Add to DSM.OPT on the client machine:
traceflags INSTR_CLIENT_DETAIL
tracefile somefile.txt
There are many other traceflags - see the Trace Facility Guide at
http://www.tivoli.com/support/public/Prodman/public_manuals/AB/STOR/TSM/SH26
-4121-01.html/ansttfrm.htm
Some of the more useful flags for your situation ,ight be (just guessing of
Some of the more useful flags for your situation ,ight be (just guessing of
course):
Fileops: shows how it scans the disk structures
Instr_client: shows time apportioned between client, server, and network
Instr_client_detail: shows time apportioned between various sections of
client logic.
Config: shows all possible options, and whether their values are from server
cloptset, client options file, or the default.
Enter: gives a somewhat cryptic overview of client logic
Comm: shows how client connects to server
A frequent problem is huge or deeply nested directories. The TSM client has
to walk the directory tree, and the server has to send back all that info,
just to find out that perhaps nothing has changed. The new TSM 4.2
Journaled backups should help with this - until then, can you unmount the
filesystem and do an image backup? or exclude that piece of the directory
structure? or gzip it into one file and backup that file? or do an
incremental-by-date backup, which eliminates most of the client-server
dialogue?
Hope this helps.
> -----Original Message-----
> From: Bonnie OShea [mailto:BOShea AT huntoil DOT com]
> Sent: Wednesday, July 04, 2001 11:34 AM
> To: 'lmorris AT servergraph DOT com'
> Subject: RE: Solaris 2.6 bad performance with TSM
>
>
> It seems to point to high idle wait times, and I am not sure how
> to turn on
> client tracing. Can you tell me? What should I be looking for if the idle
> wait times are very high?
>
> -----Original Message-----
> From: Lindsay Morris [mailto:lmorris AT SERVERGRAPH DOT COM]
> Sent: Tuesday, July 03, 2001 6:18 PM
> To: ADSM-L AT VM.MARIST DOT EDU
> Subject: Re: Solaris 2.6 bad performance with TSM
>
>
> Does your accounting log (dsmaccnt.log, in the server/bin directory)
> indicate a lot of media wait, comm. wait, or idle wait?
> If it's idle wait, turn on client tracing,;
> If it's media wait, turn on server tracing;
> If it's comm wait, see if you can get ftp to do the same thing, then bug
> your network gurus.
>
> If you're backing up directly to tape (q se f=d will tell you),
> try going to
> a disk pool instead.
>
> Interesting problem. Please let us know how it comes out.
>
> > -----Original Message-----
> > From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU]On Behalf
> > Of
> > Bonnie OShea
> > Sent: Tuesday, July 03, 2001 5:17 PM
> > To: ADSM-L AT VM.MARIST DOT EDU
> > Subject: Solaris 2.6 bad performance with TSM
> >
> >
> > We are running TSM 4.1.2 server on an enterprise E4500 server
> with Solaris
> > 2.6. Our performance is terrible, and we have tried all of the
> TCP tuning
> > and still nothing. What happens is this: When a backup is first
> > kicked off,
> > the speed is great, anywhere from 86 to 12 MB per second. Within
> > 3 minutes ,
> > it drops to about 4-5 MB/sec, then keeps dropping over time until
> > it settles
> > in about 500K/sec. Then if you stop the backup, and start it again, it
> > starts up at 500Kb/sec. If you stop the backup and wait about 3
> hours and
> > start it, it starts up really fast again, then slows down. We
> are puzzled
> > and frustrated and so far neither Sun nor Tivoli has been able
> to help us.
> > Anyone?
> >
> >
> >
> > Bonnie O'Shea
> > Manager, Information technology
> > Hunt Oil Company of Canada Ltd
> > (403) 531-1536
> > boshea AT huntoil DOT com
> >
>
|
