node_rejoin event is registered before node is actually rejoined during switchover
Opened this issue · 0 comments
spaskalev commented
I am using an event script to reconfigure database proxies on a postgres cluster configured with repmgr.
In a cluster of three (10,11,12) when performing a switchover from 10 to 11 the following events are registered through the event script
on 10: node_rejoin
on 11: standby_promote
on 12: standby_follow
On standby_promote and on standby_follow the new primary is already up and working as such when the event script is called.
However on node_rejoin I noticed a race that every once in a while the event was registered while node 10 was still running as primary. After a short time it is reconfigured to a standby.
repmgr version 5..