cancel
Showing results for 
Search instead for 
Did you mean: 

Restore of TruCluster

Andrew Li
Occasional Contributor

Restore of TruCluster

Hi all,

We have a ES45 cluster here and we're trying to develop a disasterous recovery procedure for it.

We have done it for non-cluster machines and never had any problems. Our approach is to do a normal installation (i.e. treat it as a new machine), then restore our backup back onto the system. We can successfully build a minimal cluster machine, but after we restore our backup of the cluster root partition, member1 and member2 boot partition, the machine will not come up normal again. We've had a few goes, the first time was because the HWID of the disk changed (different to the backup), the second time was because the quorum disk was used by a cluster of a different name (which should not be, becuase the qdisk is the same disk of the same name).

When it panics, here are the console messages:
panic (cpu 0): kernel memory fault
clsm: checking for peer confipanic (cpu 1): malloc: invalid size
panic (cpu 1): CNX QDISK: configuration error. Qdisk in use by cluster of different name.

Does anyone know if this approach (dumping old backups onto newly installed system) will work?

If not, how else would you rebuild a cluster system?

And why would the HWID of the disk change?? Shouldn't it be the same??

Many thanks,
Andrew
11 REPLIES
Han Pilmeyer
Esteemed Contributor

Re: Restore of TruCluster

Looks like you already have a disasterous recovery procedure. :)

You can find hints about how to do this in chapter 11 of the Cluster Administration Guide (http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51B_HTML/ARHGYETE/TITLE.HTM).

If you have some additional disk space to spare, I can recommend using Rob Urban's cluster cloning scripts. This allows for much faster recovery under most conditions. This is used by many customers already (http://www.unix-wissen.de/Tru64/clone-cluster.html).

hwid's to (disk) devices are assigned when we first discover new WWID's of a device. When the hardware configuration changed between the time of the backup and the restore, then new hwid's may have been assigned.

Eric van Dijken
Trusted Contributor

Re: Restore of TruCluster

Don't think that restoring old backup on a fresh/newly installed cluster system will work. Don't ask my why, but HWID are not always the same (even when using the same hardware)

In this document Legato describes the recovery of an Entire HP Trucluster (Chapter 11)
Even if yer not using Legato it may be of some help to yer problem.

http://web1.legato.com/infodev/publications/NetWorker/disrec/7.1/disrec_7.1.pdf
Watch, Think and Tinker.
Johan Brusche
Honored Contributor

Re: Restore of TruCluster


Andrew,

The Andrew Li ,I know, works for Alcatel in Australia.
We have written such procedures for Alcatel V5.xclusters. You will receive a copy shortly via mail.

_Johan_

_JB_
Ralf Puchner
Honored Contributor

Re: Restore of TruCluster

The fastest way is by using the cluster creating disk. So if you'll keep your cluster creating disk up-to-date (patches, settings etc.) the whole cluster can be installed within an hour.

Backup only those things which are not part of the creating process (log files, applications directories, settings) and the cluster creating disk ;-)

Help() { FirstReadManual(urgently); Go_to_it;; }
Johan Brusche
Honored Contributor

Re: Restore of TruCluster


a DISASTER Ralf, means "my creating disk was reduced to ashes together with al the rest in the fire"-situation.

-j-
-

_JB_
Han Pilmeyer
Esteemed Contributor

Re: Restore of TruCluster

Not only that, but it's virtually impossible to keep the "cluster creating" disk up to date with all changes. Especially if you have changing disk configurations (which customers using point-in-time copies typically do). Also finger trouble will soon get in the way of an exact copy.
Andrew Li
Occasional Contributor

Re: Restore of TruCluster

Thank you all for your responds, I've finally got a procedure that seems to have worked smoothly.

Eric, Phillip, thanks for the procedure, it didn't work if I just do what the procedure described, but it's been a big pointer and I've made modification so that it fits our environment.
hua_1
Frequent Advisor

Re: Restore of TruCluster

hi, Andrew Li:
very glad to read ur discussion about trucluster restore.
i have a quesiton now , the same as u.
system info: 2GS160+MA8000 trucluster system, systemm image is installed in MA8000,incl. boot,cluster_root,cluster_usr and cluster_var, this system used as a oracle data server.
now we plan to REpalce the MA8000 with a new EVA5000, and system image in MA8000 should be move to the EVA5000.
just like make a backup system in EVA5000. as u did.
could please give me ur procedure?or give me some advice?



great thanks!

Re: Restore of TruCluster

Hi, Andrew Li
Can you share the procedure with me? I have looking for this procedure for long time, but can not find. Now I need this procedure to create the backup for Trucluster system.

Thanks
Huong

Re: Restore of TruCluster

Hi, Andrew Li
Can you share the procedure with me. I have looking for this procedure for long time, but can not find. Now I need this procedure to create the backup for Trucluster system.

Thanks
Huong
Andrew Li
Occasional Contributor

Re: Restore of TruCluster

hua, Bui Xuan Huong,

Unfortunately, when we tried to use the procedure again, it did not work. Doing exactly the same thing works on one cluster but not on another. We've sent our logs to HP, they've had a look and they could not explain it either.

The last reponse from HP was that we should be using a validated procedure by our headquarter in France, and the call have been closed.

Andrew