[Veritas-bu] Windows flash-backup experiences (again)
ewilts at ewilts.org
Wed Jul 11 09:50:32 CDT 2007
Your approach has some really ugly consequences. First, you need to know
which host actually hosted the data on the day that you're restoring from.
If your recovery was from a backup taken 6 months ago, was the data on the a
node, b node, c node, or d node? Secondly, the day the disk fails over from
one to another, you will trigger a full backup since NetBackup will not know
that it's the same disk - it simply sees a new disk on a new host. To make
matters worse, you probably had something bad happen to the first node to
make the disk fail over in the first place - you now aggravate that by
putting extra stress on the remaining node(s) doing full backups.
If you have large clusters - and we have >100 luns with >20TB of data on a
single cluster, you'll see that your approach isn't scalable.
Welcome to NetBackup! :-)
Ed Wilts, Mounds View, MN, USA
mailto:ewilts at ewilts.org
I GoodSearch for Bundles Of Love:
> -----Original Message-----
> From: Leidy, Jason D [mailto:Leidy.Jason at con-way.com]
> Sent: Wednesday, July 11, 2007 9:32 AM
> To: Ed Wilts; Martin, Jonathan; veritas-bu at mailman.eng.auburn.edu
> Subject: RE: [Veritas-bu] Windows flash-backup experiences (again)
> I backup ALL_LOCAL_DRIVES on a few of our clusters, it gets everything.
> For instance, I'll put both physical nodes that host the cluster in a
> policy that uses the ALL_LOCAL_DRIVES option. I'm new to the backup
> world, is this not the best practice? It gets the quorum and everything
> (I'm able to restore the data).
More information about the Veritas-bu