Removing dead server from replication ring

[Geoff Kuenning]

I have a fairly simple replication setup, A->B->C->A. Machine B has gone away and cannot be replaced, so I need to remove it from the ring. The problem is that it took a month to determine that B would never return, and in the meantime updates have happened. So A has a correct copy of the database, but C is patiently waiting for B to come back up so that it can pull the updates.

One obvious option would be shut down both remaining servers, mysqldump A and reload C, then reconfigure the master/slave setup to be A->C->A. But I'm wondering whether I might be able to do something simpler. If I were to simply modify C so that its master is A, would it correctly read the logs on A and replay just the transactions it missed? I don't know the log format, so I'm unable to answer this question myself.

FWIW, these servers are relatively inactive (a few updates per day, primarily to a single table). I checked the mysql-bin files in /var/lib/mysql and they are all present (back several years in fact) so I can be confident that no data would be lost. I just don't know whether the server on C would have enough information to replay onlty the transactions it missed.