This is a summary of my query regarding the automatic killing of idle
processes. I would like to thank all the people that got back to me so
quickly, so here they are:
Ry Jones--record time of 26 minutes
Kent Arnott--very close second with 39 minutes
Olivier Boebion--still very impressice at 9:46
Randy Styka--just a hair under 12 hours
Don Newcomer--about 17 hours, but Carlisle is a nice layed back place
Richard Jackson--also about 17 houres, but GMU is a verrrry busy place
Ken Brown-- about 18 hours, but with the right answer for me
This was my first question to the list and I must say I am very impressed
with the responses. All of them were of value, and all of them were came
exceptionally faster than I had dreamed possible.
Here are the solutions not in any particular order:
Set environmental variables:
csh and tcsh
set autologout=60
where 60 represents the number of minutes if inactivity.
ksh and bash
TMOUT=3600 ; export TMOUT
where 3600 represents the number of seconds of inactivity.
Worth of note, but I am dealing with sophisticated users who would disable
this in a heart beat.
Comercial products:
Computronics sells a product called Logmon that analyzes cpu usage to
establish inactivity. This apperantly prevents eroneous logoffs during
long operations. I received one sales pitch and one testominal. It sounds
like a good product, but perhaps more than we need since we are looking at
a 5 day time-out interval.
use idletime:
idletime is a nifty little perl script writtne by Kent Arnott.
See
http://www.tamucc.edu/~karnott/idletime.html for a copy.
Kent deserves special praise for his selfless assistance in answering a few
questions I had. Had idletime not come along, I would have used his
product.
use idlekiller:
idlekiller is used by my alma mater, GMU. It was written by Phil Farrell
of Stanford. I found it to be available at most of the Linux ftp sites
including
ftp://sunsite.unc.edu/pub/Linux/system/Admin/idle
Richard Jackson of GMU had some reservations about its efficency because it
makes extensive use of awk, but then again, I am not running a system with
20,000+ students.
use idled:
Ken Brown recommended this program. It is a daemon process that is
extremely versital. It can be configured on the fly. The only problem is
with notifying users on DUx systems prior to killing the process.
Apperantly DEC has a small problem that cause the notice to be sent, but
then the process isn't killed. I seem to recall that there is a patch for
wall, but I haven't tried idled with that. I just took the recommended
approach and disabled the feature.
This program is fantastic. It compiled on the first try. It commes with a
sample configuration file that has about 10 lines of configureation and
about 200 of comments. It comes with configureations for most of the
common systems. It can be confiured to base its decisions on I/O or user
input. You can vary the privledges among different groups. And finally,
it seems to get regular updates.
I have compiled an run the daemon on 3.0, 3.2-3, and 3.2c systems. My
next target is HP-UX 10.01.
More information can be had at:
http://www.cs.hope.edu/~crider/idled/
Thanks again for all the assistance.
-cliff
ckrieger_at_psi.prc.com
Received on Thu Jun 13 1996 - 23:15:26 NZST