Not really a summary, rather a workaround:
Below is my original post. I got responses from:
scott r kulp <srk@aloft.micro.lucent.com>
"Peter L. Wargo" <plw@ncgr.org>
Scott suggested reseating the CPUs. Pete vaguely remembered something from
an FE handbook. I was going to drop by his place to pick them up but just
got too busy (He is in the same town).
I have come to somewhat of a realization that a 16x64 memory configuration
in an Ultra2 (2x296) OBP 3.19 Solaris 7 doesn't work. I tried three
different configurations of 12x64 that were rock solid under a high memory
load. Went to 16x64 and it croaked. Took the extra 4x64 out of that
machine and put it in another machine (8x64) and it stood up under a high
memory load.
I've done enough testing to convince myself that the 16x64 configuration
just doesn't work. I probably have not done enough testing to convince Sun
of this given the fact that these are all 3rd party mem modules.
If anybody is running smoothly in my problem configuration I would love to
hear about from them.
Tim
On Tue, 27 Apr 1999, Tim Carlson wrote:
> Hey folks (still working on a summary to my screen timeout question):
>
> Got a new problem.
>
> Upgraded a couple of Ultra 2 machines to Solaris 7 (no patches). Started
> to see the same error pop up on the two machines:
>
> panic[cpu1]/thread=2a1001e7d60: CPU1 Multiple Privileged UE Errors: AFSR
> 0x00000001.80200000 AFAR 0x00000000.6fc00000 UDBH MemMod U0504 U0404
>
> Now I was thinking.. "odd... upgrade the two machines and they start
> reporting memory errors in the same place?". So I did a little looking
> around and decided that the latest OBP/POST patch couldn't hurt.
>
> OBP 3.19.0 1999/02/04 21:01
>
> Well, one of the machines that I upgraded seems to be happy with this OBP
> patch (Ultra 2.. dual 296Mhz 768RAM (12x64)). At least it hasn't crashed
> yet. But the other machine (Ultra 2.. dual 296Mhz 1024RAM (16x64)) crashed
> again last night with the same errors. Both machines have a pretty
> constant load of 2 with simulation work running on them.
>
> I've got another Ultra 2 running Solaris 7, but it is only running the
> 32bit kernel because for some reason it didn't want to do the OBP upgrade.
> And yes I know where the jumpers are.. We have a service call in on that
> machine.
>
> A search of sunsolve doesn't show any bug reports of this nature for the
> Ultra 2 although there is an open bug report of a simlar error on an
> E3000.
>
> Has anybody seen this error? I'd rather not play around with memory
> configurations as these machines are in pretty constant use.
>
> Thanks
>
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:18 CDT