Terry Khatri wrote:
> I need to do the subject setup using PostgreSQL, where when one node
goes down another takes over
> without manual intervention, the way it is done in Oracle using grid
infrastructure and in SQL Server
> 2012 using their Alway-On feature
>
> My question is, has anyone done that ? can anyone please help me
implement it, I have googled on the
> subject, some people have done that and EnterpriseDB offer their
services to implement it but there
> are no details how actually to do it.
>
> Your step by step guide will be much appreciated.
No step-by-step guide, but we have had such a configuration
running for years.
One gotcha I remember was that we had to modify
/usr/share/cluster/postgres.sh to wait for a while
if the regular checks got
FATAL: the database system is starting up
because otherwise you could end up with a disabled
service after a failover (+ crash recovery).
We had a 2 node cluster and had problems with it
because of the quorum disk. For unknown reasons storage
had occasional "hiccups" that caused the quorum disks to
be unavailable for short spells, which in turn caused
unnecessary failovers.
I'd recommend a three node cluster.
Yours,
Laurenz Albe
