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/s3.dblist
Checklist:
NEW primary: db1223
OLD primary: db1157
- Check configuration differences between new and old primary:
sudo pt-config-diff --defaults-file /root/.my.cnf h=db1157.eqiad.wmnet h=db1223.eqiad.wmnet
Failover prep:
- Silence alerts on all hosts:
sudo cookbook sre.hosts.downtime --hours 1 -r "Primary switchover s3 T363672" 'A:db-section-s3'
- Set NEW primary with weight 0 (and depool it from API or vslow/dump groups if it is present).
sudo dbctl instance db1223 set-weight 0 sudo dbctl config commit -m "Set db1223 with weight 0 T363672"
- Topology changes, move all replicas under NEW primary
sudo db-switchover --timeout=25 --only-slave-move db1157 db1223
- Disable puppet on both nodes
sudo cumin 'db1157* or db1223*' 'disable-puppet "primary switchover T363672"'
- Merge gerrit puppet change to promote NEW primary: https://gerrit.wikimedia.org/r/c/operations/puppet/+/1024752
Failover:
- Log the failover:
!log Starting s3 eqiad failover from db1157 to db1223 - T363672
- Set section read-only:
sudo dbctl --scope eqiad section s3 ro "Maintenance until 06:15 UTC - T363672" sudo dbctl --scope codfw section s3 ro "Maintenance until 06:15 UTC - T363672" sudo dbctl config commit -m "Set s3 eqiad as read-only for maintenance - T363672"
- Check s3 is indeed read-only
- Switch primaries:
sudo db-switchover --skip-slave-move db1157 db1223 echo "===== db1157 (OLD)"; sudo db-mysql db1157 -e 'show slave status\G' echo "===== db1223 (NEW)"; sudo db-mysql db1223 -e 'show slave status\G'
- Promote NEW primary in dbctl, and remove read-only
sudo dbctl --scope eqiad section s3 set-master db1223 sudo dbctl --scope eqiad section s3 rw sudo dbctl --scope codfw section s3 rw sudo dbctl config commit -m "Promote db1223 to s3 primary and set section read-write T363672"
- Restart puppet on both hosts:
sudo cumin 'db1157* or db1223*' 'run-puppet-agent -e "primary switchover T363672"'
Clean up tasks:
- Clean up heartbeat table(s).
sudo db-mysql db1223 heartbeat -e "delete from heartbeat where file like 'db1157%';"
- change events for query killer:
events_coredb_master.sql on the new primary db1223 events_coredb_slave.sql on the new slave db1157
- Update DNS: https://gerrit.wikimedia.org/r/c/operations/dns/+/1024753
- Update candidate primary dbctl and orchestrator notes
sudo dbctl instance db1157 set-candidate-master --section s3 true sudo dbctl instance db1223 set-candidate-master --section s3 false (dborch1001): sudo orchestrator-client -c untag -i db1223 --tag name=candidate (dborch1001): sudo orchestrator-client -c tag -i db1157 --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 = 's3';"
- (If needed): Depool db1157 for maintenance.
sudo dbctl instance db1157 depool sudo dbctl config commit -m "Depool db1157 T363672"
- Change db1157 weight to mimic the previous weight db1223:
sudo dbctl instance db1157 edit
- Apply outstanding schema changes to db1157 (if any) -> Pending will be done after the reimage
- Update/resolve this ticket.