SUMMARY: System shutdown hangs at the end

From: Luca Pizzinato (Pizzinato@eumetsat.de)
Date: Thu Mar 20 1997 - 06:36:00 CST


Sorry for the delay, but I'm not allowed to shutdown the
system at any time. Anyway, I wrote

>
> Hi List,
>
> does anybody have any idea why my Sparc1000
(Solaris
> 2.5) cannot arrive to the 'ok' prompt when a 'shutdown
-g0
> -i0 -y' is issued? It arrives very close to the end, then it
says
> that /home is busy, /opt is busy, then 'System is down.',
but
> the prompt is not reached.
>
> Thanks in advance. I'll summarize.
>
> Luca Pizzinato
> pizzinato@eumetsat.de

I attached all the answers I received.
The solution of Stephen P Richardson certainly works,
becouse it fixed my problem. In few words, the "killproc
automoun" command, following "umountall -F autofs" in
the /etc/init.d/autofs script, kills the automount daemon,
while some autofs haven't already been umounted, and this
is why the system shutdown could't go any further.
Commenting out "killproc automount" fixes the problem,
becouse automountd is killed smoothly later on by "killall"
in the /etc/rc0 script
Before I tried also the suggestion of Ray Trzaska (replace
"umountall" with "umountall -s" in /etc/rc0), but it didn't
work.
Some guys suggested to use "init 0" instead of shutdown,
but even if it works I really wanted to understand what
was going on.

Thanks to

Stephen P Richardson spr@myxa.com
Ray Trzaska rtrzaska@uk.mdis.com
Clarkson, Michael clarkson@amgen.com
Amanul Haque ahaque@psa.pencom.com
Rick von Richter rickv@mwh.com
Frank Pardo fpardo@tisny.com
Robin Marquis rmarquis@ordsvy.gov.uk
Simon Convey simon@iway.nl

Luca Pizzinato
pizzinato@eumetsat.de
Tel. +49.6151.807323
Fax +49.6151.807304


attached mail follows:


I have seen this on a 2000 with a large number of stripes on ssa arrays.
it has also cropped up in other cases with striped disks.

the 'fix', for me, was to edit /etc/rc0 to change the umount from parallel
to a sequential umount.

change in the last few lines from:
---------------------------------
# unmount file systems. /usr, /var and /var/adm are not unmounted by umountall
# because they are mounted by rcS (for single user mode) rather than
# mountall.
# If this is changed, mountall, umountall and rcS should also change.
/sbin/umountall
/sbin/umount /var/adm >/dev/null 2>&1
/sbin/umount /var >/dev/null 2>&1
/sbin/umount /usr >/dev/null 2>&1

echo 'The system is down.'
---------------------------------

to:
---------------------------------
# unmount file systems. /usr, /var and /var/adm are not unmounted by umountall
# because they are mounted by rcS (for single user mode) rather than
# mountall.
# If this is changed, mountall, umountall and rcS should also change.
/sbin/umountall -s
/sbin/umount /var/adm >/dev/null 2>&1
/sbin/umount /var >/dev/null 2>&1
/sbin/umount /usr >/dev/null 2>&1

echo 'The system is down.'
---------------------------------

ray.


attached mail follows:



        Yes, there is a solution:

        comment out the line:

            killproc automoun

        in /etc/init.d/autofs (which is also linked to
        /etc/rc2.d/S74autofs /etc/rc0.d/K69autofs
        /etc/rc1.d/K68autofs.) The automounter will be gracefully
        shutdown by killall later in the shutdown process, meanwhile
        commenting it out of this script allows umountall to finish
        before the automounter process is terminated. If your machine
        is still coming down too fast, you can try adding a short
        sleep call right after the umountall in autofs. I have done
        this on many machines without any problem.

-- 
Regards,
Stephen

--------------------------------------------------------------------- Myxa Corporation Voice: (610) 436-0380 334 West Union Street FAX: (610) 429-9207 West Chester, PA 19382-3329 Email: spr@myxa.com ** Welcome Page: http://www.myxa.com **

attached mail follows:


Hi,

Is this problem related to the position of the S1000 key switch, I can't
test at the moment as the systems are live, though I seem to remember
that if the key is in the position so that STOP-A can't be used, using
shutdown behaves the way you described. Trying moving the position of
the key switch and trying again.

