openstack team mailing list archive

[Florian Haas <florian@xxxxxxxxxxx>]

On Tue, Feb 21, 2012 at 12:51 PM, Darren Birkett
<darren.birkett@xxxxxxxxx> wrote:
> Surely it depends what you want out of it?  If just pure high
> availability (ie ability to continue functioning  when primary node
> goes down), then doesn't a master/master with floating IP fit the
> bill?

Master/master with a floating IP, MySQL-on-DRBD with a floating IP,
MySQL-on-SAN with a floating IP, even MySQL-on-RBD (if your storage
pool is RADOS based) with a floating IP might be an option.

> It sounds like what you're after is better performance/throughput
> rather than HA - the ability to write to multiple masters at the same
> time.  Even here if nova *was* able to hit multiple databases, a
> master/master setup isn't going to suit, as each write to one master
> still needs to be replicated to the other so throughput is going to be
> the same.  You'd then need to look to something like a native mysql
> cluster to get actual increased throughput.

Essentially -- and there are much more acclaimed MySQL experts on this
list who will correct me if I'm wrong -- most general-purpose write
scalability options for MySQL ultimately boil down to sharding the
database, anyhow. Which is an problem different from, and orthogonal
to, replication and high availability.

Cheers,
Florian

-- 
Need help with High Availability?
http://www.hastexo.com/now