[Bucardo-general] Connection disruptions in multi-master replication

[David Jung]

Dear all,

I have been exploring the use of bucardo for multi-master replication. It works fine as long as the network connection between the servers is stable, but I have some concerns about reliability when the network connection intermittently breaks up.

In my test setup I have three servers A, B, and C (which are separate linux computers, each running a postgresql13 database). Because I want replication to continue between the remaining servers when the connection breaks up between any pair of servers, I decided to run an instance of bucardo on each server. Bucardo on server A is set up with a source<->source sync between the databases of servers A and B, server B syncs B and C, and bucardo on server C syncs C and A.

Now if I for example set up server A to drop any packets from C via iptables, any changes on C are still synced to B, but A misses out on them completely. This would be expected, but when I re-establish the connection from C to A the changes are still not synced - A just remains inconsistent with B and C indefinitely. Worse still, I can't seem to find any hint that there was a sync failure in the bucardo logs of either A and C. "bucardo status" shows the syncs to be "Good" and "Active".

Is there some log that I am missing that would show me these failed syncs or some configuration that I missed so that failed syncs are repeated when the connection is re-established?

Best regards,
David