SLP duplication to tape using incorrect SRCSERVER

[nholtz]

NBU 7.5.0.6 (might be 7.5.0.2?) running on RHEL Linux 6.4 (master) and 5.11 (media server).

Master Server @ DC -> Data Domain DC
                                  | private circuit DC - HDQ, shared traffic, no issues
Media Server @ HDQ -> Data Domain HDQ
            | FC SCSI to tape drives
            |-> SpectraLogic T120 LTO5x3

SLP handles initial backup of data to the DC DD, DD OST/Boost duplication job to DD HDQ, followed by copy to tape.  However, what seems to be occurring more and more is that the master server @ DC is the "srcserver" for writing to tape device.  This causes the device to become unavailable, but it never writes to tape.  Tape device will become unavailable for as long as I let this job run, and the job will never write, or write _very_ slowly and not finish.  I have a cron job that seeks out duplication jobs (specifically, writing to tape) with a srcserver of Master Server @ DC, and cancels the job.  This has been mostly effective, but I'm seeing less and less jobs from the Media Server @ HDQ writing to tape.  We've got a backlog growing and growing, and I think the jobs I'm cancelling are just requeueing, instead of using the correct media server for srcserver. (I think this is right, but not really sure)

Here's what I'm using to 'see' the jobs that are writing to tape:

$BPDBJOBS -report -most_columns | awk -F',' '{print $1,$2,$3,$36,$43}' | grep ' 4 1 3 '

Output looks like:

[root@nbmaster ~]# nbjobs
Currently writing to tape:
3385128 4 1 3 bckback2
3385127 4 1 3 bckback2
[root@nbmaster ~]#

If I see 'nbmaster' instead of 'bckback2', I know there is a problem with the job, and use

/usr/openv/netbackup/bin/admincmd/bpdbjobs -cancel $job

to cancel the job.

How do I ensure the HDQ Media server is the only SRCSERVER for the duplication to tape jobs?  This seems to be the core issue, as tape devices get allocated to the Master Server instead of the Media Server.

[Marianne]

You can create this touch file on the master server:  USE_BACKUP_MEDIA_SERVER_FOR_RESTORE 

See http://www.veritas.com/docs/000005346 
It applies to duplications and restores.

PS:
Are you aware that all NBU versions up to 7.6.x reached EOSL on 1st of Feb 2017?

[nholtz]

Unfortunately (for me) that file has existed since 2012.

[root@nbmaster config]# pwd
/usr/openv/netbackup/db/config
[root@nbmaster config]# ls -l USE_BACKUP_MEDIA_SERVER_FOR_RESTORE
-rw-r--r-- 1 root root 0 Aug 23  2012 USE_BACKUP_MEDIA_SERVER_FOR_RESTORE
[root@nbmaster config]#

We are in contact with our Veritas Sales team wrt upgrading to a supported version.  Thanks for pointing that out, though.