Showing results for 
Search instead for 
Do you mean 

Cloning a Cluster

SOLVED
Go to Solution

Cloning a Cluster

I have to upgrade a Tru64 cluster from V5.1A to V5.1B, consisting of 4 members in the near future. The client, and I are nervous that something goes wrong and they loose the entire production environment. I have two questions:

1) How safe is this rolling upgrade. Can it destroy anything and prevent the cluster to boot up correctly?

2) Is there a tool or script whereby you can clone the entire cluster root, usr and var and boot disk to another set of disks in case something goes wrong, we can then boot this set?

I hope there is somebody that could help me on this.
6 REPLIES
Honored Contributor

Re: Cloning a Cluster

rolling upgrade is without a problem if you read the documentation carefully, check the requirements (firmware, hardware etc.) and create a backup.

A cloning instruction can be found on www.tru64unix.compaq.com -> documentation -> best practice.

But a backup is also a possiblity.
Help() { FirstReadManual(urgently); Go_to_it;; }
Frequent Advisor

Re: Cloning a Cluster

Ralf,

I was told an entire cluster restore from backup isn't possible. You'd have to rebuild the cluster again using the system seed disk which (hopefully) you still have around. Am I missing something.

Francois,

Before you start I'd suggest you look at the following sections in the 'TruCluster Server - Cluster Administration' manual:

11.1.5 - Backing Up & Repairing a member's Boot Disk
11.1.7 - Recovering the Cluster Root File System to a Disk Known to the CLuster.
11.1.8 - Recovering the Cluster Root File System to a New Disk.

Vic
Honored Contributor

Re: Cloning a Cluster

victor,

a backup is possible but it is much faster to use the original build disk instead!
Help() { FirstReadManual(urgently); Go_to_it;; }
Valued Contributor

Re: Cloning a Cluster

Hi,

The rolling Upgrade process itself works like a charm.
But the gotcha's for version changes are the firmare compatibility issues.
With a four node cluster, assuming you can take down two nodes, a better strategy would be to do an install on the 4th node and install the new version.

The advantages of this are :
if the system crashes, only the 4th node is affected. ( I'm assuming the firmware on all four nodes are at the same level..)

Apply all the patches for 5.1B, add the second member in. Check networking etc, and if there are no problems you can proceed.

Not the fastest but most sure method to avoid any potential problems with a production system.

Hope this helps,
Regards,
Orrin.

Re: Cloning a Cluster

Thanks everybody, but it seems that a backup will be the only option at this time. This is what I got from the best practice I was referred to....

"Configuration Cloning is not supported in a cluster environment. Do not save the configuration of a cluster member because individual cluster members share disks and files, and the information saved in the CDF relates to the entire cluster, not to individual members."

Esteemed Contributor

Re: Cloning a Cluster

If you can spare the additional disk space that it requires, then a very good solution is to use the cluster clone scripts that Rob Urban created. You can get them here:

http://www.unix-wissen.de/Tru64/clone-cluster.html
//Add this to "OnDomLoad" event