ADSM-L

Re: Monthly Backups, ...DIRMC

2002-04-11 07:49:10
Subject: Re: Monthly Backups, ...DIRMC
From: Jim Healy <James.Healy AT AXA-TECH DOT COM>
Date: Thu, 11 Apr 2002 07:49:25 -0400
Be careful with dirmc, I'm not sure but I think it might be causing a
problem doing restores from the GUI with point with point in time backups.
Yesterday I was trying to do a point in time restore with a date of march
20 on a specific sub-directory in NT. I Drilled down to the where the
directory should have been displayed and it it wasn't yet when i did a view
of the files with active and inactive displayed it showed files that would
have been there when the point in time was specifing. I fell back to the
trusty command line with the point in time option and it worked fine.

Beware!




"Don France (TSMnews)" <DFrance-TSM AT ATT DOT NET>@VM.MARIST.EDU> on 04/11/2002
01:48:55 AM

Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>

Sent by:  "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>


To:   ADSM-L AT VM.MARIST DOT EDU
cc:

Subject:  Re: Monthly Backups, ...again!: The Real Issues


Bill,

Just out of curiosity, in your restore situation -- was it an NT platform
(I
see it must have been alot of tiny files)?!?

If you are using NT, Netware or AIX as file server platforms, the DIRMC
option pays for itself... BIG TIME.  I recently responded to a customer,
clearly a smaller shop, but the storage pool was non-collocated for over a
year, using DLT library in a Win2K-to-Win2K context;  there was a total
RAID
failure for the E: drive, but I originally engineered it with DIRMC on
disk,
migrated to FILE on disk, then copy pool'ed to tape.  After about 30 hours,
1.6 million files and 316 GB of data were restored to the point-in-time
specified as the last-known-good;  this performance was achieved by using
two concurrent restore sessions (across 10 high-level directories), CLASSIC
restore (not the default), and the "dirsonly" first, then the data -- only
slowdown was due to tape mounts (which were consolidated within each
session), because the customer had more tapes than slots, so needed to
respond to demand mount requests for tapes "mountable not in library".

This experience is usually a wake-up call for the customer to evaluate
RESTORE requirements;  if 5-10 GB/Hr is satisfactory, then, so be it.  2GB
restore in 8 hours, did you check your accounting records for media wait --
and does that reconcile with the time?  How about the number of tape mounts
(available from the summary records or count the approp. message # in the
activity log)???  Sounds abit suspicious, to me.

I've seen several ideas shared in this thread, any one of which could be
the
right answer for a given context;  your 3-class system sounds interesting
--
as does Bill Colwell's, and Paul's, and Nick's, and Alex.  Also, with 5.1
as does Bill Colwell's, and Paul's, and Nick's, and Alex.  Also, with 5.1
the new IMAGE backup would seem a good substitute for monthly backupset.
Ultimately, I like Jim Taylor's answer the best... get the dialog on
RESTORE
needs, then figure out what of the various suggestions will work for a
given
customer/server/class-of-servers.

Of course, the key political question is truly to get a dialog on RESTORE
REQUIREMENTS;  focus on the business needs for that first, then benchmark
and/or collaborate for possible solutions -- ultimately, the customer needs
to take this seriously enough to "pay" for trial exercises a couple times a
year... else, they're just putting their heads in the sand and courting
disappointment.

Regards,

Don France
Technical Architect - Tivoli Certified Consultant

Professional Association of Contract Employees (P.A.C.E.)
San Jose, CA
(408) 257-3037
don_france AT att DOT net



<Prev in Thread] Current Thread [Next in Thread>