4.8-RELEASE problems

[Andrew Bogecho]

Hello all,

After very little sleep and many hours wasted, this turned out to be a
memory problem. We resolved this by having only one stick in and trying a
make installworld, this action rebooted the machine with each stick other
than the first. We have since replaced the memory and the machine has been
running fine for 2 days now.. Thank you for your input Lowell.

To all a nice day.

Andrew.

> "Andrew Bogecho" <andrewb at cs.mcgill.ca> writes:
>
>> I had initially suspected bad interaction with the new raid card so I
>> removed the card, but still had the same problem when using a local
>> disk.
>
> Hmm.  Is there any kind of power-saving functionality enabled in the BIOS?
>
>> I then run memtest from the ports and got the following errors on the
>> "first" run:
>>
>>   Test 15:          Walking Ones:  Testing...  47
>> FAILURE: 0x00020000 != 0x00010000 at offset 0x01efcb30.
>> Skipping to next test...
>>   Test 16:        Walking Zeroes:  Testing...  52
>> FAILURE: 0xffffefff != 0xfffff7ff at offset 0x0101bbc0.
>> Skipping to next test...
>>
>> But, no errors for any of the continuing runs. Is memory a problem here?
>
> That does look suspicious, all right.  You could try running memtest
> again from time to time.
>
>> I had initially installed FreeBSD 5.0-RELEASE, that run very well, but
>> nis
>> and amd would die every 24 hours. As these were "very" necessary
>> services,
>> I decided to go back to 4.x. On 5.0-RELEASE there were no reboots at
>> all.
>
> You could always try -CURRENT, I suppose.  A little risky, but
> probably less so for production use than 5.0-RELEASE was.
>
>> How should I proceed now? I am thinking of maybe only running a single
>> CPU
>> kernel to see if that runs better.
>
> Not likely to matter, but worth a try, anyway.
>
> You could run a low-priority CPU-hogging job (or several) to see if
> it's really connected to low usage levels, or if it's actually
> time-sensitive.
>
>