Veritas-bu

[Veritas-bu] Yet another bpstart/bpend script question

2007-01-24 07:38:55
Subject: [Veritas-bu] Yet another bpstart/bpend script question
From: ewilts at ewilts.org (Ed Wilts)
Date: Wed, 24 Jan 2007 06:38:55 -0600
On 1/23/2007 11:12 PM, Jayanth wrote:
> Thanks Jasan...if you could try giving a path in the 1st backup stream to be
> very small like in KBs and that stream need not be part of the DB, could be
> part of File System. That way you could ensure that the 1st stream is
> completed much before the other streams. Not sure if this would work but I
> am just thinking loud.

No, this won't work.  For any combination of multiple streams and 
bpstart/end scripts, you will have a not-insignificant probability of 
logic failure.  I've seen this in practise at our shop and I've seen 
many, many postings here from people who are "close" but still see 
script failures in weird and wonderful ways.

You may think that your smaller stream will complete first, but if 
there's a failure in your job and the stream restarts, you'll run into 
another failure scenario. All it takes is for a bad spot on tape, stream 
1 gets retried, and suddenly stream 2 not only starts first but can end 
first.

> For the Exhange cluster server, once the backup job fails because of a
> cluster failover or any other error, a mail alert is sent but most of the
> time this happens during the night and the backup admins are all at home,
> they would come in to office in the morning and restart the job and the
> backup would happen during production hours.
> 
> So, what we are trying to figure out is: is it possible to detect the
> cluster failover and restart the backup job so the backup is done before
> production starts...I know this is a weird wish list but just checking if
> anybody has tried it before or has some clues on how we could achieve this.

A cluster failover is simply a communication error between the media 
server and the Exchange host - you'll get a status 41 or something 
similar to that.  Enable retries on failures and your job will retry. 
If you can receive the mail alert automatically, you can also restart 
the backup job automatically.

If this happens to you a lot, you should have a look at your Exchange 
cluster :-)

        .../Ed

-- 
Ed Wilts, Mounds View, MN, USA
mailto:ewilts at ewilts.org