I think that what you are seeing is due, at least in part, to the client
multithreading model. There is some overhead involved in starting and
stopping sessions, and the rapid "start/stop" introduced by the
consecutive selective commands (why not just use a client macro instead of
the redirected input?) could be contributing to performance problems.
Note: I don't understand how the -optfile option is working into this,
since -optfile is not a valid option with the SELECTIVE command; it is
only valid whe invoking TSM.
A couple of suggestions:
1) Try RESOURCEUTILIZATION 1. This will, as you guessed, get you down to
one session through which all the files will be backed up.
2) Better yet, go to the 4.2 client, and use the -filelist option. What
you do is:
a) Create a file containing a list of the files you want to back up
(without the SEL command or other options... just a list of file names,
one per line). For example, textfile.in might contain:
file1.txt
file2.txt
file3.txt
b) Run the command
dsmc s -filelist=textfile.in
This will process the files much more efficiently than using redirected
input or a client macro, as these methods process 1 file per transaction.
The -filelist options allows multiple files to be processed in a single
transaction, like a regular incremental or selective backup for a drive or
wild-carded file specification.
Regards,
Andy
Andy Raibeck
IBM Tivoli Systems
Tivoli Storage Manager Client Development
Internal Notes e-mail: Andrew Raibeck/Tucson/IBM@IBMUS
Internet e-mail: [EMAIL PROTECTED]
The only dumb question is the one that goes unasked.
The command line is your friend.
"Good enough" is the enemy of excellence.
"Kent J. Monthei" <[EMAIL PROTECTED]>
Sent by: "ADSM: Dist Stor Manager" <[EMAIL PROTECTED]>
09/26/2001 16:31
Please respond to "ADSM: Dist Stor Manager"
To: [EMAIL PROTECTED]
cc:
Subject: Peculiar issue with ADSM Client -> TSM Client upgrade on WinNT
4.0
We use the B/A Client to backup a list of ~300 files. The list of files
is
maintained in a separate text file named 'textfile.in'. Each line of the
text file is of the form 'sel <filename> -optfile=<optfilename>', i.e. 1
'select' subcommand per line. The schedule prompter starts a script that
simply does a 'dsmc < textfile.in'. This is the first of several
scheduled
nightly backups of that node. The next schedule starts 30 minutes later.
We just upgraded to the TSM 4.1.2.12 Client (WinNT 4.0). Our servers are
Sun Solaris 2.6, running TSM 4.1.2 Server.
With the ADSM v3.1.0.7 B/A Client, this worked fine, starting a single TSM
Server session, processing all the subcommands in 'textfile.in' and
reporting all the results to the *SM Server - 1 session, 300 files,
consistently averaging less than 20 minutes start to finish. This was
very
reliable & rarely failed.
With the new TSM v4.1.2.12 B/A Client, we're observing the following
problematic behavior:
- the schedule prompter immediately starts 3 TSM Server sessions
- 2nd session starts processing subcommands from 'textfile.in'
- 3rd session immediately ends
- after completing each subcommand, 2nd session spawns a
short/separate TSM Server session that immediately ends
(300 files = 300 subcommands = 300 additional Server sessions,
each about 30 seconds)
- after 30 minutes (our idle timeout), 1st session terminates with
ANR0482W
- 2nd session continues processing subcommands
- when 2nd session finishes processing all subcommands from
textfile.in, it terminates normally
- IF/F 1st session terminated with ANR0482W, server issues error
message ANR2576W and 'q event' reports 'failed',
otherwise 1st session ends normally when 2nd session does and 'q
event' reports 'completed'
- IF/F server issues error message ANR2576W, all subsequent TSM
Server
schedules for this node report 'missed'
and the TSM Server Activity Log contains an ANR2716E
The same scenario is occurring randomly/intermittently on multiple clients
in different subnets on each of 4 different TSM servers, in 2 US and 2 UK
locations. All were recently upgraded to TSM B/A Client, and also
received
a Winnt TCP/IP hotfix. We have ruled out network- and locality-related
issues because of the mix of clients, servers, schedules/times and
locations.
The key questions are:
- why the 300 spawned TSM Server sessions, 1 per subcommand/file?
(this is the primary reason the timeout now gets triggered)
- can that be avoided by, for example, setting 'resourceutilization'
to
1? (I dug far enough to confirm it defaults to 2)
- when the ANR0482W occurs, why does the ANR2576W occur, and why do
all
other schedules fail with ANR2716E?
(this hints at a deeper problem, either in TSM B/A client or in
Winnt
tcp/ip)
- the ANR2716E consistently reports 1 of 4 specific tcp port numbers
1035, 1501, 1757 or 2013 - are these port numbers significant?
-rsvp with any thoughts or recommendations for resolving this, thanks!
Kent Monthei
GlaxoSmithKline R&D
