February 2020
Intermediate to advanced
734 pages
38h 13m
English
Remember the witness node? Pausing the repmgr service only prevents the daemons from taking action when they detect an outage. Each will continue to watch the other nodes, especially the upstream primary. If we check the witness node logs, we have an objectively uninvolved history of the manual switchover we demonstrated previously.
Here is an example of that activity:

It's apparent that the witness node noticed the connection from pgha1, and that triggered six 10-second checks where it would have declared that node as unreachable. This would determine how the witness would vote if we hadn't paused the daemon cluster management. ...
Read now
Unlock full access