looks like the problem i had with recovery log full.
support said to upgrade 4.1.4 ? may be a good idea. 4.1.4 includes fix for
expiration errors. if u have a clean look at actlog and its errors , you
may get an idea why is it happening. these are the following events may
cause server to crash frequently (as i have experienced and had no choice
other than rebuilt TSM server from scratch upon adivse of L1,US)

1. expiration is not happening , errors in expiration (OBJECT not found in
TSM database table) - this is due to index corruption. to overcome this
problem , apply 4.1.4 and do 'audit db'.

2. recovery log full due to client backup pinning the log tail. try
overcome this by splitting client sessions (if it is db2 like databases,not
much idea on other dbs)
try avoiding server processes like expiration,migration,reclamatoin,db
backup etc while client sessions are running (run these operations
mannually).
If you are backingup data to disk and then to tape pool for large sized
objects, try backing up directly to tape pool. OR write a script which may
page you when recovery logs goes beyond 70-80 percent.

3. In process of extending recovery log and starting server process ,
server seems to be hanging for a long period while recovery log undo
process. keep patience and let it come up. dont kill this process. this may
cause tsm db corruption and you may force to do restoration of latest
available db.

hope this helps you.
thanks and regards
sreekumar.



                    "Malbrough,
                    Demetrius"           To:     ADSM-L AT VM.MARIST DOT EDU
                    <DMalbrough@TT       cc:
                    IINC.COM>            Subject:     Re: TSM Crashing
                    Sent by:
                    "ADSM: Dist
                    Stor Manager"
                    <ADSM-L AT VM DOT MAR
                    IST.EDU>


                    12/21/2001
                    23:44
                    Please respond
                    to "ADSM: Dist
                    Stor Manager"





Good idea!!! Match up the time of the crash in the errpt -a |more report
with the dsmserv.err & actlog! Any ANR9999D errors may be critical!!!