Another late summary...
I asked about how to avoid users leaving behind orphaned processes after a
crash of an X-terminal (cf. "VXT2000+ Blues"). I have found out that such
orphaned or strayed processes can occur when logged on to the console as
well. And it is not just Netscape processes that get left behind, but they
tend to be the most troublesome since they get stuck in a "hyperactive"
state and start to chew up CPU time.
>From Eli Burke <eburke_at_cslab.vt.edu> came the easiest suggestion, which was
to replace the netscape command with this little script:
#!/bin/ksh
LD_LIBRARY_PATH="/usr/3.2D/shlib"
export LD_LIBRARY_PATH
ulimit -t 1000
ulimit -c 0
nice -5 /usr/local/bin/netscape-alpha3.0 $*
"This sets core files to be zero length, and prevents the netscape
processes from takeing up more than 1000 CPU seconds. (you can vary
that to your liking..)"
This will kill the netscape processes when they reach the ulimit time of
1000 seconds.
other responses came from:
Hans Kowallik <hans_at_physics.orst.edu> (who suggested using xclosedown to
end xsessions. More information on xclosedown can be found at:
http://www.sct.gu.edu.au/~anthony/info/X/)
Debby Quayle <dquayle_at_hamilton.edu> (who suggested applying the OSF400-107
or equivalent patch)
Thanks for the help!
Original Question:
>Date: Thu, 3 Apr 1997 22:30:44 +0200
>From: Kevin Reardon <kreardon_at_cerere.na.astro.it>
>Subject: orphaned netscape processes
>
>Dear OSF Managers,
>
> I have noticed that users who connect to our AlphaServer 2100A 4/266
>with a VXT2000 X-terminal sometimes end up leaving "orphaned" Netscape
>processes running after they have logged out (see 'ps' output below). This
>doesn't happen very often (once every week or so from 15+ VXT users
>connecting every day) and wouldn't be so much of a problem, but these
>Netscape processes begin to eat up an entire CPU!
>
> Once you've noticed them, they can easily be killed by root, but not
>by the user who owns them (as their PPID is now 1). I have auto-renicing
>enabled, so they just sit in the background and only chew up CPU time when
>no other interavtive processes are running. But it could result in reduced
>performance for other long batch jobs we run. I do not think it is directly
>related to the VXT terminal crashing either, and I don't see anything in
>any of the system logs. Has anyone else has seen this and if there is a way
>to avoid such "rogue" processes?
>
>thanks,
>kevin reardon
>
>Output from ps -ef | grep user
>user 1065 2497 0.0 - ?? 0:00.00 <defunct>
>user 2179 1 99.0 13:47:32 ?? 06:51:01 netscape
>user 2497 1 0.0 13:19:19 ?? 0:03.49 netscape
>
>Output from ps Awvg | grep netscape
> 2179 ?? R N 06:52:13 0 0 15.3M 7.8M 99.0 1.5 netscape
> 2497 ?? IW 0:03.49 29812 0 15.3M 0K 0.0 0.0 netscape
>
>Output from top
> PID USERNAME PRI NICE SIZE RES STATE TIME CPU COMMAND
> 2179 user 60 4 16M 8208K run 411:47 98.00% netscape
>
>Output from last -n 3 user
>user ttyqd vxtna.na.astro.i Thu Apr 03 13:52 - 14:08 (00:15)
>user ttyqd vxtna.na.astro.i Thu Apr 03 13:44 - 13:52 (00:07)
>user ttyr9 vxtna.na.astro.i Thu Apr 03 12:59 - 13:44 (00:45)
>
Received on Fri May 16 1997 - 16:37:44 NZST