Closed legrandlegrand closed 1 year ago
Hi @legrandlegrand ; thanks for reporting an issue here. One thing that is strange to me is why would you first stop a node and then perform a switchover? pg_autoctl is meant to take care of doing the failover automatically, you don't have to ask for it when a node is down, just wait for it to happen. It takes about 30s.
It might be a bug that you found here, when you attempt a manual switchover when a node is down but not detected down by the monitor yet. But I am still going to close that issue. Please consider re-opening it if that workflow is important in your case.
Thank you Dimitri, for this answer. I had to evaluate pg_auto_autofailover for a postgres HA project that is now frozen :o( Then this issue had many goals:
Hello, I'm testing pg_auto_failover, et found it very interesting ! I had a strange behavior, that I would like to share with you to get a better understanding.
Configuration:
initial state:
then trying to perform a switchover with pg3 node down
and it hangs until node 3 restart (using pg_auto_failover run) and connections to postgres is not possible.
My questions: 1/ is that the expected behaviour ? 2/ if expected behaviour, should SWITCHOVER be allowed ? 3/ if node 3 can not be restarted, what is the faster way to restart service and allow postgres connections
Thanks in advance Regards PAscal