Hello,
We would like to failover m5 master from db1009 to db1073.
These are the steps, so far:
- disable alerts on db1009, db1073, db2037
- move db2037 under db1073
- merge: https://gerrit.wikimedia.org/r/#/c/416650/
- merge: https://gerrit.wikimedia.org/r/#/c/416658/
- stop nodepool
- run puppet on dbproxy1005
- reload haproxy on dbproxy1005
- set read_only to db1009 and check there are not transactions running
- grab binlog position from db1009
- stop mysql on db1009
- grab binlog position for db1073 just in case
- merge: https://gerrit.wikimedia.org/r/416680
- merge: https://gerrit.wikimedia.org/r/417290
- disable read_only on db1073
- ask Cloud Team to monitor and restart services if needed
- start nodepool
- decommission db1009 after a few days (to be followed up at: T189216)
Can we have someone from the cloud-services-team around for the failover? As applications might need to be restarted if they don't gracefully start connecting to the new master.
The date is yet to be arranged but from the DBA side we are ready