When: During a pre-defined DBA maintenance windows
Affected wikis:: https://noc.wikimedia.org/conf/highlight.php?file=dblists/s4.dblist
Checklist:
NEW primary: db2140
OLD primary: db2179
- Check configuration differences between new and old primary:
sudo pt-config-diff --defaults-file /root/.my.cnf h=db2179.codfw.wmnet h=db2140.codfw.wmnet 16 config differences Variable db2179 db2140 ========================= ========================= ========================= general_log_file db2179.log db2140.log gtid_binlog_pos 171966513-171966513-34... 171966513-171966513-34... gtid_binlog_state 171966513-171966513-34... 171966513-171966513-34... gtid_current_pos 0-180359175-3375835514... 0-180359175-3375835514... gtid_domain_id 180363364 180367386 gtid_slave_pos 0-180359175-3375835514... 0-180359175-3375835514... hostname db2179 db2140 log_bin_basename /srv/sqldata/db2179-bin /srv/sqldata/db2140-bin log_bin_index /srv/sqldata/db2179-bi... /srv/sqldata/db2140-bi... pid_file /srv/sqldata/db2179.pid /srv/sqldata/db2140.pid report_host db2179.codfw.wmnet db2140.codfw.wmnet rpl_semi_sync_master_e... ON OFF rpl_semi_sync_slave_en... OFF ON server_id 180363364 180367386 slow_query_log_file db2179-slow.log db2140-slow.log wsrep_node_name db2179 db2140
Failover prep:
- Silence alerts on all hosts:
sudo cookbook sre.hosts.downtime --hours 1 -r "Primary switchover s4 T374512" 'A:db-section-s4'
- Set NEW primary with weight 0
sudo dbctl instance db2140 set-weight 0 sudo dbctl config commit -m "Set db2140 with weight 0 T374512"
- Depool NEW from any specific group (API, vslow, dump) if present.
sudo dbctl instance db2140 edit # If some changes were made: sudo dbctl config commit -m "Remove db2140 from API/vslow/dump T374512" removed API
- Topology changes, move all replicas under NEW primary
sudo db-switchover --timeout=25 --replicating-master --read-only-master --only-slave-move db2179 db2140
- Disable puppet on both nodes
sudo cumin 'db2179* or db2140*' 'disable-puppet "primary switchover T374512"'
- Merge gerrit puppet change to promote NEW primary: https://gerrit.wikimedia.org/r/1072105
Failover:
- Log the failover:
!log Starting s4 codfw failover from db2179 to db2140 - T374512
- Switch primaries:
sudo db-switchover --replicating-master --read-only-master --skip-slave-move db2179 db2140 echo "===== db2179 (OLD)"; sudo db-mysql db2179 -e 'show slave status\G' echo "===== db2140 (NEW)"; sudo db-mysql db2140 -e 'show slave status\G'
- Promote NEW primary in dbctl
sudo dbctl --scope codfw section s4 set-master db2140 sudo dbctl config commit -m "Promote db2140 to s4 primary T374512"
- Restart puppet on both hosts:
sudo cumin 'db2179* or db2140*' 'run-puppet-agent -e "primary switchover T374512"'
Clean up tasks:
- Clean up heartbeat table(s).
sudo db-mysql db2140 heartbeat -e "delete from heartbeat where file like 'db2179%';"
- change events for query killer:
curl -sS 'https://gerrit.wikimedia.org/r/plugins/gitiles/operations/software/+/refs/heads/master/dbtools/events_coredb_master.sql?format=TEXT' | base64 -d | sudo db-mysql db2140 curl -sS 'https://gerrit.wikimedia.org/r/plugins/gitiles/operations/software/+/refs/heads/master/dbtools/events_coredb_slave.sql?format=TEXT' | base64 -d | sudo db-mysql db2179
- Update candidate primary dbctl and orchestrator notes
sudo dbctl instance db2179 set-candidate-master --section s4 true sudo dbctl instance db2140 set-candidate-master --section s4 false sudo cumin 'dborch*' 'orchestrator-client -c untag -i db2140 --tag name=candidate' sudo cumin 'dborch*' 'orchestrator-client -c tag -i db2179 --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 db2179 for maintenance.
sudo dbctl instance db2179 depool sudo dbctl config commit -m "Depool db2179 T374512"
- Change db2179 weight to mimic the previous weight db2140:
sudo dbctl instance db2179 edit
- Apply outstanding schema changes to db2179 (if any)
- Update/resolve this ticket.