Process for removing/restoring NetBackup 7.5 in HP Serviceguard to a Standalone configuration?

[JL_DXC]

Hello,

We are testing what it would take to remove/restore NetBackup 7.5 in HP Serviceguard to a Standalone configuration (first in our lab environment, then in our live/test environment).

LAB TESTING:

I made a tar of /usr/openv before doing anything, then I halted the netbackup package (cmhaltpkg netbackup) and then removed the netbackup pkg ("cmdeleteconf -fvp netbackup" from the hpux 11.31 master server:(

# cmdeleteconf -fvp netbackup
Completed the package deletion

# cmviewcl -p netbackup -v
cmviewcl: "netbackup" is not a configured package name.

It seemed to work, but when we start netbackup in standalone mode, we see this:

# netbackup start

...
NetBackup will not run without /usr/openv/db/bin/NB_dbsrv running  <==

...

# ll /usr/openv/db/bin/NB_dbsrv
lrwxr-xr-x   1 root       sys              7 Sep 13  2015 /usr/openv/db/bin/NB_dbsrv -> dbsrv11

# /usr/openv/db/bin/nbdb_ping
Database [NBDB] is not available.

I stopped NBU, restored the tar, and restarted NBU but received the same errors about nbdb.

So I ran a "truss" against the "nbdb_ping" to see where it's failing; I saw several "open" calls to the bp.conf file, so after reviewing it, it seems pretty obvious that the problem with nbdb_ping is because VXDBMS_NB_DATA is pointing to a non-existent filesystem (i.e. a clustered filesystem that was either removed via "cmdeleteconf" or re-used, since this is a lab/sandbox cluster).

LIVE TESTING:

It seems that taking a backup/tar of everything in the shared filesystem and restoring it to its equivalent standalone locations will be necessary IF (and I stress IF) we pursue this in our live environment, but if anyone has gone through this process successfully, I would be curious to know if I am on the right track..? 

Also, is there is an "official" process for removing HP Serviceguard-Clustered NBU and going back to Standalone? I reviewed the 7.5 Admin Guide but only see reference to installing it, which begs the question: is the only recommended solution here to back everything up, do a complete standalone re-install, and restore from backup?

Thanks in advance.

[mph999]

I hate to use the words 'not supported' but ...

https://www.veritas.com/support/en_US/article.TECH31385 

VERITAS has developed comprehensive methodologies to handle a variety of reconfiguration requirements including the following scenarios:

• Migration from an existing master server to a NetBackup appliance master server
• Merging of two or more domains
• Splitting of a domain into two or more domains
• Clustering of a non-clustered Master Server
• Renaming of a Master Server or Media Server
• Moving a Master Server to a new machine involving a host name change
• Moving a Master Server to a different platform/operating system (Windows to Linux/Unix or Linux/Unix to Windows)
• Un-clustering a clustered Master Server

Even restoring back from a catalog backup doesn't remove the 'cluster config, as it's contained with NBDB.

It may be that you can get it to run, however, the risk of problems down the line is why it's unsupported eg. upgrades.

The way it's done is by using catman, which is carrried out by consulting.  This is a tool that unpicks the cluster config returning to a standalone setup.

[JL_DXC]

Thanks, kind of suspected that.

[mph999]

Yes, sorry about the bad news ... a bit out of my control I'm afraid ...

[Marianne]

Take it from someone who has tried (in a lab environment fortunately) - clustered nbdb will not start in a non-clustered installation. Not even if you perform a fresh installation of NBU with the Virtual hostname now as system name and then recover the catalog.

Only catman tools can migrate the catalog from clustered to non-clustered (or vice-versa).

Catman tools will also be able to migrate the 7.5 catalogs to a newer (supported) version of NBU. 

(All NBU versions up to 7.6.x have reached EOSL on 1 Feb)