Re: Zimbra Backup Strategies

[Matt Mencel <MR-Mencel@wiu.edu>]

We've got similar problems here too.  Because of the way that Zimbra creates their backups...with LOTS of small files, backup to tape is unrealistic, and trying to do a full restore for DR would be impossible.  What we're looking at now is some way to do SAN snapshots, or block level backups, or RAW backups.  Something that goes deeper than the file level.  So the whole partition is backed up as an image...nightly or weekly we haven't decided.  This would only be for DR though.  We will not be restoring any data from this backup.  If we want to be able to go further back than 30 days we'll just make the zimbra backup scheme go back further.

As far as our setup, we have 5 mailbox servers and about 15000 users (3000 faculty/staff and 12000 students).  It varies a bit, but generally they look like this.

Primary Storage (FC SAN) - 30 days of user mail - 25-30GB per server in use
Secondary (HSM) Storage (SATA SAN) - Everything older than 30 days - 200-250GB per server in use
Backup Storage - 30 days of backups - 300-350GB per server in use

HSM is a nice feature.  We run it nightly.  If you've never run it before, the first time will take awhile to archive off all the old stuff.  But nightly maintenance after that is very quick.  Usually just a few minutes...last night's took 10-12 minutes for us.  We've watched the SAN activity, and our primary FC storage is always very busy, but the secondary SATA storage is very lightly used....so it works very well.

We're actively pursuing new backup software right now and are looking at several products.  Currently we use Syncsort's Backup Express, but we're also looking at Netbackup, Bakbone Netvault, and a few others.

Matt

----- "Steve Hillman" <hillman@sfu.ca> wrote:

> ----- "Tim Ross" <tross@calpoly.edu> wrote:
> 
> > With our current setup,
> > the fastest transfer rate we could get would be approximately 2
> > Gigabit/sec from our Zimbra servers to the SAN storage.  We are
> > thinking that we would probably do a full backup once a week with
> > incrementals for the rest of the week.
> > 
> > What backup strategies/hardware are others on this list using?
> 
> We're just in early pilot right now, so backups are still small and
> easily manageable. I'm not thrilled with the idea of having to have
> 'disk x 2' just to hold a backup image that we then spool to tape, but
> it appears to be all we can do for now. So what we've done to start
> with is use an iSCSI mount to a Sun Thumper for the backup area. We
> had originally hoped to put our main message store on Thumpers as
> well, but that's proving problematic due to reliability issues with
> the Thumpers and the bleeding edge Solaris version we have to run to
> get iSCSI.
> 
> So we've roughly broken it down like this (or will, once we're done)
> 
> - /opt/zimbra -- iSCSI mount to NetApp
> - /zimbradata/volume0 -- first volume for msg/index storage, also
> iSCSI to NetApp
> - /zimbradata/backup -- iSCSI mount to Sun Thumper
> - /zimbradata/hsm0 -- not set up yet, but iSCSI mount to Thumper where
> msgs older than 30 days will migrate to
> - /opt/zimbra/redolog -- to do, move this to a separate iSCSI mount on
> a different NetApp
> 
> Right now, a NetBackup client on the Zimbra servers backs up all of
> these volumes to tape over the net.
> 
> Is anyone actively using HSM? If so, what are you HSM'ing to? I assume
> the HSM process runs overnight - how long does it typically take? How
> much data lives on HSM vs primary storage?
> 
> 
> -- 
> Steve Hillman                                IT Architect
> hillman@sfu.ca                               IT Infrastructure
> 778-782-3960                                 Simon Fraser University
> Sent from Zimbra