We are having a problem with a couple of 445 servers. All of them has 40Gb of memory and when oracle creates large databases and the memory usaage is more than 32Gb the server hangs. I have tested that this is not an oracle error. I just did a dd on /dev/zero to a 50Gb large file. If we have 32Gb of memory the dd is working. If i use 40Gb of memory the server stops responding when there is approx. 8,5 Gb left.
OS : suse 8/9
oracle: 9i

That sounds more like a SuSE vs memory or 445 vs 40GB issue to me.
(You're talking about IBM 445 e-Series machines?)

Is there anything visible on the machines, do you see a kernel
panic on their console, any indication as to what may be going
on in any of the logs?

And a completely random note: in the past I've seen weird problems
with IBM PC 330 machines if the RAM chips were inserted in the
"wrong order". On those machines with 4 sockets if we populated
the first two sockets with the smaller chips the machines would
freeze at random times during heavy load. IBM techies on site
found that out using a low-level debugging tool via serial console
and the issue was addressed in a new firmware release.


Cheers,
Tink

these didnt solved problem
problem still pending
RAM are in corect order
BIOS are updated.

any new commnets? plz help

heloo any updation?