Need help planning Disaster Recovery

[mhlevy]

Hi all,

I've been tasked with coming up with a DR plan, in order to get our Zimbra system up and running as quickly as possible after any sort of crash. Of course, there is $0 budgeted for this, the only assets I can use are my time and existing resources. I do have access to a Windows server share for storing backups.

We're currently running Zimbra Network Edition 6 (Release 6.0.10_GA_2692.SLES11_64_20101215162357 SLES11_64 NETWORK edition), though I will be updating the Zimbra to v7 before too long.

Our environment is a 4 server system: Ldap, 2 x MTA, and Postoffice servers.

The number 1 goal is getting the system back up as quickly as possible after any sort of failure.

I was wondering if I can use a backup solution to back up the /opt/zimbra directories on the servers, and if need be, restore them to replacement servers (with the same OS, and IP configurations), rather than reinstalling the Zimbra software on the servers. I know that pipes and sockets are used in the spool directories on the MTA servers, so that pretty much leaves me with cpio as the backup and restore utility.

I realize that in order to get a clean backup, I will need to issue a "zmcontrol shutdown" to halt the zimbra services. I'm wondering just how often these DR backups should be scheduled.

I am planning on copying the backup directories on the postoffice and ldap servers to the Windows server at lease once a day, probably using rsync. If I do so, should I exclude those backup directories when performing the cpio DR backups?

Am I going in the right direction, or is there a better way that I'm just missing? I have seen the WIKIs on the subject, "http://wiki.zimbra.com/wiki/Network_Edition_Disaster_Recovery" and "http://wiki.zimbra.com/wiki/Ajcody-Notes-Multi-Server-Restore-DR" but they haven't been all that helpful to me.

I just thought of one other question... Regarding the backup directory on the ldap and postoffice servers: The DR wiki talks about how a full backup will backup both the ldap account information and the mail store, but since these are now on different servers, if I wanted to force a full backup through the admin console, am I also going to have to force a full backup on the ldap server using the CLI? It appears that will be the case.

Thanks in advance,

Mark

[bdial]

one thing you could do is backup /opt/zimbra every night, then backup the redologs throughout the day if you'd want to have a DR plan other than restoring form the night before. what i'd do make a script that would maybe rsync /opt/zimbra to a remote location while zimbra is running, then shutdown zimbra, then do another rsync to catch whats changes. if you're doing this in the middle of the night chances are that it's very little thats changed.

Then if you backup /opt/zimbra/redologs periodically through the day, you could replay those over a restored system to get back to where you want to be for that day.

zimbra network edition backups run on all servers that it needs to. so if you've scheduled a backup to run nightly, it'll run on your mailox server as well as your ldap server.

[mhlevy]

Thanks for the reply, and I'll take this into consideration.

It looks like the direction I may be going is to use MondoRescue in order to do a bare metal restore of the Linux box, rather than a fresh install. And planning to do a cpio backup of the /opt/zimbra directories on the ldap and mta servers. Finally using a combination of cpio and Zimbra's backup for the actual ldap and account information.

Thanks!

Mark