ADSM-L

Re: Monthly Archives

1997-04-29 09:23:00
Subject: Re: Monthly Archives
From: Bill Colwell <bcolwell AT DRAPER DOT COM>
Date: Tue, 29 Apr 1997 09:23:00 -0400
I submitted a requirement at the SHARE conference in 3/96 which would meet
your needs.  I called it 'discrete point-in-time backup and restore'.  I
asked that IBM build in a way to snapshot the active set and keep track of
the snapshot.  It should be possible to bring together the meta data for
the files of the activeset and treat this list as a new kind of object.  At
the same time each activeset database record must be flagged to prevent
deletion while it is still part of a snapshot.  During restore, the client
could select from a list of snapshots and select one to use as a substitute
active set.

The actual file versions aren't moved so it should be possible to snapshot
all the nodes in a very short time, no more than the time it takes to do
expire inventory (which I run weekly in 4 hours or less).

I can't think of any current command which will meet your needs; as you
say, '1.5 TB of data -- inadequate time.'

If you would like to see this feature in the future please ask your IBM rep
to concur on this requirement.

Bill Colwell
The Charles Stark Draper Laboratory
Cambridge Ma.

_________________________Reply Header_________________________
Author: ADSM-L AT vm.marist DOT edu
Subject: Monthly Archives
04-24-1997 11:46 AM

A Business Requirement:  To create an Archive of all data across all
platforms (UNIX, Novell, NT) after the close of business on the last
business day of each month.  This is in addition to the normal daily
incremental backup.

Problem:  1.5 Tb of data -- inadequate time.  Is it "really" necessary to
resend the data across the network when it already exists on tape (onsite
backup and offsite copy)?

Possible solution:  On a weekend (or other ADSM server low utilization
window), restore data for one node to a given point in time onto DASD; write
an archive tape from the restored DASD; delete restored data from DASD;
repeat cycle until all nodes completed.

Are there any other solutions or options?

"Export Server" of data is not a viable option because exported files are
not managed to ADSM and you must "Import Server" all files in order to
restore a few to a given node.

Solutions surrounding restoring the ADSM database to a prior point in time
is not a viable option because it in effect brings the server down (it's
used extensively) and is simply a lot of work, not to mention being messy.
<Prev in Thread] Current Thread [Next in Thread>