Conversation
Notices
-
It would be really nice if you could spin up a new empty #MySQL server, say "slave to this master", and it would sync data automatically.
-
Having to take a server offline to set up a slave kinda defeats the purpose, y'know?
-
@evan holy crap, you can't? I shouldn't be that surprised I guess, the kernel just added the ability to do the same with RAID volumes.
-
No, you have to freeze the master, do a data dump, save some obscure variables, install the data dump on the slave, and let it catch up.
-
...and it still usually doesn't work.
-
@evan I figured something like that was the alternative, but I also thought it supported live migration, wonder if it's being worked on.
-
Yeah, I have to admit I last time I did that I had to do it three times before I got it right.
-
If you already have a slave, you can take it down, copy its files to another server, and start that server up again. It's still error prone.
-
Also, once it's working don't forget to get Nagios to monitor the replication status (io status, sql status and seconds behind)
-
and now we all know the thing that will ultimately kill MySQL when it is replaced with something that does this trick properly.
-
Having wrestled with this myself lately I wondered if it were really that hard. Seems like a brittle hacked together system.
-
@gomerx some of the new NoSQL systems are much more flexible.
-
Yes, definitely. In my case converting our code to a NoSQL backend isn't going to happen any time soon.
-
I'm trying to do it with our servers. It's pretty fun.
-
@mohanpram no. Replication requires that you set up a full data dump on the slave.
-
@evan if you run a full data dump of an empty server just after setting it up, does that not enable you to create slaves from it much later?
-
All Evan Prodromou content and data are available under the