On Wed, 24 May 2006, Dan Trainor wrote:

D Canfield wrote:
Perhaps I'm missing something obvious, but I would have assumed Maildir to be *less* prone to consistency errors. I may not be able to say "here's exactly what the server looked like at 11:59 on 5/23" but I don't see how the backups would be invalid. I could see issues with mbox storage where you'd start reading a large mailbox, and the server rewrites half the file before you finish with it. Again, I may be missing something...

In any case, how would the LVM snap help me with the original problem? Wouldn't I still need to backup millions of small files even with a snapshot?

Thanks
DC

Ryan Novosielski wrote:

You'd almost need an LVM snap anyway, wouldn't you, for consistency's sake? Seems to me Maildir is particularly vulnerable to inconsistency if backed up while the server is running.

 ---- _  _ _  _ ___  _  _  _
 |Y#| |  | |\/| |  \ |\ |  | |Ryan Novosielski - User Support Spec. III
 |$&| |__| |  | |__/ | \| _| |[EMAIL PROTECTED] - 973/972.0922 (2-0922)
 \__/ Univ. of Med. and Dent.|IST/AST - NJMS Medical Science Bldg - C630

On Wed, 24 May 2006, D Canfield wrote:

I'm just curious how other users are backing up maildirs using bacula. We're using cyrus as our mail server, so we've got about 162GB of data stored in 4.4 million files, or an average of 37K per message. Most other products I've tried (both commercial and open source) have not been too happy with this arrangement and end up requiring vast amounts of disk space and overhead to index this kind of mail spool.

Does anyone know if bacula is any better with this? If not, I'm thinking my best resort would be to have each user's mail spool gzipped before backing it up (we don't restore individual emails for users, so this kind of "resolution" is sufficient for us). Any ideas on a good way of doing this that won't require me to run an external crob job to do the zipping and require almost twice the disk space?

Any other insights would also be appreciated.

Thanks!
DC


Hi -

I think the point that Ryan is trying to convey is that an LVM snapshot would be much quicker than any other product out there, including Bacula. You could then archive that snapshot, and be sitting pretty.

I think Ryan suggested against Bacula for those 4.4 million files because it would take a while, leading to inconsistancies. At least that's what I think ;)

THanks

I would think with all of those files -- granted I don't know the internals -- you'd run into trouble. Aren't the index files kept someplace other than the individual messages? If so, if it backs up the mail dir one minute and the index another minute, you already have a confused mailbox. In UNIX mailbox format, you backup the mailbox as it is -- it is entirely self-contained. Of course, either case allows you to have a lack of consistency across mailboxes... User A's mailbox might be backed up at 2:00 and user B's might be backed up at 2:45.

With LVM, at least you've frozen the snap. That doesn't necessarily help your speed situation, but low speed + high rate of change is asking for it, IMO.

 ---- _  _ _  _ ___  _  _  _
 |Y#| |  | |\/| |  \ |\ |  | |Ryan Novosielski - User Support Spec. III
 |$&| |__| |  | |__/ | \| _| |[EMAIL PROTECTED] - 973/972.0922 (2-0922)
 \__/ Univ. of Med. and Dent.|IST/AST - NJMS Medical Science Bldg - C630



-------------------------------------------------------
All the advantages of Linux Managed Hosting--Without the Cost and Risk!
Fully trained technicians. The highest number of Red Hat certifications in
the hosting industry. Fanatical Support. Click to learn more
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=107521&bid=248729&dat=121642
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users

Reply via email to