Here is the setup in question:
Fast-Start Failover: ENABLED
Threshold: 30 seconds Target: dbabc
Observer: (none)
Lag Limit: 30 seconds (not in use)
Shutdown Primary: FALSE
Auto-reinstate: TRUE
LogXptMode = 'SYNC'
NetTimeout = '30'
Can you help me understand what happens in the following scenarios?
- Standby was unavailable for some time, say 35 seconds, while the primary was being updated. As a result the databases became
UNSYNCHRONIZED, therefore automatic failover is no longer allowed. Then the primary gets isolated from the standby and observer for more than 30 seconds while the standby and observer are still able to communicate.
In this condition, would redo generation be stalled on the primary after reaching the 30 seconds threshold? Would it continue accepting new commits since it's already running UNSYNC?
- Primary gets isolated from the standby and observer for more than 30 secs while the standby and observer are still able to communicate. The primary was being written to, as a result redo generation was stalled on the primary and the observer initiated an automatic failover to the standby.
In this condition, when the communication is re-established between all the sites, the observer will reinstate the former primary as a standby.
