On Thu, Jan 21, 2016 at 11:34:18AM -0800, John R Pierce wrote:
> the most difficult part is reliably determining that A) the master has
> crashed, and B) fencing the failed old master so it doesn't wake up and
> think its still in charge.
>
And, depending on your workload, C) that you actually want to fail over.
I've seen an awful lot of people want automatic failover who also
can't afford for the already-committed transactions on the master to
be lost. Unless you're running synchronous, be sure you have the
workload that can actually accept lost writes.
A
--
Andrew Sullivan
ajs@crankycanuck.ca