ADSM-L

Re: large client multi-streamed backups

2001-09-11 11:39:19
Subject: Re: large client multi-streamed backups
From: "Glass, Peter" <Peter.K.Glass AT WELLSFARGO DOT COM>
Date: Tue, 11 Sep 2001 09:30:57 -0500
Len,
In many cases, all files are grouped under the same filespace, but are to be
split off into multiple backup sessions beneath that. For example:
 Session 1
                /usr/oracle/ad/dev/sys/sys01
                /usr/oracle/dev/data/rtdat0t01.db
                /usr/oracle/dev/index/rtdat0i01.db
                /usr/oracle/dev/data/rtdat0t04.db
                /usr/oracle/dev/index/rtdat0i04.db
                /usr/oracle/dev/rollbk/rbs01.db
                /usr/oracle/dev/data/rtusr0t01.db
Session 2
                /usr/oracle/dev/data/rttmp0t01.db
                /usr/oracle/ad/dev/log/redo01.db
                /usr/oracle/dev/log/redo01a.db
                /usr/oracle/ad/dev/log/redo02.db
                /usr/oracle/dev/log/redo02a.db
                /usr/oracle/ad/dev/log/redo03.db
                /usr/oracle/dev/log/redo03a.db
                /usr/oracle/ad/dev/controlfile/ctrl01.db
                /usr/oracle/dev/sys/ctrl02.db

Both sessions are to each back up files from under the /usr/oracle/ad and
the /usr/oracle/dev directory paths. These directory paths also contain
files of other databases that will be active during the backup, so wildcards
cannot be used, in order to avoid active files.

For now, the only way I can see to do this is to set up a separate schedule
for each group, and specify each file individually in the OBJECTS field. But
isn't there a cleaner way to do this?

Regards,

Peter Glass
Distributed Storage Management (DSM)
Wells Fargo Services Company
> * 612-667-0086   * 866-407-5362
> * peter.k.glass AT wellsfargo DOT com
>
> -----Original Message-----
> From: Len Boyle [SMTP:SNOLEN AT VM.SAS DOT COM]
> Sent: Monday, September 10, 2001 8:59 PM
> Subject:      Re: large client multi-streamed backups
>
> In article
> <A7BF798B0884D411A1AB00508BDF0B3E05202F36 AT msgmsp17.wellsfargo DOT com>,
> "Glass, Peter" <Peter.K.Glass AT WELLSFARGO DOT COM> says:
> >
> >We are converting our Netbackup clients to TSM. Some of these clients are
> >large DB2 and Oracle clients with tight processing windows. Most of these
> >clients have chosen not to use TDP or SQL Backtrack. The solution in
> >Netbackup was to spread their workloads across several backup classes
> that
> >ran concurrently.
> >Can TSM clients be setup to multi-stream their backups in similar
> fashion?
> >Unless I'm missing something, I'm not seeing where TSM has the kind of
> >granularity that allows a client node to be setup to run its backups in
> >several concurrent sessions, where each session backs up its own
> >pre-determined set of files.
> >Any advice on this, or where I might find a solution?
> >Thanks, in advance.
> >
> >Peter Glass
> >Distributed Storage Management (DSM)
> >Wells Fargo Services Company
> >> * 612-667-0086   * 866-407-5362
> >> * peter.k.glass AT wellsfargo DOT com
> >>
> Hello Peter
>
> The short answer is yes,TSM can do multi-stream backups at the filespace
> level.
> By default it will use two resource streams. One for backup data and one
> for
> status. You can define another resource stream for each filespace.
> Now you may have other problems which cause this to fail. You will
> have to give the list a little more info on the nature of the material
> to be backuped up to get a better answer.
>
> I assume that since this data is data-base files, then you are basically
> doing
> full backups of these files. In this case instead of doing an incremental
> backup of the filesystems, you could start a session for each data base
> part
> and do a selected backup of same. You can have n tsm client sessions
> active
> at one time.
>
> len
<Prev in Thread] Current Thread [Next in Thread>