Doh,
I should read through all my email in the morning before I start
writing replies....I see other folks brought up both my suggestions. Ah
well, I'll send them anyways to affirm their suggestions.
One of the things I like to do, especially when the data in
question is in one of our non-collocated storage pools, is to prefetch
the data to the dispools in advance.
At the earliest indication that we may be having to restore a
large chunk of data from a client, we will start a "move nodedata" on
the data in question. If you read about the "move nodedata" command it
specifically states that this is just what it can be used for.
Depending on how bad the host looks, we may do it only on one
filesystem or perhaps all the data for the node. I can start it off
using from 1 to all the tape drives to have it multi-thread moving the
data from the tapepool to the diskpool.
The one downside is that it moves all the data (the inactive and
active data) to the diskpool, so you are actually moving more data than
is necessary. I think I saw someone in this list request an "active=yes"
flag or something for the command, but I'm not sure that request was
actually done through official channels.
Another thing you can do is a selective backup of all the data
on the host in advance. It can kinda throw your versioning off, as you
will get 2 identical files as versions, but it will put the data right
there on disk or on one tape in a contiguous fashion.
Ben
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED] On Behalf Of
Jim Zajkowski
Sent: Wednesday, January 11, 2006 4:11 PM
To: ADSM-L@VM.MARIST.EDU
Subject: Pre-fetching a restore?
Hi folks,
I'm working on our internal "late night admin guide," and one of the
things I'm thinking of is how can I get TSM prepared to do a restore.
Here's what I mean: let's say I know that there is going to be some
filesystem maintenance on a client. Since we've been burned by that
kind of operation in the past, I'd like TSM to prefetch the appropriate
data from tape and have it ready to go (on disk) for a restore.
Archives would do the trick except that uses the client, potentially
during business hours.
Backupsets look like they might work but they're kind of rigid... can a
backupset be restored followed by restoring the latest incrementals? So
I could create a backupset on Friday before the procedure on Saturday,
and then be able to restore the incremental we took before beginning the
disk operation after that?
Am I out of my tree? Do people do this?
Thanks,
--Jim
