s).
sudo db-mysql db1238 heartbeat -e "delete from heartbeat where file like 'db1160%';"
- change events for query killer:
events_coredb_master.sql on the new primary db1238 events_coredb_slave.sql on the new slave db1160
- Update DNS: https://gerrit.wikimedia.org/r/1042596
- Update candidate primary dbctl and orchestrator notes
sudo dbctl instance db1160 set-candidate-master --section s4 true sudo dbctl instance db1238 set-candidate-master --section s4 false (dborch1001): sudo orchestrator-client -c untag -i db1238 --tag name=candidate (dborch1001): sudo orchestrator-client -c tag -i db1160 --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 db1215 zarcillo -e "select * from masters where section = 's4';"
- (If needed): Depool db1160 for maintenance.
sudo dbctl instance db1160 depool sudo dbctl config commit -m "Depool db1160 T367378"
- Change db1160 weight to mimic the previous weight db1238:
sudo dbctl instance db1160 edit
- Update/resolve this ticket.