List:Replication« Previous MessageNext Message »
From:Johan De Meersman Date:November 18 2013 10:24am
Subject:Re: old master to slave?
View as plain text  
----- Original Message -----
> From: "Regan Yelcich" <reganyelcich@stripped>
> Subject: old master to slave?
> 
> Can someone explain the process I should go through to bring an "old" master
> back online as a "slave" after a successful failover using mysql-master-ha?

The classic wisdom is that you just start from scratch again after a failover; especially
if you failed over because there was a problem on the old master.

That being said, there's a few things that can speed you up.

 - If you did a controlled failover, you can note the new master's log positions between
turning the old one off and activating the new one. You can then restart the old master
as a slave on those positions, provided of course that the slave was fully updated with
the last of the old master's logs.

 - If you use circular replication (aka master-master), the old master will simply pick up
any changes that happen on the other side. The term "master-master" tends to suggest a few
things that this is not, though, so caveat administrator. Read the docs, and read them
repeatedly.

 - The wonderful Percona Toolkit (formerly Maatkit) contains, among other things, the
pt-table-checksum and pt-table-sync tools, which can help bring a "broken" replication
set back up to speed. They're very good, but there's a few things they can't do (unique
key-less tables are a bugger, for one), so again, read the docs thoroughly.

/johan


-- 
Unhappiness is discouraged and will be corrected with kitten pictures.
Thread
old master to slave?Regan Yelcich18 Nov
  • Re: old master to slave?Johan De Meersman18 Nov