WAL Archiving, warm standby - edge cases questions:

From: Grzegorz Dostatni <dostatnig(at)yahoo(dot)com>
To: pgsql-admin(at)postgresql(dot)org
Subject: WAL Archiving, warm standby - edge cases questions:
Date: 2008-03-10 22:23:06
Message-ID: 322877.37825.qm@web30401.mail.mud.yahoo.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

Assume we have a primary database server and a warm
standby. WAL files get shipped over to the standby and
applied as they arrive.

Questions:
1. I know the recommendation is to fail over to the
standby and re-configure the primary as a new standby.
Is it possible to do that without any data loss? I
know I can copy un-archived WAL files to the archive
area, but is that guaranteed not to result in any data
loss? I suppose that would require turning on the
option that forces a sync of the WAL file before
finishing a transaction. I guess I'm asking as to the
recommended procedure that minimizes data loss /
outage window when purposely failing over to the
standby.

2. How would you deal with archiving the new master DB
server? Should there be a second WAL archive area? If
you copied un-archived WAL files, and the DB server
tries to archive them - you're going to have a
problem. The documentation says that the archive
script should not overwrite files silently. Can those
un-archived WAL files be copied to the pg_xlog
directory directly? Will postgresql look in there
first before activating the restore_command?

3. If you want to have an automatic failover to the
standby - does that mean that your standby needs to
archive WAL files as well? Otherwise you'd have a
window where the current primary is not doing any
archiving. OR you'd have 2 archive copies of your WAL
files.

Mostly I'm looking for input from people who are
running warm standby in production environment,
especially with respect to any gotchas. I did manage
to successfully setup a standby server and I am able
to fail over to it. I found the messages on this list
that recommend failing over periodically, and I am
wondering if people are accepting data loss for what
should be a maintenance event.

Thanks in advance for any input. As I mentioned
before, I believe I have the base case covered. Just
trying to play with a bit of *what if*.

Greg

____________________________________________________________________________________
Never miss a thing. Make Yahoo your home page.
http://www.yahoo.com/r/hs

Browse pgsql-admin by date

  From Date Subject
Next Message Scott Marlowe 2008-03-10 22:43:11 Re: Reindex + relation doesnt exist
Previous Message Peter Koczan 2008-03-10 18:58:17 Re: Newly converted to 8.2.5 and getting this error