Ignite Tactics

[Mike Hassell]

Fellow HP-UXers,

So I've just gone through my first full scale disaster recovery test at a hot site and I've learned a few things, which I'd like to review with everyone.

1. Do you include more than just vg00 in your make_tape_recovery tapes? if so than why?

2. Do you install your backup software (Legato Networker, Omniback, Veritas) in /dev/vg00 and include this on your make_tape_recovery tape?

3. Do you keep scripts on hand that create all your specific lvols?

My DR test was successful, however my recovery time could use some work. I included all of /dev/vg00 as well as specific mount points (Legato and Sybase) to keep from having to reinstall these apps. One thing that I ran into was the ignite tape trying to create all of my /dev/vg02 lvols as well, I assume this is the case because I have included specific mount points that are contained within /dev/vg02 (/dev/vg02/networker, /dev/vg02/sybase119, etc.)

I'm in the process of updating my procedures and so fourth and would like to review these details with forum members. Your thoughts? Thanks.

-Mike

[Christopher McCray_1]

we use make recovery to back up vg00 only. We install our backup products on vg00 and even in some instances other applications such as oracle so that make recovery will restore them as well. Congratulations on a successful test and good luck in the future. I hope my info was of some help.

Chris

[Alan Riggs]

1. I take only vg00.

2. I install Omniback on vg00. On the backup server, I also install oracle on vg00. On other servers I do not install non-OS software in vg00.

3. I keep mapfiles and the vgdisplays of all volume groups in /root. Between that and /etc/lvmconf I feel comfortable restoring any particulat VG/LV definitions as required. I do not have automated scripts for the purpose.

[Darrell Allen]

Hi Mike,

1) no
2) yes
3) that's a good idea that can save valuable time in a true DR. Be sure to keep them updated. Keeping them in vg00 obviously is a plus.

Congratulations of your successful test.

Darrell

[Sridhar Bhaskarla]

Mike,

1. No. I don't include anything other than vg00. Because I can easily import other volume groups once I have my vg00 restored. I keep vg00 as minimal as possible so as to shorten up the make_recovery process.

2. No. Most of my additional software but related to System Administration goes into vg01. So, when I import other volume groups, everything will be imported back.

3. Yes. I do have scripts that takes backup of important configuration files into a single archive and keeps it on a central repository. This includes the map files of the volume groups.

After restoring vg00 from make_recovery, you an use these map files to import the volume groups back. Most of the system configuraiton is copied back through make_recovery.

-Sri

[Patrick Wallek]

Mike,

First off, good job on a successful DR test. It is important for you, and your management, to know that your procedures work. While tweaking those procedures now, make sure you don't break them.

1. No, I just backup vg00. The reason - All of my other VGs are just too damn big to fit on a tape. Besides, that's what I've got my other backups for.

2. Yes, I do have my backup software, database binaries and other critical components installed in VG00. When I do have to do a DR test, once I have vg00 recreated, I just have to restore the data and don't have to install any other software.

3. No, I don't keep scripts to recreate my LVOLs, although I probably should. I do keep a print_manifest printout handy though, so I know exactly what things look like.

The point one or two of the other responses made about being able to reimport the VGs is fine if you are recovering the OS on your production machine. But if you are at a DR site and your are restoreing you whole environment, then it is kind of difficult to import your VGs other than VG00 when they don't exist yet.

[harry d brown jr]

1. On my R&D servers, yes, in production or development (we consider dev to be production anyways) no. I usually make my R&D servers with just vg00, mostly because I build then destroy the servers quite often, and I usually don't want to screw around with building things from scratch - start the recovery and then go for a beer (S).

2. Yes, in production our backup software is part of vg00 that get's tossed into the basic root directories that make_recovery will pick up.

3. Not only scripts, but all systems have their lvm info (a config file for our lvm create scripts) on a trusted system, that allows us to recreate them in emergencies. This hasn't been much of an issue, and most definitely won't be when we migrate all of our systems to an EMC SAN over the next year.

[James R. Ferguson]

Hi Mike:

1) I only include vg00 in my Ignite ???make_tape_recovery???.

2) We use EMC2???s EDM (Enterprise Backup Manager) as a backup for both Unix and NT. The client-resident software exists in a non-vg00 logical volume group on my servers. During a disaster recovery we recover the EDM server as one of our first tasks, and then push (reload) the client software. Following the installation of the client component, we then proceed to recover all other non-vg00 filesystems.

3) Yes, I have scripts that define the filesystems on each server. I find that scripts make excellent documentation. While the actual disk devices are going to differ at a remote disaster recovery facility, it is easy to ???stub??? in the appropriate device files and have the remaining ???logical??? definitions ready to configure with commands pre-composed.

Regards!

...JRF...

[Mike Hassell]

Thanks to everyone for the great replies. I will be sure to update my documentation as well as create some scripts that will be included in our 'make_tape_recovery' tapes for the lvol creation as I have so many of them.

JRF,

Quick question about EDM, just wondering what your overall thoughts were on the product compared to your basic backup utility. We are currently using Legato Networker for our HP9000 servers, however we have been looking at EDM for our enterprise wide backup solution as 90%+ of our critical data is on EMC Symms.

Thanks for all the info everyone!

-Mike

[A. Clay Stephenson]

Hi Mike:

1) No, I only include vg00 in the make recovery. I want this phase to go as smoothly as possible.

2) No, in fact /opt is not even a part of my vg00. We use OmniBack and my method to get the applications back is to maintain a swdepot with all the applications and patches currently in use. This is backed up weekly using cpio (could use fbackup). To get all the applications back, I restore that swdepot and then do a swinstall.

3) Yes, I do have scripts to recreate all the vg's and lvols. The system actually creates this script every week and stores this data
in vg00 before the weekly make_tape_recovery.