Thanks for the many replies I received for information on killing
processes which don't want to die. Here was the original post:
> I've got a process which I can't kill and its tying up a tape drive.
> ps -x gives the following
> 11910 ? DW 0:00 mt -f /dev/nrcwd0 eom
> I've tried 'kill -9 11910' but that doesn't work. How do I kill such
> Tom Milliman
The overwhelming response was to simply reboot. Apparently processes
in the 'DW' state are in the kernel waiting for an event (in this case
from the tape drive) at a priority higher than the priority at which
signals are received. Therefore the signals will be queued until the
event being waited for occurrs (making it immune to all signals!).
The next most common suggestion was to try power cycling the tape
drive. The rational here is apparently to get the tape drive to send
some signal to the device-driver which was stuck waiting for a signal.
There was a suggestion that there are known problems with the Sun scsi
tape device drivers which could cause this problem. We are using a
third-party device driver but of course it could have a similar
I did power cycle the tape drive and about 2 1/2 - 3 hours later the
process disappeared. I'm not sure what exactly caused it to go away.
A handful of other responses suggested other approaches which I didn't
try. These ranged from sending a variety of signals to killing and
restarting the rpc.lockd process. One interesting response suggested
looking at the Sun patch 100358-01 which was used to make a modem
process (?) killable. I'll probably take a look at this but have not
Thanks again to the list for the informative and timely response.
| Thomas E. Milliman INTERNET: email@example.com |
| Space Science Center Voice: (603) 862-2867 |
| University of New Hampshire Fax: (603) 862-4685 |
| Durham, NH 03824-3525 |
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:54 CDT