Does a single Postgres instance automatically failover to the replica, or must this be done manually by me? If it is automatic, how long before the replica becomes primary and can start receiving requests? In other words, how much down time is there for DB requests?
As of right now, the primary will be marked “unhealthy” after being in a failing state for roughly 20 seconds. Once the primary keeper is marked “unhealthy”, the leader election process will start and automatically issue a failover so long as there’s an eligible/healthy standby available. The new primary should start accepting requests right away.
Hope that helps!
1 Like