catalog recovery

Clustered master can only be recovered to another clustered setup. Single node cluster is fine.
Virtual hostname and mountpoint(s) must be the same.
Verify that catalog backup folder is available in correct path: /catalog_backup02
Local nodename can be different.

See this TN:  http://www.symantec.com/docs/TECH58358

Thanks Marianne, can you further elaborate on few other things below.

 

While installing Installics (PBX) it asks for the machine name, does that need to be cluster name or physical node name. ( i have entered physical node name )

After PBX while installing netbackup 6.5 what server name do we need to specify, do we enter phyical node name or cluster name in netbackup setup.

While taking to tech support in symantec, we have been told that the PBX can be done with physical. Please advice further.

We are testing on test environment to get the test environment to 6.5 configuration and then take it to 7.1.x upgradation and catalog recovery is part of that so that we can have a real scenario before upgrading the production. ofcourse we have NOT installated the tape devices and tape library or media servers or all storage units in the test master server as the catalog is on disk and same disk is visible to production.

We are trying to make a single Storage unit with basic disk as it was in production master server which was used for catalog backup.

Please advice further.

Wizard, always use the cluster name. You do not want to be using nodenames at all, if the system fails over you want to continue using the same name, the Cluster name!!

 

NBU 6.5 is so old that I cannot remember PBX installation anymore.

Best to read up in Installation as well as Cluster/High Availability Guide.

All NBU 6.5 Manuals: http://www.symantec.com/docs/TECH52878

I do remember that I have always installed PBX as standalone (non-clustered). This is also how PBX is installed in 7.x cluster.
I sort-of remember that NBU 6.x wants to install PBX as clustered. Cannot remember how to bypass that.

The only place where virtual name MUST be entered is towards the end of NBU installation where name of 'configured' host must be entered. You will then be prompted to start cluster_config.
Make sure that you have all the answers to prompts as per worksheet in HA guide. Mountpoint for 'shared drive' must be the same as production cluster to ensure successful catalog restore.

 

PS: Any reason why you are still on 6.5? It will reach EOSL in 2 months time.
If you upgrade production to 7.x in production, you will install same vesion in DR.
You will see that cluster installation in 7.x is a breeze...

Agreed Marianne. I've seen people install PBX as clustered, but it doesn't have to be, unless worried that PBX processes may go down. If that's the case have a mechanism to restart it.

PBX is always running on both nodes normally, doesn't need to be started just on failover.

This is a test environment, where we are testing to build the master on a lblank server , then recover catalog in it as the master server and then take it to 7.1 if this is success then we would be doing the production master on 7 version as we need to have a good roll back plan as well in case anything on upgrade goes wrong.

Hence we need to have a clone of master server built in isolated testing network to have this as a proof that the catalog recovery can be made. 

Right now we are stuck at a point where we bring the test master server on same level with netbackup installation and patch, make the catalog myount point mounted on test server and run a berecover -wizard and it says cannot locate the image files on the location. i was doubting it was cluster which caused the pain as we were doing it on single test server as it was unable to locate the image files. 

Moreover to support my point that the cluster should not be a problem with that, we have partial catalog restore as well where  clustered master catalog can be restored on a different master which is NOT clustered (only the image database though) but this should not be stuck at a point where it says cant locate the image file when we specify the path for DR file imported on test master.

any more  pointers ??? 

According to this technote, you cannot recover a clustered master onto a single node non-clustered DR system

http://www.symantec.com/business/support/index?page=content&id=TECH58358

 

It Can be done in partial recovery. . .(image database can be and then we can build on relational database ) 

 

According to file attached as pdf which is a guide to site disaster recovery, please look on page 4 under 2.2 

"Because the server topology is not affected by the recovery process the Master Server topology at the DR site doesn’t need to reflect the topology at the production site (i.e. a catalog backup from a clustered Master Server can be restored to a standalone Master Server at the DR site)."

Now ... read this tech note http://www.symantec.com/business/support/index?page=content&id=HOWTO72778#v15108283

 

when we give the path of dr file it says cant find images, when it should go at next step, and ask for the full recovery or partial recovery, but atleast should find the files. 

The Shared NFS mount point is given exactly the same name and location as the production

any more pointers ???????? 

 

I always consider recovering catalog to a non-similar environment not a true DR. 

>> when we give the path of dr file it says cant find images, when it should go at next step, and ask for the full recovery or partial recovery, but atleast should find the files

Can you post a screenshot