Migrating NBU 6.5.3 Master From HP-UX to Linux

[AIXManRon]

Ok, I know that the recommended solution for the question I'm about to ask is to engage Symantec's Consulting Services, but I'm going to ask the question anyway, simply to see if there is anyone out there who has been successful in transitioning a NBU 6.5.3 master server from HP-UX (or Solaris) to Linux with or without Symantec's help. 

We have a migration plan that we developed which has been tested multiple times including with the production catalog, and we are confident that the process will work, but we're doing our due diligence and soliciting others' input before we pull the trigger next week.  Anyone have any experience with this?  Any gotchas we need to keep in mind?  Anyone find anything bad happening after a migration that wasn't expected?  I'm happy to post the plan here if someone wants to take a look to see if we've missed something obvious.

Thanks very much!

Ron

[Abesama]

I did master server migration once with Symantec Consulting, across different OS, different hostname.

I guess your migration keeps the same hostname, because you could test it yourself and you are confident it will work - I don't think you can really do that with the hostname changed on new master.

in my migration, it was like this.

From = Solaris9, old_hostname, 6.5.3
To = RHEL, new_hostname, 6.5.3

Because this one was with new hostname, after the migration I found few of the SQL backup scripts (on the backup clients, of course) had to be modified to reflect new master server name in there.

Symantec consulting's migration binary does not update the vault profile properly, so I had to update it manually.

The job database started with jobID 1 - in other words, it did not migrate the job database, so we had to keep the old server around for a little while.

If you post your plan here, then that'll probably cover it - but let me ask anyway, did you test it with the EMM too?

I'm surprised that EMM database can be migrated across different OS.

Abe

[Mouse]

Both with Symantec Consulting and without...
What can I say - if you want to keep old hostname and have tried to dump/recover catalog, that should work (of course, bear in mind that nobody will help you if something will go wrong).

During one migration (Windows 32 bit -> Windows 64 bit) we lost our job database (that's what Abesama has said in his post). Be aware of it

[AIXManRon]

Interestingly enough, the EMM database transfers flawlessly with the bpsyncinfo -doBackup and the bprecover command.  All the GUID's, dates and other binary implementation dependent data looks fine and matches up.  I'll clean out the hostnames out of the plan and post it here later today.

[MattS]

I did this earlier this year and ran into problems.  I went from Solaris 9 to RHEL4, keeping the same server name and IP addresses.  The problem was simple but I had a hard time discovering it, and support actually did help me on it (i think they will help if you keep the same server name)...

After i installed NB on the new master and disabled the old one and it's NICs. I restored the catalog from tape as if this were a disaster recovery scenario.  I can't remember the exact errors but after looking through the logs from the import I remembered that, long ago, we had to add some space to the Solaris master and linked the NB database and some other directories to a new disk.  Well, when you restore your catalog, it expects to see the exact same file structure that existed on the old master (theres a better term for this i think).  In other words it acknowledges symbolic links, and will restore the data to the same exact directory structure on your new server (minus the symbolic links.. i think). 

To resolve this, we simply checked all the directories netbackup uses for symbolic links, on the old server.  Then on the new server we found the location of the restored data, shutdown netbackup and copied it to the default installation locations.  I might be wrong, but i think netbackup started back up without issue.

The new master was a VM so we were able to take snapshots whenever we hit a spot that we were unsure of, saved us a ton of time.

Other than that, i had no issues.

Matt

[AIXManRon]

Hi, All,

Just wanted you to know the migration was successul.  Our biggest issues were NFS performance problems (we ran the bpsyncinfo -doBackup to an NFS mount exported from the new backup server), which we fixed with a reboot of the HP-UX server, and then a bad GBIC on the new tape SAN.  Other than that, the catalog and EMM db migrated flawlessly and we were able to remove a media server and rename the master all in one 12 hour window.  All I can say is having a bullet-proof migration plan and a post-migration test plan made all the difference in the world.  I still need to sterilize the migration plan, but will get it up here by end of week.  Sorry for delay if anyone is keenly looking for it.

Ron