We have user who is consistently able to hang our DEC 3000/400's running
OSF/1 v3.0 with a program he runs. These systems are configured for lazy
swap, and have 64MB of memory and 467MB of swap.
He says his program's memory needs increase during the course of a run.
After a certain point the console starts displaying "swap space below 10%
free" and the system hangs. Today, interestingly, the system stayed up,
network access stayed up, but you couldn't telnet in, and the messages file
showed the following:
Jun 30 10:28:06 kestrel5 vmunix: swap space below 10 percent free
Jun 30 10:28:39 kestrel5 vmunix: swap space below 10 percent free
Jun 30 10:29:59 kestrel5 last message repeated 3 times
Jun 30 10:31:34 kestrel5 last message repeated 5 times
Jun 30 10:31:35 kestrel5 vmunix: process (pid = 332) killed because of no swap
space
Jun 30 10:31:35 kestrel5 vmunix: process (pid = 400) killed because of no swap
space
Jun 30 10:31:35 kestrel5 vmunix: process (pid = 281) killed because of no swap
space
Jun 30 10:31:45 kestrel5 vmunix: swap space below 10 percent free
Jun 30 10:32:23 kestrel5 last message repeated 3 times
Jun 30 10:32:28 kestrel5 vmunix: process (pid = 8639) killed because of no swap
space
I'm guessing it killed the inetd at some point.
My questions are:
* why doesn't the system kill the offending job rather than letting
the system die? Is this just basic unix behavior?
* Is there any way, short of adding more mem and swap, to resolve
this?
Thanks!
Judith Reed
judith_at_npac.syr.edu
systems_at_npac.syr.edu
Received on Fri Jun 30 1995 - 18:21:50 NZST