Hi group,
Sorry for the late response,
last week I asked:
------------
#DU 4.0B, DEC 3000-400
#
#I have a user under whose account there are running
#64 instances of comsat, the first of these processes
#occurring from 10-Mar-97, the rest being started
#until now about every day some instances.
#As maxuprc is 64, the user is no more able to work.
#I will kill these processes now, but:
#-Does anybody have any idea what may have caused this?
#-How do I prevent it in the future?
#-Will any information get lost by killing comsats?
#...
-----------
The solution came just some hours later from:
Peter Stern <peter_at_wiscpa.weizmann.ac.il>
I have had this same problem for a while now, ever since upgrading
to DU 4.0 and Digital Support has come up with no satisfactory solution.
The problem definitely comes form biff y being set. It occurs because
one can reproducibly close a dtterm window under certain circumstances
but if you do ls -l /dec/tty* for that terminal (e.g. ttyp0, etc.)
you get: crwx-w---- which means that the terminal looks executable, but
it's not there anymore so the comsat process hangs. If you just kill
comsat (no harm done) without chmod -x on the /dev/tty* (or setting
biff n the next mail sent will start a new comsat process on all the
user's terminals which are crwx.
I hope that this is partially clear. You can either set biff n for
everybody or periodically check for comsat processes (with cron?),
chmod -x on the particular user's "dead" terminals and kill all the
comsat processes.
===============
Amy E. Skowronek <amy_at_aloha.nascom.nasa.gov> wrote:
...
#I had one instance where comsat was using up all the processes, and it
#turned out that the user had put "biff y" in the .login or .cshrc
#(I forget which), so *every* window got a message when a new piece
#of mail came in. And she had a lot of windows. I don't think any
#information gets lost when you kill the comsats, as all it does
#is watch the spool.
#
#Turning biff off in the .login should help. The user can still run
#biff in a window, but it should not be the default for every window.
...
=====
so the cause at ours was also "biff y" in the .login of the
particular user. No problems anymore after removing it from .login.
Thanks to everyone,
Rainer Landes, eMail: Computer-Administration_at_Physik.uni-karlsruhe.de
Tel(+49)721 608 3578
http://www-comp.physik.uni-karlsruhe.de/
Computer facilities of the Faculty of Physics, Univ. of Karlsruhe, GER
Received on Fri Apr 11 1997 - 20:24:32 NZST