On Tue, Aug 06, 2013 at 12:08:57PM -0700, Steve Atkins wrote:
>
> You can't do it by sharing the disk files, at all. The two instances will trash each others data.
>
Right. This is why products that do this sort of hardware fail-over
have something akin to the "STONITH" (Shoot The Other Node In The
Head) program, to make sure the failed box is actually turned off
before the failover box starts up.[1]
Given Postgres's hot standby features today, you don't need this sort
of solution, I'll wager.
Best,
A
[1] Or anyway, they have those kinds of program unless they are
designed and sold by IBM, whose product once made my afternoon more
amusing than I wanted by blowing up the data area on fail over not
once, or even twice, but three times. (This was attributed to
operator error, because the operator had dared to run the failover
sequence.)
--
Andrew Sullivan
ajs@crankycanuck.ca