Solaris 8 system won't shutdown gracefully 1

[angelshark]

Hello,

We have a crontab entry that reboots the system with the command 'init 6' every monday morning (4:00) but it stopped working 2 months ago.

The system won't shutdown or reboot using the commands: init or shutdown. When I type those commands it gives back the command prompt after a couple of seconds. The command 'reboot' does work, however it does not shutdown the system gracefully (it won't run the /etc/rc0.d scripts) so that's not a solution.

Have you ever run into this problem? Any help or suggestions are appreciated.

Thanks in advance.

 

[KenCunningham]

I had the same problem a while ago and had to resort to the reboot command. See Thread60-169326 for the discussion at that time (can't quite believe how long ago it was!!).

If anyone has an update I'd be interested to hear it.

 

[kHz]

Have you added any new kill scripts (i.e., /etc/rc0.d/K*)?

 

[bfitzmai]

Angle... When I had boot and shutdown problems, I put echo commands (Example: echo "S76lmgrd") in my RC sripts that told what script was running. That way I could localize the problem area.

 

[angelshark]

Hello,

I didn't add any new kill scripts recently. However I installed the recommended Solaris 8 patch bundle recently and that might be what triggered this problem.

Let me know if you have another suggestions.

Thanks.

 

[KenCunningham]

The only other thing I would add is to question why you reboot your server every week? There's usually no need (these aren't Windows machines after all). We only reboot if a patch requires it or if there's another pressing need.

 

[angelshark]

I agree with you but my boss wants to reboot every server once a week. He says that cleans the zoobie processes from memory that users leave running.

I would do as you say. Only if a patch or if a hardware upgrade needs it.

 

[KenCunningham]

Zombies don't really have detrimental effect unless they're hogging resources or causing licensing issues. However, I guess if the boss says it, it must be true, huh?

 

[angelshark]

Yes, you are right about that.

What do you suggest that we do about the shutting down problem? Should we apply the latest patches?

The kernel version is SunOS 5.8 Generic 117350-08 Aug 2004.

Thanks.

 

[KenCunningham]

You could try the latest patch cluster in the hope that it fixes the problem. Let us know.

 

[VicMcD]

I had a similar problem on Solaris 8 but found this in Sunsolve. I have not tried a reboot yet.

Document Audience: SPECTRUM
Document ID: 78857
Title: Solaris[TM]: shutdown -i6 -g0 -y failed to reboot machine.
Synopsis: Solaris[TM]: shutdown -i6 -g0 -y failed to reboot machine.
Update Date: Fri Oct 29 00:00:00 MDT 2004
Products: Solaris 9 Operating System, Solaris 8 Operating System, Solaris 7 Operating System, Solaris 2.6
Technical Areas: Hang, Reboot

Keyword(s):shutdown, init, reboot, hang
Problem Statement:

Intermittently run levels are changed via the init(1m) command or the
shutdown(1M) command. The system can fail to complete the change in run levels.

This is most commonly seen when changing from run level 3 to run levels 0, 5 or
6.

The problem has been present in all releases of Solaris[TM] since Solaris[TM]
2.0.

Resolution:

Now, there is only one workaround. The official patch will be released when the
verification of the IDR patch is finished.

Reducing the number of services started from inittab will reduce the chances of
seeing this.

Temporary Workaround:

If you put the following at the bottom of /sbin/rc6:
exec /sbin/uadmin 2 1

At the bottom of /etc/rc5:
exec /sbin/uadmin 2 6

At the bottom of /etc/rc0:
exec /sbin/uadmin 2 0

then it will reduce the probability of seeing the problem:

Watch out when editing these files. I found that changes to rc6 were replicated in rc5 and rc0. The only way round it was to :-
cp rc6 rc6.old
mv rc6 rc6.cpy
vi rc6.cpy
mv rc6.cpy rc6 and so on....

 

[KenCunningham]

Useful, thanks. How remiss of Sun not to realise that these files are linked and advise accordingly!

 

[sunixadm]

Hi,

If you use the workaround, your services will "NOT" be shut down gracefully!

This could have a terrible impact on DB type apps.

-Joe

 

[angelshark]

If that's the case, it's not a "real" workaround cause the reboot command will do the same thing.

 

[VicMcD]

The /sbin/rc* scripts run the equivalent /etc/rc*.d scripts before calling the uadmin command. The databases would be shut down at this point.

 

[angelshark]

What does the uadmin command do exactly?

Thanks.

 

[angelshark]

I have made the modifications to the /sbin/rc* files but unfortunately the server still doesn't reboot.

Loading the latest patch bundle isn't an option at this point.

Could an hung application prevent the server from rebooting ?

Any other suggestions ?

Thank you very much.

 

[VicMcD]

Sometimes the system can spend a long time trying to unmount busy file systems. Why don't you put some echos in the /sbin/rc* scripts to see where it is getting to?

 

[coffeysm]

I would like to add I have the same problem with a E450 and a brand new V880. This occurred after patching, I used the uadmin command before even reading this post and I would not recommend it, unless it is absolutely necessary. Depending who you ask the uadmin command is great or horrible, my opinion is it's flaky. 9 times out of 10 I had to run fsck. The only thing good about it is it brings the system down in less than a minute.

 

[coffeysm]

I forgot to mention if you do add that to your scripts. Add the sync command. that way it flushes out all the buffers from memory and might help minimize having to run fsck.

exec sync && sync && uadmin 2 x