Hope that helps

Michael Clarkson
Unix Systems Specialist
AMGEN Ltd
Cambridge,UK

>----------
>From: Luca Pizzinato
>Sent: 14, March 1997 8:39 am
>To: sun-managers@ra.mcs.anl.gov
>Subject: Shutdown hangs at the end - NO SOLUTION ???
>
>Dear SysAdmins,
>
>some weeks ago I wrote
>
>############################################
>
>Hi List,
>
>does anybody have any idea why my Sparc1000 (Solaris
>2.5) cannot arrive to the 'ok' prompt when a 'shutdown -g0
>-i0 -y' is issued? It arrives very close to the end, then it says
>that /home is busy, /opt is busy, then 'System is down.', but
>the prompt is not reached.
>
>Thanks in advance. I'll summarize.
>
>Luca Pizzinato
>pizzinato@eumetsat.de
>Tel. +49.6151.807323
>Fax +49.6151.807304
>
>############################################
>
>I received only an answer by Ray Trzaska, pointing me to
>the command "umountall" in the /etc/rc0 script, and
>suggesting me to replace it with "umountall -s".
>Well, it didn't work but the problem must be somehow
>related to conflicts between that command and the fact that
>automounted filesystems are still mounted and the
>automountd daemon is already died, so they cannot be
>umounted. There is also a specific bug in the SunSolve CD
>about that (but no workarounds)
>
>Now, I finally asked support to Sun, and the Sun engineer
>who picked up the call told me that there is nothing I can
>do, that's the way Solaris2.5 works, use allways /etc/halt or
>reboot, and then closed the call!
>It seems to me very strange that is impossible to use the
>"shutdown" command when autofs is used.
>
>If anybody of you have a suggestion, it will be very
>appreciated.
>
>Luca Pizzinato
>pizzinato@eumetsat.de
>Tel. +49.6151.807323
>Fax +49.6151.807304
>


attached mail follows:


There is an lofs bug filled out with sun and they have a patch for it.
I had the same probelms with 1000's and 2000's. I had to beat up on sun
to get a ptch release.

Basically, if you look at the umountall script, it unmounts nfs, ufs
and lofs (autofs) filesystems. If you hve loopback file systems (lofs)
through the automounter, and the actual file systems reside on the local
machine, then you see this problem. The machine will EVENTUALLY come down,
but the time is directly exponentially related to the # of
lofs you have.

Sorry that I do not have the bug id, or the patch id, but there is a lofs
and automount path on sunsolve1.sun.com. That should help you out.

Hope this helps . . .

Luca Pizzinato said:
>
> Dear SysAdmins,
>
> some weeks ago I wrote
>
> ############################################
>
> Hi List,
>
> does anybody have any idea why my Sparc1000 (Solaris
> 2.5) cannot arrive to the 'ok' prompt when a 'shutdown -g0
> -i0 -y' is issued? It arrives very close to the end, then it says
> that /home is busy, /opt is busy, then 'System is down.', but
> the prompt is not reached.
>
> Thanks in advance. I'll summarize.
>
> Luca Pizzinato
> pizzinato@eumetsat.de
> Tel. +49.6151.807323
> Fax +49.6151.807304
>
> ############################################
>
> I received only an answer by Ray Trzaska, pointing me to
> the command "umountall" in the /etc/rc0 script, and
> suggesting me to replace it with "umountall -s".
> Well, it didn't work but the problem must be somehow
> related to conflicts between that command and the fact that
> automounted filesystems are still mounted and the
> automountd daemon is already died, so they cannot be
> umounted. There is also a specific bug in the SunSolve CD
> about that (but no workarounds)
>
> Now, I finally asked support to Sun, and the Sun engineer
> who picked up the call told me that there is nothing I can
> do, that's the way Solaris2.5 works, use allways /etc/halt or
> reboot, and then closed the call!
> It seems to me very strange that is impossible to use the
> "shutdown" command when autofs is used.
>
> If anybody of you have a suggestion, it will be very
> appreciated.
>
> Luca Pizzinato
> pizzinato@eumetsat.de
> Tel. +49.6151.807323
> Fax +49.6151.807304
>

Cheers . . .

"People will accept your ideas much more readily if you tell them that
Benjamin Franklin said it first."
_____________________________________________________________________________
 | _ \ Amanul Haque
 | | \ \ Pencom System Administration
 | |_/__/__ __ 9050 Capital of Texas Highway North, Austin, TX 78759
 | _/ ___/| \ Email: ahaque@pencom.com
 | | \___ \| \ \ Pager: (708) 643-7331
 |__| /____/|_|__\___________________________________________________________
           Pencom Web Page : http://www.pencom.com


attached mail follows:


Try this...

# cd /etc/init.d
# vi volmgt
Line 21 should read;
    /usr/bin/kill ${PID} 1>/dev/null 2>&1
add another line under this to read;
    sleep 5

save the file.
# cd /etc/rc0.d
# ln ../init.d/volmgt K73volmgt

Now try to do your shutdown command.

Let me know if this works.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  Rick von Richter | Phone: 619-552-6222
  Systems/Network Admin | Fax: 619-552-6221
  Maintenance Warehouse | Email: rickv@mwh.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     Science is true. Don't be misled by facts.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

