[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: replication and failover




On Nov 1, 2006, at 3:38 PM, David Bishop wrote:

Next step in our project* is to provide some sort of failover.  After
years of fantastic uptime, we've had two seperate network connectivity
issues in the last week (one ISP related, one literally 'oops, bumped
the cable' related).  We have an opportunity to setup a machine in a
server room a couple of states away, which ought to be wide enough to
cover 'natural disaster or terrorist activity', but I'm unsure how to
proceed.

The main (only, really) problem is replicating the mysql database. I'd
prefer the alternative server to be perfectly usable all the time, and
run in some sort of round-robin fashion.  That means, however,
duplicating all the data on our primary server over to the secondary. I
looked into the two replication methods for mysql and have found
problems with both.  Traditional master/slave won't work, as that has
one point of failure for writes: the master (negating the whole reason
for doing this). And it appears clustering won't work, as we won't have anything even close to 100Mbit connection between the two (which the docs
say is needed).  Any ideas for how to go about this?


I think what might work best for your situation is a multimaster replication system. It takes some changes to your auto-incrementing defaults and the my.cnf file, but overall it should do what you need.

http://www.onlamp.com/pub/a/onlamp/2006/04/20/advanced-mysql- replication.html describes it in detail.

Hope that helps!
Eric



Reply to: