Have you ever posted a blog entry, then reached for it, only to find it not there? That
is what happens if you try to "automatically" split reads and writes.
There is too much chance that the 'write' (posting the blog) will not be replicated to the
slave before you 'read' from the slave.
The application must take this into consideration, it must not depend on some trivial
Galera (Percona XtraDB Cluster) has a 'simple' way for the 'read' to indicate that it
needs to wait for replication to catch up. Otherwise, you are on your own.
> -----Original Message-----
> From: Manuel Arostegui [mailto:manuel@stripped]
> Sent: Saturday, October 12, 2013 7:41 AM
> To: nik600
> Cc: replication@stripped
> Subject: Re: solution to split read-only query between master and slave
> 2013/10/12 nik600 <nik600@stripped>
> > Hi what do you use to split read/write requests between one master and
> > many slave mysql server?
> > The mysq-proxy is in alpha state from years...ad it seems that you
> > have to manage the split at application-level.
> > What do you suggest to use to split sql queries between many
> > slave-replicated and one master node?
> It is usually done from the application/code level. It is the easiest way.
> Forget about mysql-proxy. It is kinda dead and never worked too well to be