[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: Failover Master setup

I didn't say move just the IP.  Move the mount (via SAN or NAS) and the IP.
Take the -entire- application with you on failure.  That way, you
don't have to worry about staying in-sync in real-time, which neither
slurpd nor syncrepl can guarantee.

On 6/24/05, Digant C Kasundra <digant@uta.edu> wrote:
> Simply moving an virtual IP is insufficient.  An app that needs to talk
> to the master will also likely want to make queries against it.  So, in
> a failover setup, the secondary master has to stay in sync with the
> master.  In other words, slurpd has to be replicating to it at all times
> and the secondary should not be replicating to anyone else (it should
> act as a slave).  When the primary goes down, the secondary has to
> realize it is now in charge or replicating to the slaves.
> -- DK