Page MenuHomePhabricator

Switchover s6 master (db2114 -> db2129)
Closed, ResolvedPublic

Description

When: During a pre-defined DBA maintenance windows

Affected wikis:: https://noc.wikimedia.org/conf/highlight.php?file=dblists/s6.dblist

Checklist:

NEW primary: db2129
OLD primary: db2114

  • Check configuration differences between new and old primary:
sudo pt-config-diff --defaults-file /root/.my.cnf h=db2114.codfw.wmnet h=db2129.codfw.wmnet

Failover prep:

  • Silence alerts on all hosts:
sudo cookbook sre.hosts.downtime --hours 1 -r "Primary switchover s6 T363713" 'A:db-section-s6'
  • Set NEW primary with weight 0 (and depool it from API or vslow/dump groups if it is present).
sudo dbctl instance db2129 set-weight 0
sudo dbctl config commit -m "Set db2129 with weight 0 T363713"
  • Topology changes, move all replicas under NEW primary
sudo db-switchover --timeout=25 --replicating-master --read-only-master --only-slave-move db2114 db2129
  • Disable puppet on both nodes
sudo cumin 'db2114* or db2129*' 'disable-puppet "primary switchover T363713"'
  • Merge gerrit puppet change to promote NEW primary: FIXME

Failover:

  • Log the failover:
!log Starting s6 codfw failover from db2114 to db2129 - T363713
  • Switch primaries:
sudo db-switchover --replicating-master --read-only-master --skip-slave-move db2114 db2129
echo "===== db2114 (OLD)"; sudo db-mysql db2114 -e 'show slave status\G'
echo "===== db2129 (NEW)"; sudo db-mysql db2129 -e 'show slave status\G'
  • Promote NEW primary in dbctl
sudo dbctl --scope codfw section s6 set-master db2129
sudo dbctl config commit -m "Promote db2129 to s6 primary T363713"
  • Restart puppet on both hosts:
sudo cumin 'db2114* or db2129*' 'run-puppet-agent -e "primary switchover T363713"'

Clean up tasks:

  • Clean up heartbeat table(s).
sudo db-mysql db2129 heartbeat -e "delete from heartbeat where file like 'db2114%';"
  • change events for query killer:
events_coredb_master.sql on the new primary db2129
events_coredb_slave.sql on the new slave db2114
  • Update candidate primary dbctl and orchestrator notes
sudo dbctl instance db2114 set-candidate-master --section s6 true
sudo dbctl instance db2129 set-candidate-master --section s6 false
(dborch1001): sudo orchestrator-client -c untag -i db2129 --tag name=candidate
(dborch1001): sudo orchestrator-client -c tag -i db2114 --tag name=candidate
sudo db-mysql db1215 zarcillo -e "select * from masters where section = 's6';"
  • (If needed): Depool db2114 for maintenance. → This is needed as the server will be decommissionned T358741 and will serve as testing device before this.
sudo dbctl instance db2114 depool
sudo dbctl config commit -m "Depool db2114 T363713"
  • Change db2114 weight to mimic the previous weight db2129:
sudo dbctl instance db2114 edit
  • Apply outstanding schema changes to db2114 (if any)
  • Update/resolve this ticket.

Event Timeline

Change #1024758 had a related patch set uploaded (by Gerrit maintenance bot; author: Gerrit maintenance bot):

[operations/puppet@production] mariadb: Promote db2129 to s6 master

https://gerrit.wikimedia.org/r/1024758

Mentioned in SAL (#wikimedia-operations) [2024-04-29T14:51:47Z] <arnaudb@cumin1002> START - Cookbook sre.hosts.downtime for 1:00:00 on 27 hosts with reason: Primary switchover s6 T363713

Mentioned in SAL (#wikimedia-operations) [2024-04-29T14:52:03Z] <arnaudb@cumin1002> dbctl commit (dc=all): 'Set db2129 with weight 0 T363713', diff saved to https://phabricator.wikimedia.org/P61421 and previous config saved to /var/cache/conftool/dbconfig/20240429-145203-arnaudb.json

Mentioned in SAL (#wikimedia-operations) [2024-04-29T14:52:11Z] <arnaudb@cumin1002> END (PASS) - Cookbook sre.hosts.downtime (exit_code=0) for 1:00:00 on 27 hosts with reason: Primary switchover s6 T363713

I've encountered an issue running db-switchover:

Checking if db2151/(none) needs to be moved under the new master...
Disabling GTID on db2151/(none)...
Waiting some seconds for db to catch up...
[ERROR]: db2151/(none) failed to be moved under the new master

I've resumed replication and ran: stop slave; CHANGE MASTER TO MASTER_USE_GTID=Slave_pos; start slave; after it caught up on db2151 and db2129. db2151 is left depooled.

Change #1024758 merged by Arnaudb:

[operations/puppet@production] mariadb: Promote db2129 to s6 master

https://gerrit.wikimedia.org/r/1024758

Mentioned in SAL (#wikimedia-operations) [2024-04-29T15:55:13Z] <arnaudb> Starting s6 codfw failover from db2114 to db2129 - T363713

Mentioned in SAL (#wikimedia-operations) [2024-04-29T15:55:58Z] <arnaudb@cumin1002> dbctl commit (dc=all): 'Promote db2129 to s6 primary T363713', diff saved to https://phabricator.wikimedia.org/P61430 and previous config saved to /var/cache/conftool/dbconfig/20240429-155557-arnaudb.json

Mentioned in SAL (#wikimedia-operations) [2024-04-29T15:58:39Z] <arnaudb@cumin1002> dbctl commit (dc=all): 'Depool db2114 T363713', diff saved to https://phabricator.wikimedia.org/P61431 and previous config saved to /var/cache/conftool/dbconfig/20240429-155838-arnaudb.json

ABran-WMF claimed this task.
ABran-WMF triaged this task as Medium priority.
ABran-WMF updated the task description. (Show Details)
ABran-WMF moved this task from Triage to Done on the DBA board.