Implement switchover to a designated Postgres instance without using repmgr
Per multiple discussions, it was agreed that it is more reliable to use a regular, low-level approach of performing switchover.
Repmgr is to be serving as an autofailover solution after migration is complete. But during migration the switchover is to be performed using some designated Postgres node in GCP (supposingly, that one which follows the Azure master) and using "promote" procedure described in Postgres manual https://www.postgresql.org/docs/9.6/static/warm-standby-failover.html – either using pg_ctl promote
command or the trigger file.
After that, the remaining GCP nodes are to be attached to the new master and repmgr is to be reconfigured.