Re: anyone have script for multi-session/multi-filespace restore?
2004-06-24 14:06:09
Well I have looked at everyones responses. And I would like to thank
everyone who has responded. I know it takes time and effert to do so.
Fundimentally I don't think I can use collocation by Filespace.
I would be using 6x as many tapes for the task if I put the COPYPOOL
tapes like that. It would also cost me at least 50% more for the local
tapepool. But I think it was a popular suggestion because I did not
include enough detail in this e-mail. I did so on the first e-mail and
then made a bad assumption... . Anyway the Large system I need to
restore is an ORACLE DB that we do FULL BACKUP (cold) once a week and
then save the logs during the week. I do not believe that this is the
best way to do things but spending $$ seems to be a no no. Anyway I
need the RESTORE to work for a full system, (like DR), to provide
recovery for a local or DR site RESTORE or to populate a test system.
The part that makes this different is that there are no daily
incremental updates on any of these Filesystems.
My solution:
1) I do a select node_name, volume_name, Filespace_name from
volume usage where node_name='MYNODE' and filespace_name like '/U0%' and
stgpool_name='TAPEPOOL' this lists all tape volumes that have any of
the filespaces I need
2) I use SAS to make a smaller file that contains the list of
UNIQUE tape volumes. The program then creates the TSM commands to do
step 3.
3) select volume_name, last_write_date from volumes where
volume_name='thevolser'
4) here I do some sorting to merge all the data together,
through away all but the most recent tape used for every filespace and
then sort them again to have a list of tape volumes and every filespace
on them to do individual restores, knowing which file spaces to ask for
so all are listed on 1 tape. (ps: the STK tape silo tape system
allocates scratch tapes in a ascending volser order. So a Filsystem on
2 tapes 'should' start on the tape with a numerically smaller volser
number)
I still think Tivoli/IBM should be able to do this internally. This is
no more than a standard RESTORE NO QUERY that would ALLOW FOR A
WILDCARDED FILESPACE.
Thanks everyone.
Matt
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Andy Carlson
Sent: Wednesday, June 23, 2004 2:30 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: anyone have script for multi-session/multi-filespace
restore?
If you have enough scratch tapes, you could consider colocation by
filespace, then divide the filespaces roughly by size into 8 groups (or
however many tape drives you want to assign) and start 8 restores. They
would be guaranteed to be on separate tapes becuase of the collocation.
On Wed, 23 Jun 2004, MC Matt Cooper (2838) wrote:
> Hello all,
> I received no responses from my previous post 'Need your
> thoughts on best RESTORE approach'. So with just my own research I
> have found that TSM does NOT have anything to manage a
> multi-filespace, multi-session restore that is not relying on the user
> to prevent 2 sessions asking for the same tape. (please tell me I am
wrong)
> ( reference the section in the (AIX) client book on LARGE
> RESTORES, NO QUERY, it is set up for a single filespace (NO
> WILDCARDS) In my case I will have 60+ filesspaces on 6+ tapes with 8
> tape drives available and have to start from the command line
> (actually scripted) to start a session for each tape drive. The real
> problem is preventing more than 1 session from asking for the same
> tape. I believe I can find the necessary information to write the
> script from quering the volumeusage and volume tables and doing the
restore comands from there.
> I would guess I am not the first one to approach this problem and hate
> to spend my time trying to script this when someone else, (who is most
> likely much better at scripting than me) has already done it.
>
> Does anyone have a script that will arrange/control the submission of
> multiple restore sessions to avoid having sessions having to wait for
> other sessions to finish using a tape?
>
>
> Thanks in advance
> Matt
>
>
|
|
|