We are having a problem on an AIX 4.3 system, whereby users somehow exit the system in a way such that their process continues to run.
In the who listing, the user may or may not be listed. Processes are still listed in ps, and are still assigned to the pseudo-tty. Processes continue to receive CPU cycles.
One way to detect these orphaned processes is to echo "nothing" to the device, and detect the return code;
echo "\c" > /dev/pts/100 2>/dev/null
if test $? = "1"
then
# bogus tty
fi
That can certainly be automated and scripts used to manage the problem, however I am looking for a root cause.
Is it possible that one of the signals has been mapped, and is causing this behavior? Is there some telnetd configuration that can be tweaked, in order to cause these processes to get cleaned up whenever a telnet socket dies?
Any hints or comments will be very appreciated.
Thanks!