Are there critical issues using Ignite to DR test a rp7420 and rx4640 that have Serviceguard?

[Rhonda Mokri]

Getting ready for a DR test of a rp7420 (B11.11 v1) and a rx4640 (B11.23 v2) that each are members of a Serviceguard A.11.16.00 cluster. We are only testing the recovery of the primary production servers and not the failover servers at the DR test and plan to use Ignite (version C.6.3.324) make_tape_recovery backups to restore the OS and Data Protector v5.5 for data. Complicating things a bit is that the Data Protector cell manager is on a server that is not being DR tested so I will need to figure out what are the key issues recovering the cell manager to one of the DR servers.

As I have never recovered a HP-UX server before, I do not have first hand knowledge of likely issues which is why I have been poring through the ITRC trying to determine what are common problems and how to avoid them. The key question that I have is whether or not it Would it be more advisable to avoid using the ignite tape as it will mean that I will have to negotiate Serviceguard related issues if I use it to build the DR server? If I use the OS presented to me at the DR test site, I could recreate the data volume groups and filesystems, install Data Protector and restore the data much quicker than struggling with potential Ignite issues trying to recover to larger DR servers. Does anyone think there would be major Ignite issues recovering from a rp7420 to a rp8400 and from a rx4640 to a rx8620?

Thanks!

[Marcel Burggraeve]

To save loads of time make a separate backup ( fbackup, via pc to usb stick, burn on a CD etc etc ) of the Data Protector database.
You really don't wanna scan your backup tapes on a fresh Data Protector installation to create a new database.
Having a seperate database ready for use will save you loads of time.

[Steven E. Protter]

Shalom,

One common problem is not keeping your Ignite version up to date. Version 7.x is released and deals with past flaws in previous versions.

I recommend new archives be made using a current stable version of Ignite. http://software.hp.com

I think there are potential problems going from system to system.

An rx4640 is an entry level machine that is not cell based. An rx8620 is a high end cell based machine with different hardware. You will need to practice this recovery to insure it works. It might but there may be an extra reboot or two to deal with hardware issues. Also there is a finite possibility that the recovery will fail.

rp7420 and rp8400 are a little closer in terms of design but the same risks exist and this needs to be tested.

You could use ServiceGuard continental or metro cluster to create a cluster between rpxxxx machines and another cluster between rxxxxx machines. It's expensive and you need to deal with access to storage or data replication, but its an alternative.

SEP

[Rita C Workman]

Rhonda,

Your ignite version may be an issue. We do subscription DR now and were required to upgrade our ignite-ux to be uniform with their hardware. Like Stephen, I have some concers about the 4640 going to an 8620, but that's why they call it a test.

My biggest worry would be DataProtector and that whole hardware configuration. We always have our admin server that houses Data Protector figured into every test. Depending on your tape library configuration you can have issues there too. It's always easiest to get things up and running if your DR configuration as close as possible to your production environment. For DR testing we put /etc/opt/omni and /var/opt/omni tar'd on a tape and take that with us to rebuild things quickly. Getting your servers O/S up is one thing - restoring it's data is another. Your situation is more complicated, if I read it right, your planning on trying to bring up data protector on a another server (not the Data Protector cell manager server), which would mean your also going to have to change your cell manager info. Not a plan I would like to undertake. At everyone of our DR tests, we restore the admin (data protector) cell manager first, and get the tapes ready. Then we get the prod boxes up and start restoring data.
For DR, even though we are also in a clustered environment, I remove that from the equation. Failover is nothing I'm concerned about in a DR. For us - failover is production world only, but not required in a disaster.

Another big point, is who your subsciption provider is. Some are big, but not as helpful as others. So hope you have a good provider.

Rgrds,
Rita

[Rhonda Mokri]

I appreciate the feedback from everyone but need to understand what the implications are for the server that I am going to install the DP cell manager onto at the DR test. The server is a DP client in our production environment so it already has the directories /etc/opt/omni and /var/opt/omni. If I understand everyone correctly so far, the cell manager would be easier to recover if I copy the /etc/opt/omni and /var/opt/omni directories - what hazard is this to the recovery of the original client server that I AM DR testing? Once I copy these directories to the DR server, do I still need to recover the IDB database from the daily IDB backup tape or would this be unnecessary?

Rita, in your response you noted that for DR you remove clustering from the equation - could you explain how?

Thanks.

[Rita C Workman]

Rhonda,

You say your going to bring up DP on a server that is a 'client' server originally during your DR test. A 'client' is not a Cell Manager. Generally you push out agents to your client machines - you don't load "it all" on the clients. Your Cell Manager is configured to be the Manager - not your clients. So what I was saying (albeit poorly) was that you will have put all your cell manager info to that client, AND you will have to reconfigure your Data Protector so it will now view that "client" as the NEW CELL MANAGER. Now that method is not my preference - so let me know how you do on it.

For your Cluster question.
We have all production boxes rebuilt via the ignite. Actually, we use HP's disk, so they even set up my all my volume groups for us. [HP offers a variety of DR options - we use them for quick restore ]
All my packages are failover tolerant (as is cron, environment & users), so when the boxes are built at DR...I just turn off starting up the cluster in the /etc/rc.config.d/cmcluster AUTOSTART_CMCLD option. Restore all our data from tape (at this time) and bring it up using the same commands that the package uses. For DR - I don't care about MC/SG or heartbeats. All boxes up, all applications started locally on box.........No Cluster Running.

Rgrds,
Rita

[Bill Hassell]

Probably the largest task has not been mentioned yet: very different I/O cards in each machine, specifically LAN and fibre and SCSI. The drivers for the target (DR) machine must be present in the current machine even though they are not used. If you have multiple LAN cards, you will need to configure them to match your current machine so that IP addresses and routes will work correctly (plan all this ahead of time). Similarly, you'll need to match SCSI devices. If you are using fibre, especially through a SAN switch (and perhaps some link-enhancing software like PowerPath), you have a lot of work to do. You'll need to collect all the details about your current setup including the SAN, and then configure the DR SAN to match your DR system, followed by connecting all the LUNs to the correct device files for each volume group. And don't forget alternate paths.