John Scalia <jayknowsunix@gmail.com> writes:
> Hi all,
>
> I'm wondering as I've never seen anyone doing this if I can/should
> setup WAL archiving from the primary server to 2 different standby
> servers. These machines are configured to be streaming replication and
> I've got the archive command on the primary doing an scp to both
> standbys, but for some reason the second one isn't getting
> anything. What the archive_command looks like is "scp %p
> user@standby1:/<dir>/%f; scp %p user@standby2:<dir>/%f". That's not
> wokring correctly, but I'm wondering if it's even necessary for the
> 2nd standby?
Look in your server log for errors on the 2nd scp.
Also check your $PGDATA/pg_xlog/archive_status directory for a pile-up
of *.ready files.
If your archive command isn't fully operational and is returning
non-zero to the caller, then Pg thinks archiving is failing and will
*not* recycle the related xlog files.
I'd suggest also you get on the server box as user postgres and make
sure you can manually ssh without being prompted for a password and/or
asked to verify host keys... fix mismatching lines in ~/.ssh/known_hosts
file etc.
As for whether or not you should be sending WALs to both instances, it
depends on what you need to accomplish.
HTH
>
> Just askin',
> Jay
>
> PS. There are two standby's to deal with the problem of synchronous_commit and the failure of one of the standby's.
--
Jerry Sievers
Postgres DBA/Development Consulting
e: postgres.consulting@comcast.net
p: 312.241.7800