SMP freeces up after 6-14 days (3.3/3.4)

SMP freeces up after 6-14 days (3.3/3.4)

Post by Karl M. Joc » Mon, 10 Jan 2000 04:00:00



i have one problem machine where i have no more idea what to do. this is
a Tyan 1662 SMP Motherboard with 512MB Ram and 2 Pentium Pro 200 CPU`s.
the machine runs fine under heavy load for 6-14 days. then it simply
freeces up. the only chance is to power off and restart. witch gives
sometimes problems when rebooting. till now fsck repaired everything.
machine also has a adaptec 2940u2w controller with 2*8GB hdu's.

i think i will install a single user kernel for now. but this is really
downgrading :-(. and the machine needs both cpu's because of the high
loadage.

problem already happend with 3.3. continues now with 3.4.

many thanks for any ideas.

best regards,

karl

 
 
 

SMP freeces up after 6-14 days (3.3/3.4)

Post by David Malo » Mon, 10 Jan 2000 04:00:00



Try installing a kernel with DDB installed. When the machine freezes
up press CTRL-ESC to break to the de*. They type "t" to get a trace
of where the kernel was. Then do "c" to continue. Do this a few times to
see where the kernel is stuck.

If you can't break to the de*, this means the kernel is stuck
somewhere with interrupts disabeled, which will make the problem hard
to find ;-(

        David.

 
 
 

SMP freeces up after 6-14 days (3.3/3.4)

Post by Steven G. Kar » Tue, 11 Jan 2000 04:00:00





> Try installing a kernel with DDB installed. When the machine freezes
> up press CTRL-ESC to break to the de*. They type "t" to get a trace
> of where the kernel was. Then do "c" to continue. Do this a few times to
> see where the kernel is stuck.

> If you can't break to the de*, this means the kernel is stuck
> somewhere with interrupts disabeled, which will make the problem hard
> to find ;-(

Following up on David's suggestion.  If you are running X11 on the
system, you won't be able to break to the de*i at the console.  
You might want to set up a serial console for debugging.  See the
handbook for details.

--
Steve

 
 
 

SMP freeces up after 6-14 days (3.3/3.4)

Post by sa.. » Wed, 12 Jan 2000 04:00:00


Ok, I'm probably wrong because I know nothing about these things but I'm
going to say it anyway:

There was discussion on the linux kernel mailinglist about if some stuff
about whether an optimization to some smp spinlock asm code on intel
architecture should be used. They came to the conclusion that it works
on all processors but older pentium pros causing random crashes on these
machines. Someone mentioned that FreeBSD uses this optimization, maybe
this could be the cause?

--
        Wolter

 
 
 

SMP freeces up after 6-14 days (3.3/3.4)

Post by Karl Joc » Tue, 18 Jan 2000 04:00:00


I temporary have installed a single cpu kernel. This worked perfect
till today. But with the single cpu kernel the machine reboots. Again
no entry in the log files. Because of the weekend i was not in the
office, but it looks like a hard boot. Not really sure where to start
searching.....

just started a cvsup to do a make world tomorrow. Will then compile a
new kernel too and restart the machine. Will wait what happens
next.....

best regards,

karl

Quote:>>>>>>>>>>>>>>>>>> Ursprngliche Nachricht <<<<<<<<<<<<<<<<<<


(Steven G. Kargl) zum Thema Re: SMP freeces up after 6-14 days
(3.3/3.4):




> > Try installing a kernel with DDB installed. When the machine freezes
> > up press CTRL-ESC to break to the de*. They type "t" to get a
trace
> > of where the kernel was. Then do "c" to continue. Do this a few times
to
> > see where the kernel is stuck.

> > If you can't break to the de*, this means the kernel is stuck
> > somewhere with interrupts disabeled, which will make the problem hard
> > to find ;-(
> Following up on David's suggestion.  If you are running X11 on the
> system, you won't be able to break to the de*i at the console.
> You might want to set up a serial console for debugging.  See the
> handbook for details.
> --
> Steve

 
 
 

1. updating Kernel 3.3 to 3.4

I had a lot of luck removing the 3.3 source, and downloading the 3.4
STABLE source, and doing a buildworld; installworld; and then
recompiling my kernel.  I have had serious trouble getting the compat22
working properly with Netscape, it only works right on one machine, I
recompiled the src twice on there, rebuilt XFree86 3.3.6 twice and then
it worked without library errors.

Thanks,
Doug

--
Douglas A. Maske
CEO/President
Thin Blue Communications, Inc.
Management/Information Systems

Simple Solutions For The Small Business, & Home.

Address: 640 Anthony Trail Northbrook IL. 60062
Web: http://www.thinblue.net
Home: (847) 776-1606
Work: (847) 562-1886
Fax: (847) 562-1895
Cellular: (847) 452-0709

2. login invisibly ???

3. Anybody upgraded SP System from AIX4.3.3 and PSSP 3.2 to AIX5.1 and PSSP 3.4??

4. agppart SiS 745 Patch

5. Slackware 3.3, Slackware 3.4, SC875, n_53c8xx.s bootdisk

6. 48MB RAM only detected 16MB

7. Upgrade Slackware 3.3 to 3.4

8. proxy ARP daemon for *bsd?

9. 3.3/3.4 installation hangs after root fs mounted on MFS

10. Slackware 3.4 == 3.3!

11. 3.4 is not as stable as 3.3

12. How to downgrade from 3.4 to 3.3

13. FreeBSD Power Pack 3.3 and 3.4