THANK YOU to everyone who responded to my question regarding my Alpha not
shutting down when given a shutdown command. (I was comforted by the
responses -- I thought something was seriously wrong with my system.)
I haven't taken any further action on this, but I'll let everyone know
if/when I find out anything new.
Here's a summary of the responses:
From: Peter Stern <CFSTERN_at_WEIZMANN.weizmann.ac.il>
It happened to me. I have seven workstations and there is no way I
could predict to which and when it would happen, but it has to do
with noninterruptable processes that get left. The good news it that
it goes away with V3.0 (or more), or you can get a patch.
From: Brian Weaver <weaver_at_pr1.k12.co.us>
Mary, I have the exact same problem with version 2.0, and tech
support says it's a known problem. The good news is, it's fixed
in version 3.2, the bad news is there isn't a patch for 2.0.
From: "Richard L Jackson Jr" <rjackson_at_portal.gmu.edu>
This is a known bug. A OSF/1 3.2 patch exist for this problem -- OSF320-057.
You will have to check with the CSC for a 2.0 patch. The patch fixes
/sys/BINARY/ldtty.o.
Patch ID: OSF320-057
Some systems were experiencing hangs during shutdown due to a tty
being incorrectly held open. This would cause the "wall" command
to hang thereby causing the shutdown to not complete.
A kernel rebuild is required.
From: Phil Rand <prand_at_paul.spu.edu>
Mary, this has happened to me several times. I hadn't noticed the connection
with idle users.
From: Andrew Gallatin <gallatin_at_isds.Duke.EDU>
Shutdown has been semi-broken for some time (since 2.0 or so, I'm at
3.2). I don't think I've ever heard a satisfactory explaination for
it.
I've found that 'init 0' will halt a machine as safely as shutdown &
much more reliably.
From: Carole Thompson <carole_at_callutheran.edu>
This has happened to me a couple of times: DEC Support never has a
reason, explanation, or sympathy for that matter. However, if this
happens, try doing an 'init s', which effectively shuts down. I also
have dead procs when shutdown is disabled, usually talkd. My strategy is
to reboot regularly, so that dead procs are cleaned up, and the proc
table never get full of junk. I'm interested in less thuggish solutions,
if you hear of any.
From: Andreas Priebe <apr_at_aip.de>
I have no idea what this could be - just once I noticed on our SUN Sparc10
under SUNOS 4.1.3. that a shutdown WITHOUT message (ie. "shutdown -h +5"
instead of "shutdown -h +5 'Message ...'") gives the start of the shutdown -
but the machine is never going down.
But again - no idea if this is related to your problem.
And to be honest - I never made deeper investigations :-(
From: howard_at_cycor.ca (Philip M. Howard, Sys. Admin. Cycor)
I use '/usr/sbin/shutdown -k <time> <message>' myself... then issue the
'halt' or 'reboot' command separately.
Apparently the hanging processes common to OSF/1 cause shutdown to fail. The
'-k' issues the messages but does nothing to bring the system down... it's a
'fake' shutdown, if you like. That gets my users off and I issue the
'reboot' manually.
Sometimes, when the system has built up a lot of zombies this shutdown
command fails too... all I can do then is manually create a /etc/nologin
file and do a manual 'reboot'.
From: Jarkko Hietaniemi <jhi_at_snakemail.hut.fi>
We've seen this several times. One possible explanation is that when
the shutdown tried to announce the soon coming end some tty/pty was so
badly hung that the announcement routine got stuck. There is a patch
for this that claims to fix the problem:
---
Patch ID: OSF320-057
Some systems were experiencing hangs during shutdown due to a tty
being incorrectly held open. This would cause the "wall" command
to hang thereby causing the shutdown to not complete.
From: ap_at_budapest.ozonline.com.au (Andrew Prendergast)
a ...
kill -TERM 1
should work ;)
Received on Fri May 26 1995 - 10:41:29 NZST