Hello managers,
i asked, why so many processes survive a power-off of a Xterminal-
CDE-session, how to identify and get rid of them.
The answers (from my Digital Support Center):
1.) The reason is, that such a CDE-Xterminal-Session runs via
a tcp/ip connection and that tcp/ip was not designed to "handle"
such brute power-offs. That is: an application on the server
may (or may not!) use the timeout to decide, whether the connection
is still running or not.
2.) For this timeout there are 2 kernel-parameters under digital unix:
tcp_keepidle (default 14400 half-seconds) and
tcp_keepintvl (default 1500)
thus defining a default timeout of 2 hours
3.) There is no default-procedure/utility to identify and or kill
processes that are running after a power-off broken Xterminal session.
We have to write a script ourselves to identify candidates and
decide manually whether to kill or not thise candidates...
4.) Lucio Chiappetti <lucio_at_ifctr.mi.cnr.it>
has sent me a script (author: Phil Farrell, Stanford University School of
Earth Sciences) to kill idle processes.
(I did not take time to look deeper inside it yet, because i don't
think that all idle processes are good candidates to be killed and
i have not found a good algorithm yet to decide which processes are and
which not...)
With regards,
Bernt Christandl
----------------------------------------------------------------------
- Bernt Christandl / Max Planck Institut - Extraterrestrische Physik -
- D-85740 Garching / Phone: +49/89/3299-3342 / Fax: +49/89/3299-3569 -
- email: beb_at_mpe.mpg.de -
----------------------------------------------------------------------
Received on Mon Feb 02 1998 - 15:03:34 NZDT