> From sun-managers-relay@ra.mcs.anl.gov Fri Mar 14 01:37 PST 1997
> Date: Fri, 14 Mar 1997 09:39:18 +0100
> From: Luca Pizzinato <Pizzinato@eumetsat.de>
> To: sun-managers@ra.mcs.anl.gov
> Subject: Shutdown hangs at the end - NO SOLUTION ???
> Mime-Version: 1.0
> Content-Disposition: inline
>
> Dear SysAdmins,
>
> some weeks ago I wrote
>
> ############################################
>
> Hi List,
>
> does anybody have any idea why my Sparc1000 (Solaris
> 2.5) cannot arrive to the 'ok' prompt when a 'shutdown -g0
> -i0 -y' is issued? It arrives very close to the end, then it says
> that /home is busy, /opt is busy, then 'System is down.', but
> the prompt is not reached.
>
> Thanks in advance. I'll summarize.
>
> Luca Pizzinato
> pizzinato@eumetsat.de
> Tel. +49.6151.807323
> Fax +49.6151.807304
>
> ############################################
>
> I received only an answer by Ray Trzaska, pointing me to
> the command "umountall" in the /etc/rc0 script, and
> suggesting me to replace it with "umountall -s".
> Well, it didn't work but the problem must be somehow
> related to conflicts between that command and the fact that
> automounted filesystems are still mounted and the
> automountd daemon is already died, so they cannot be
> umounted. There is also a specific bug in the SunSolve CD
> about that (but no workarounds)
>
> Now, I finally asked support to Sun, and the Sun engineer
> who picked up the call told me that there is nothing I can
> do, that's the way Solaris2.5 works, use allways /etc/halt or
> reboot, and then closed the call!
> It seems to me very strange that is impossible to use the
> "shutdown" command when autofs is used.
>
> If anybody of you have a suggestion, it will be very
> appreciated.
>
> Luca Pizzinato
> pizzinato@eumetsat.de
> Tel. +49.6151.807323
> Fax +49.6151.807304


attached mail follows:



> From sun-managers-relay@ra.mcs.anl.gov Wed Feb 26 00:32 EST 1997
> Date: Tue, 25 Feb 1997 18:47:40 +0100
> From: Luca Pizzinato <Pizzinato@eumetsat.de>
> To: sun-managers@ra.mcs.anl.gov
> Subject: System shutdown hangs at the end
> Mime-Version: 1.0
> Content-Disposition: inline
>
> Hi List,
>
> does anybody have any idea why my Sparc1000 (Solaris
> 2.5) cannot arrive to the 'ok' prompt when a 'shutdown -g0
> -i0 -y' is issued? It arrives very close to the end, then it says
> that /home is busy, /opt is busy, then 'System is down.', but
> the prompt is not reached.
>
> Thanks in advance. I'll summarize.
>
> Luca Pizzinato
> pizzinato@eumetsat.de
> Tel. +49.6151.807323
> Fax +49.6151.807304
>

After the "System is down." message, try pressing Stop/A (or L1/A).

--
  Frank Pardo  <fpardo@tisny.com>
  Transaction Information Systems
  New York City

The scholar's ink outlasts the martyr's blood. -- Irish proverb

attached mail follows:


>If anybody of you have a suggestion, it will be very
>appreciated.

I tend to use init to change runlevels, init 0 works fine on my 2.5 ss5's
which use autofs. I have heard from several sun sources (engineers & course
instructors) that init is the preferred way to do this.

Robin

-- 
Robin Marquis                  __   _____   __
____________________________  (__\_/     \_/__)  ________________________
rmarquis@ordsvy.gov.uk         `-\\\_____///-'       Something witless 
Senior Project Officer            ( \___/ )             this way comes...
voice :+44 01703 792747           |\_____/|       
fax   :+44 01703 792078          =| |///| |=       
  PGP key available                \|\|\|/ 
  Ixion BMF '92 Blade               |///| dspace.dial.pipex.com/rmarquis/
__________________________________  |___|  ______________________________

attached mail follows:


Luca Pizzinato wrote:

  Dear SysAdmins,

  some weeks ago I wrote

  ############################################

  Hi List,

  does anybody have any idea why my Sparc1000 (Solaris
  2.5) cannot arrive to the 'ok' prompt when a 'shutdown -g0
  -i0 -y' is issued? It arrives very close to the end, then it says
  that /home is busy, /opt is busy, then 'System is down.', but
  the prompt is not reached.

  Thanks in advance. I'll summarize.

  Luca Pizzinato
  pizzinato@eumetsat.de
  Tel. +49.6151.807323
  Fax +49.6151.807304

  ############################################

  I received only an answer by Ray Trzaska, pointing me to
  the command "umountall" in the /etc/rc0 script, and
  suggesting me to replace it with "umountall -s".
  Well, it didn't work but the problem must be somehow
  related to conflicts between that command and the fact that
  automounted filesystems are still mounted and the
  automountd daemon is already died, so they cannot be
  umounted. There is also a specific bug in the SunSolve CD
  about that (but no workarounds)

  Now, I finally asked support to Sun, and the Sun engineer
  who picked up the call told me that there is nothing I can
  do, that's the way Solaris2.5 works, use allways /etc/halt or
  reboot, and then closed the call!
  It seems to me very strange that is impossible to use the
  "shutdown" command when autofs is used.

  If anybody of you have a suggestion, it will be very
  appreciated.

  Luca Pizzinato
  pizzinato@eumetsat.de
  Tel. +49.6151.807323
  Fax +49.6151.807304

 Use 'init 0' That is the cleanest way to shutdown solaris. That
runlevel stops all the processes 'nicely' , closes files, syncs
filesystem, stops daemons etc. Halt or fasthalt merely does a killall !
Shutdown should do roughly the same as init 0, but it obviously getting
upset about somethig ....

Simon



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:11:48 CDT