Warnings in default bpstart_notify script - A load of old cobblers?

[Tommojunior]

Hi,

I'm running 4.5MP3 on hpux. We are looking to use the bpstart_notify and bpend_notify scripts to perform our own scripted and sql backups of some large oracle backups.

Some of these "hot" or "cold" pre script can last upto five hours, so I've set BPSTART_NOTIFY to 32767 (max allowed) in bp.conf on the master.

I've successfully ran some "sample" bpstart_notify scripts with just a simple "sleep 1000" inside and all looks good.

NOW TO THE POINT!:

The supplied (standard) bpstart_notify script which is in goodies has a warning about long scripts holding up the start of other clients policies:

# If this script will not complete within a few seconds, you should set
# the BPSTART_TIMEOUT in the /usr/openv/netbackup/bp.conf file on the server.
# You should also be aware that the time taken by this script will delay
# the initiation of other client's backups".

To put this to the test, I started my "sleep 1000" policy and then started another policy on another client and even another policy on the same client. All appears fine, no issues to date (they DONT wait for the "sleep 1000" script or policy to complete)

Is the warning a load of cobblers, or is there some other side-effect I need to be aware of? Its likely that I'll be running many of these pre-scripts across dozens of servers.


All thoughts welcome!

Cheers,

Tommojunior

 

[comtec17]

What Veritas is warning about is that if you have 4 tape drives and 6 jobs needing to run, then the script will hold 1 of those drives until it completes and the other 2 jobs will sit there and eventually time out.

 

[Tommojunior]

Yeah, that's the impression I got too. However, its not true. In my dummy "sleep 10000" job, the job quite happily starts and sits running the bpstart_notify. While its running, the device manager shows no evidence of the media server having reserved a drive, this seems to (thankfully) happen after the bpstart_notify script finishes with an exit code of zero.

I've had a chat to Veritas support, who toe the party line and say that it "should" reserve a drive, but it doesn't. I'm going to a more comprehensive test by downing all drives but one, running the "sleep 10000" job and then trying another. I wonder if the media server does actually reserve a drive (and maybe a piece of media) but doesn't actually show it in Device Manager???

I've learn't something more scary though, which is going to cause me grief:

If you have multi-streaming turned on for a particular job, it will run a bpstart_notify for each stream, hence the pre job will run a number of times concurrently. Thankfully, the script gets passed an environment variable "STREAM_NUMBER" which could be used in my script to:

1) Make only STREAM-1 run the actual pre script
2) Other STREAMs wait for STREAM-1 to finish successfully

I realise I'm really pushing the functionality here, and am well into "un-supported mode". But until we adopt the database agent products, its our only choice.

Cheers,

Tommojunior

 

[Tommojunior]

HaHa!

I stand corrected. My full test of downing all drives but one confirmed that Netbackup does indeed reserve a drive and piece of media before it starts the backup_notify script. It will never tell you this in device manager, but its true.

Cheers,

Tommojunior