"Me too." This thread is unanswered; what did you end up with on your own?
Our enterprise backups are done with Networker, but the concepts are the same.
So far, my strategy looks like:
Never back up the following at all:
store (also skip any other primary/secondary volumes)
/opt/zimbra/redolog is on its own pair of spindles for speed and isolation. It gets backed up along with the rest of the system. Yeah, redundant, but also very important to retain.
/opt/zimbra/backup lives on the other end of a 5km fibre channel link. So we have "immediate" offsite backups. I've modified zimbra's crontab to include:
It would be nice if Zimbra had a supported hook for "post-backup commands" so that you could push to tape immediately upon completion of zmbackup. I'll write up an RFE. Fo now, I think it's slightly more maintainable to alter crontab (thereby screwing up the output of zmschedulebackup) than to edit zmbackup itself.
0 1 * * 6 /opt/zimbra/bin/zmbackup -f -a all && /opt/zimbra/networker-backup.sh full
0 1 * * 0-5 /opt/zimbra/bin/zmbackup -i -a all && /opt/zimbra/networker-backup.sh incr
And the networker-backup.sh script reads:
# Call with argument either "incr" or "full"
nice save -q -l $* -s nwserver -g EmailGroup -b Email -e '+14 day' -S -i /opt/zimbra/backup 2>&1 | egrep -v '^save:/opt/zimbra/backup.*level=.*files|lost.found/: Permission denied'
-g Mark this backup a member of that group
-b Send backup to this tape pool -- by default, our backups go to offsite disk and are cloned to tape, but in the case of Zimbra that's already been done for us, so we only make one copy to tape
-S Do not store index entries for this saveset, allowing only restores of the whole blob. This speeds things up and saves index space/complexity on the backup server. Your backup system probably has some analogous option.
And finally, with the exception of the directories named above, the entire system gets backed up like any other.