Migrate NSR Server to MSCS Cluster

[lill]

We are trying to migrate our NSR Server 6.1.1 from a single W2K SP2 machine to a MSCS Cluster formed by 2 physical W2K SP2 machines.
We tried the migration a few weeks ago. I will not describe in details the migration but the important steps were:
nsrck -L6 old server
copy of the entire nsr directory (except nsr\bin) from single machine to cluster disk (shared disk on cluster))
nsrck -L6 new server

Everything seemed to be successful at least at the beginning. Indeed, we were able to take a backup of the cluster nodes (physical & virtual machines). We were able to backup a first savegroup. But when we started a second group then we had a lot of problems.
At once, all resources within NetWorker cluster were frozen, even writing and moving forward on tapes mounted on drives.
We had to perform a fallback to the single machine.

Has anyone performed such a migration ?
Thanks.
Regards,

 

[john7]

How did you set this up?

Where are the tape drives attached?

Where are the indexes?

Are you trying to failover the Backup server?

need more info. What you want to do can be done as long as you know and accept the limitations of the config.

John7

 

[lill]

About the set-up: I followed the Legato Installation Guide: "Installation of NetWorker Server in MSCS". The basic tests are perfomed befoe the migration were successful.
Once again, to migrate the old nsr server to the MSCS cluster roughly I did:
nsrck -L6 old server
copy of the entire nsr directory (except nsr\bin) from single machine to cluster disk (shared disk on cluster))
nsrck -L6 new server

The NetWorker Server is attached to a robot STK containing DLT7K tape drives.

The indexes on the MSCS Cluster are in a shared disk since we want to put the NSR Server in the MSCS Cluster.
So, yes we want to failover the backup server between 2 machines via MSCS.

Thanks.
Regards

 

[john7]

OK, Tape drives/Robots can not be failed over. Most OS's issue a scsi reset when the failover happens and this causes the robot and drives to reset. THey wil eject tapes. THis will cause the Networker server to get all screwed up. When you cluster the server you should only cluster the indexes and then have storage nodes which handle the backups for you.