Re: Warm standby problems - Mailing list pgsql-admin

From David F. Skoll
Subject Re: Warm standby problems
Date
Msg-id 4AE60742.9090607@roaringpenguin.com
Whole thread Raw
In response to Re: Warm standby problems  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: Warm standby problems
List pgsql-admin
Tom Lane wrote:

> No; there's no WAL change between 8.3.7 and 8.3.8.  What seems more
> likely is that you're somehow shipping the WAL files before they're
> quite finished.

I doubt it.  Our archive_command works like this:

1) scp the file over to the backup server as root.  It's stored in a
file named <name_of_wal>.NEW

2) When the scp finishes, ssh to the standby server and chown it to postgres.

3) Then, ssh to the standby server and rename the <name_of_wal>.NEW
file to the real WAL filename.

Only after step (3) does our standby script even see the file.

So unless there's a possibility that the master server invokes our archive
script before the WAL has finished being written, I can't see how this
could happen.

Regards,

David.

pgsql-admin by date:

Previous
From: Tom Lane
Date:
Subject: Re: Warm standby problems
Next
From: Anj Adu
Date:
Subject: Re: pg_attribute size