Shutdown problem

[André Sorge]

Hello there,

At the end of a shutdown-process on a C3650 running hpux 11.00 following message occurs:

"sync'ing of buffer cache timed out
Some buffer may not have been flushed due to a non-responding device or
resource deadlock

System will continue ...."

Does anyone have a clue about this one? Checked for hardware-errors in logfiles, pdt, pim etc., but nothing logged... Maybe a patch-problem? I did not find out anything about this...

Thanks in advance,
Andre

[Joseph Loo]

hi,

it looks like something to do with your disks. you may like to monitor it.

# dd if=/dev/rdsk/c#t#d# of=/dev/null ibs=10k obs=10k

you may also like to check:

# cat /var/adm/syslog/OLDsyslog.log
the last few lines

regards.

[André Sorge]

Hi Joseph,

already checked syslogs, but nothing found what would reveal a defective disk.
Will try the read-test of the disks, just to be shure.
I had first a defective disk or defective memory in mind, but did not find any additional info pointing in that direction.

Any more ideas?

Thanks,
Andre

[Joseph Loo]

hi,

if you have no glance, probably do a

# sar -b 5 20

to monitor the buffer cache as well. it may be a case of applications read/write too frequent.

another test is before you shutdown, close the applications, do a ps command to check the number of processes still active.

regards.

[RolandH · Accepted solution]

See this document:

klick me -->http://www4.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000065140529

DocID=4000030837

Regards
Roland

[André Sorge]

Hi Joseph,

thanks for your help! Did not manage to connect to the machine yet, but here some further info.

The machine is shutdown by a script which does a reboot -h (not a shutdown, but I´m working on it). The user did not wait for the machine to shut down properly and pressed the powerswitch... (doh!) That was the first time the error occured. I´m not quite sure though, whether it occurs again...

I will keep on investigating...

Thanks a l

[Bill Hassell]

reboot -q (or reboot -qh to halt) is the fastest way to shutdown a machine. AS long as you're not running anything important, this works great. Just a few seconds and HP-UX will reboot or halt after syncing the disks. Oh, you're running a database? Well, you will need to run a db-verify to see how many broken chains there are, perhaps a db-reload to recover all the data...reboot -qh is a giant kill -9

reboot -q is great way to restart from single user mode. As far as the non-responding device, it will pretty difficult to find the reason as the state of the system at that moment has very little information available. Certaibnly bringing the machine up to date on all patches makes a lot of sense.

[André Sorge]

Hi Roland!

You did it! The users script issues a reboot -h, so that is the "problem"... When the powerswitch is pressed, there will be issued a reboot as well (on ws only), as far as I remember.
I will replace the reboot -h command with a shutdown -h and will teach my user a lesson in patience...

[André Sorge]

Hey guys,

thanks for the great support! You rock!

Best regards,
Andre