Jan Kirchhoff <j.kirchhoff@logical-line.de> writes:
Kristian, I didn't know of that function, good to hear that.
But I was after some kind of slave_skip_counter-like function to make galera skip an event in case of problems. I'm fine with a cluster member stopping (or going read-only or something like that) because some update couldn't be applied, but if I (for whatever reason) think that's OK and just want it to skip that event, how could I do that?
I'm sorry, I quoted the wrong piece of your original email, which may have created some confusion. I meant to comment on this:
application connects to all databases, sets SQL_LOG_BIN=0 and then starts updating just a few tables (up to 15-20GB of pure sql statements per hour depending on the time of the day). This works fine, the server was processing the statements no slower than the other non-galera servers.
I couldn't see the updates on the second galera-server and figured that was because of the SQL_LOG_BIN=0. Now things got complicated. How to
My point was - you can maybe avoid the problem that SQL_LOG_BIN=0 makes the updates not be seen on other galera servers. SKIP_REPLICATION=1 works similar to SQL_LOG_BIN=0, in that it makes the statements not be replicated. But with SKIP_REPLICATION=1, data is still written to the binlog, so I thought the missing update problem would be solved for galera. Unfortunately, I do not know about the slave_skip_counter and galera. - Kristian.