Dhaval Shah wrote:
> I am in a situation where we have to deploy a hot standby to a
> postgres db server. There is no custom tablespace and all data files
> are in $PGDATA
>
> I was thinking of using PITR
> []http://www.postgresql.org/docs/8.1/static/backup-online.html] to
> achieve that and here are my thoughts:
Same here.
>
> 1. Continuously copy WAL files to the standby.
Right.
> 2. The standby is always in "recovery" mode, that is whenever it gets
> a WAL file, it recovers to that WAL file.
This I decided against, only because traffic does not warrant it. The
idea at this point was to let the WAL files pile up in the offsite
location and then work out a periodic schedule for running them, but it
would be more like once/week than continuously.
My thinking was that maintaining a 60-second failover is not really what
I'm after. If the main server goes down, we would likely wait and hope
for up to 10 minutes before starting to make DNS changes. During that
time we could always run a recovery of the WAL files and have the
standby easily ready in time.