On Tue, 20 Sep 2005, John Madden wrote:
I don't completely agree here, there are ways to do snapshots without any
downtime for example in a SAN environment or using filesystem or LVM based
snapshot methods. However, it does not mean your snapshots are really
consistent from the aplications point of view.
And that's the big question: How inconsistent can things be without
shutting down cyrus?
How consistent does it need to be?
We dump the mailboxes.db to a flat file at midnight, then start a regular
filesystem backup of the system. In the event of a complete system
failure or disaster recovery, our plan is to restore everything from the
latest backup then (most likely) rebuild the mailboxes.db from the flat
file. Our mail store is on a SAN, so it seems unlikely that we'll ever
reach this point unless the building falls down.
The more common case is individual mailbox restores. I do a fair amount
of mailbox restores at the request of users that have accidently deleted a
message. I simply create a new subfolder in the user's INBOX named
RESTORE and recover all the message files and the cyrus.header file into
it. Then I run reconstruct and quota on the mailbox to fix things up.
This has always worked without fail for me.
Andy
----
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
