When: 2nd Feb at 09AM UTC
NEW primary: es1020
OLD primary: es1021
- Check configuration differences between new and old primary:
sudo pt-config-diff --defaults-file /root/.my.cnf h=es1020.eqiad.wmnet h=es1021.eqiad.wmnet
Failover prep:
- Disable writes to es4 by merging this patch: https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/758715/
- Check there are no writes
- Silence alerts on all hosts:
sudo cookbook sre.hosts.downtime --hours 2 -r "Switchover es4 T300127" 'A:db-section-es4'
- Set NEW primary with weight 10
sudo dbctl instance es1020 set-weight 10 sudo dbctl config commit -m "Set es1020 with weight 10 T300127"
- Topology changes, move all replicas under NEW primary
sudo db-switchover --timeout=15 --only-slave-move es1021 es1020
- Disable puppet on both nodes
sudo cumin 'es1020* or es1021*' 'disable-puppet "primary switchover T300127"'
- Merge gerrit puppet change to promote NEW primary: https://gerrit.wikimedia.org/r/c/operations/puppet/+/758716
Failover:
- Log the failover:
!log Starting es4 eqiad failover from es1021 to es1020 - T300127
- Switch primaries:
sudo db-switchover --skip-slave-move es1021 es1020 echo "===== es1021 (OLD)"; sudo db-mysql es1021 -e 'show slave status\G' echo "===== es1020 (NEW)"; sudo db-mysql es1020 -e 'show slave status\G'
- Promote NEW primary in dbctl, and remove read-only
sudo dbctl --scope eqiad section es4 set-master es1020 sudo dbctl config commit -m "Promote es1020 to es4 primary T300127"
- Restart puppet on both hosts:
sudo cumin 'es1020* or es1021*' 'run-puppet-agent -e "primary switchover T300127"'
Clean up tasks:
- Clean up heartbeat table(s).
- change events for query killer:
events_coredb_master.sql on the new primary es1020 events_coredb_slave.sql on the new slave es1021
- Update DNS: https://gerrit.wikimedia.org/r/c/operations/dns/+/758717
- Re-enable writes on es4: https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/758918
- Update candidate primary dbctl notes
sudo dbctl instance es1021 set-candidate-master --section es4 true sudo dbctl instance es1020 set-candidate-master --section es4 false
- Check zarcillo was updated
- db-switchover should do this. If it fails, do it manually: https://phabricator.wikimedia.org/P13956
- Depool OLD primary for reimage
sudo dbctl instance es1021 depool sudo dbctl config commit -m "Depool es1021 until it's reimaged to buster T300127"