I recently installed the MySQL module for our backup software at work. Noticing that the new version has PIT restore capability I gleefully clicked to enable that, aimed the backup software at the master database, and away it went.
The way this software implements PIT is to do a level 0 backup of the database, send the database a RESET MASTER command so as to remove any existing binlogs, and then (later) just backup all the bin logs since they'll be replayable from the level 0 it created, etc., etc.
Except that sending a RESET MASTER command completely hoses any slaves that are replicating off the master you're backing up.
So I opened a trouble ticket with the vendor, asking them, "How do you think I'm supposed to do this, given how this interacts with my slaves?"
Their response, in a sentence:
"It shouldn't cause any issues with the replicas themselves."
Yeah, no shit, pal. That's what I'm trying to tell you. Backing up the master shouldn't break the replicas. But it does.
Sometimes we get really really smart support engineers with this vendor, and sometimes, you wonder if they've read the ticket.
Leave a comment