Sorry I posted this already but with a Subject: Re ...
instead of summary.
Many thanks to all who responded
Daniel Clar <Daniel.Clar_at_supelec.fr>
Terry Horsnell (tsh_at_mrc-lmb.cam.ac.uk)
Richard L Jackson Jr <rjackson_at_portal.gmu.edu>
Dr. Tom Blinn <tpb_at_doctor.zk3.dec.com>
Claudia A T C Burg <claudia.burg_at_asu.edu>
Ian Mortimer <ian_at_physics.uq.edu.au>
It shows that the problem happens under quite often under different
circumstances. Therefore I include most of the responses at the
end.
Regards
Otto
My original message was:
>
>Hi all,
>
>it isn't a problem, but just to improve my poor knowledge of DUnix.
>I performed a
># shutdown -h now (also the same mit -r -f)
>The message "system is going down immediately" came but not the
>message "shutdown time arrived" and nothing happened. Repeating
>this gave the same result. A
># ps
>showed all shutdown attempts with a state "I<". The system was
>rather idle with only a few users.
>
>After rebooting the system the shutdown worked again as expected.
>What can be the reason for this behavior?
>
>Thanks and regards
-----
Responses:
Terry Horsnell (tsh_at_mrc-lmb.cam.ac.uk)
I often have the same problem. (DU 4.0D, patch kit 2).
It used to be necessary to manually stop LAT before
doing a shutdown, as the process would hang waiting for
LAT to be stoppable, but this seems not to be the
problem in my case. It does seem to be associated
with NFS connections though; my greatest trouble
is shutting down two nodes with cross-mounted disks,
even 'sync; sync; sync; halt' doesnt always work.
I'll be *very* interested in your summary...
-----
"Richard L Jackson Jr" <rjackson_at_portal.gmu.edu>
We have this problem almost all of the time. It is related to wall not being
able to send a message to all users. So, one workaround is to not send
messages to the users, not good, by using the -f switch. This is a problem
that we have had with Digital UNIX for several years from 3.2x thru 4.0D.
Sometimes we send a wall message about the pending shutdown then try a graceful
shutdown without message. Some users will get the message.
-----
Dr. Tom Blinn <tpb_at_doctor.zk3.dec.com>
What really happens when you issue the "shutdown -h now" command is that a
signal is sent to init to tell it to terminate all the processes running on
the system. It tries to kill all processes, and waits for the processes to
go away. If some processes don't go away in a reasonable amount of time, it
will still wait for them for a while. While it's waiting, the system does
not go down. This sounds like what happened. But it's really hard to know
for sure, since the system has been rebooted, so no one but you really has
any knowledge of what was going on while it was "sitting there".
-----
Claudia A T C Burg <claudia.burg_at_asu.edu>
we had a similar problem here in the past. it mostly happened on one
computer where the user was always hanging windows - ie he had a lot of
dead, but unkilled windows hanging about. i don't recall the details and
technical terms, but it seems if shutdown can't communicate to all windows
properly it doesn't want to shut down. you may try ensuring that all
users have logged out and then look for hung processes to kill. when i
found his dead windows and killed them the system went down.
-----
Ian Mortimer <ian_at_physics.uq.edu.au>
I've had this happen a few times when trying to shut down a
DU system and a HP-UX system.
In every case the problem has been a user process which refuses
to shutdown. The solution is to do 'ps -ef | grep -v root' to
see what user processes are still running and kill -KILL them.
With DU you don't have to abort the shutdown. As soon as the
problem processes are killed the shutdown proceeds.
Received on Wed Oct 28 1998 - 13:50:19 NZDT