Veritas-bu

[Veritas-bu] RE: Veritas-bu digest, Vol 1 #4751 - Netware cluster backup

2006-01-18 15:42:48
Subject: [Veritas-bu] RE: Veritas-bu digest, Vol 1 #4751 - Netware cluster backup
From: Mark.Hickey AT hds DOT com (Mark Hickey)
Date: Wed, 18 Jan 2006 12:42:48 -0800
 
I had to back up a large (1.5 TB, 25-40 volumes) Netware cluster in the
past. The situation is as stated in the response - since the backup
product is not cluster aware, you have to have the volume mounted on the
server and set up the backup client as described in the Netware client
manual for netbackup (which involves failing all the volumes to each
server while setting up the client).  After you do this, you will have
one other problem.  If you have a lot of volumes, then it is hard to
figure out what got backed up for the following reason:
  - you have to attempt a backup of all volumes on 
    each cluster member each night in order to ensure that the
    volumes get backed up even if they end up failing over to
    another node between backup sessions.
  - as a result, you will get backup results for each volume multiple
    times.  You will get successful, but really small, backups for the
    volume when backed up from the cluster members where it is not
mounted, 
    and the usual size backup on the cluster member where the volume
*is*
    mounted.
  - you also have some volumes (local system volumes) that should be 
    backed up on *all* the servers.

So this is all very confusing to parse visually, especially if you are
using in-line duplication.  It is also very confusing to most of the
reporting tools out there (Bocada, etc.) because they get valid status
from the ndbd database.

What to do?  I wrote a PERL script that reads a config file listing all
the volumes in the cluster, whether or not they are clustered or not,
and what systems the volume can fail to.

The script takes the output of dumping the jobs database with bpdbjobs
and the volumes information and outputs a report that shows what volumes
were backed up on what server, what the status of all attempts on that
server were (along with the job ids) and what files could not be backed
up.  It also prints a summary of all the volumes that were supposed to
be backed up, the volumes that failed, and the volumes that were
supposed to be backed up that were not found.

I will have to look into the legality of posting this, however, since it
was written for a client.



<Prev in Thread] Current Thread [Next in Thread>
  • [Veritas-bu] RE: Veritas-bu digest, Vol 1 #4751 - Netware cluster backup, Mark Hickey <=