K570 crashing after upgrade

[Marcel Burggraeve]

Yesterday a K570/2 at a client site has been upgraded by a colleague. At first it was a K570/2 with 1GB memory ( 8 x 128MB ) and it has been upgraded to a K570/4 with 4GB ( 8 x 512MB )
The system is running 10.20 ( 4GB can/t be used by 10.20, I know ) After installation of the hardware the system will boot into HP-UX but will crash within something like 10 minutes. Unfortunately nobody was at the console when the crash happened and all I know is that the display had CBFC on it. They refitted all the hardware, booted the system and it started. 10 minutes later another crash took place. After the second crash the client didn't trust the new hardware anymore and wanted the system back to it's original state. In the original state it's running fine. Since there's a very small timeframe for maintenance at this site we're very limited in testing what hardware ( or maybe software ? ) is causing these crashes. Updating in smaller steps is not an option due to the small timeframe. At the office we have a K370 in which we can install the processor and the memory but I haven't got a clue howto properly test them and I don't even know if the system crashed under heavy duty or during some kind of initialization of software or whatever. I allready received the OLDsyslog.log but there's nothing special in it ( I see the hardware paths, configuring of btlan4, info about the memory, starting of inetd, starting of hardware monitors and the last line is from xntpd )
In a few days we will receive the crashdump as well.
Anybody got any ideas how to find it what might have happened ?
Can it crash if there's more physical memory than it can address ?
Is there an easy way of stressing the cpu's and memory ( maybe a number of seti instances ) ?

[Eugeny Brychkov]

Take all files from /var/adm/tombstones (called tsXX, where XX is decimal number), zip them and attach to your next reply. If HPMC has occured I will be able to see and decode hardware component which is causing HPMC
Eugeny

[Khalid A. Al-Tayaran]

Hi,

Like Puala and Eugeny said also check /etc/rc.log and /var/adm/syslog.log.

Use ODE to find some clues (ode mapper)...

[Marcel Burggraeve]

Monday back at the office we will contact the client and ask them to e-mail the files in /var/adm/tombstones which will be posted to the forum.
Allthough we're not very experienced with analyzing crash files we will use the given tips and probably will learn alot ;-)

[Steven E. Protter]

I would think 10.20 might have issues with more memory than it can address. I assume you are doing this in an attempt to upgrade.

You might try booting to single user mode, mounting as many file systems as possible, and then trying to anlyze the crash dumps( on 11.x they would be in /var/adm/crash

Attaching a doc that MIGHT help.

SEP

[A. Clay Stephenson]

You shouldn't have a problem with that much memory - unless it's bad. CBFC - means that a branch to HPMC (High Priority Machine Check) failed. That very strongly suggests a CPU failure. A very probable cause of your problem is mismatched firmware on the CPU's.

[Paula J Frazer-Campbell]

Hi

This did not cause the problem but needs addressing:-

May 9 17:05:10 geuze vmunix: WARNING: Insufficient space on dump device to save full crashdump.
May 9 17:05:10 geuze vmunix: Only 1610612736 of 4026532864 bytes will be saved.


Increase dump to 4 gig

Paula

[Marcel Burggraeve]

Eugeny :

These are all the files from the Tombstones directory with the following timestamps:
-rw-r--r-- 1 root root 6773 Apr 1 08:45 ts95
-rw-r--r-- 1 root root 6773 Apr 17 22:03 ts96
-rw-r--r-- 1 root root 6773 Apr 21 11:17 ts97
-rw-r--r-- 1 root root 14714 May 9 17:05 ts98
-rw-r--r-- 1 root root 7906 May 9 18:12 ts99

ts98 and ts99 are from the day of the upgrade.