ML110 G5 unexpected power down

[Wolfgang Ailec]

Hello all,
we have a very strange problem:
We are useing 56 ML110 G5 Servers on 56
different locations. The most are running without
problems, but approx. 10 servers unexpectedly shutdown power as you press the power button of the system. We have tried different firware releases of the BIOS and the LO100 card. Using the old 2008 BIOS and LO100 firmware, we got the event log message

Generic 09/07/2009 08:39:42 Power Am Upper Non-critical-going high Assertion

Generic 09/07/2009 08:39:42 Power Am Upper Critical-going high Assertion

but after upgrading to the actual BIOS and LO100 firmware relases, nothing is in the event log area. We already run the HP diagnostics CD-Rom and also the server are running a long period of time in our labs, but if we put them in the different locations, the same effect appears (unexpected power down without eventlog entry and without informing the OS about the shutdown).

Ca anyone give me a hint, especially whats the meaning of the eventlog entries above. Dones this mean, that the Power Supply voltage ist too high?
Thanks in advance,
Wolfgang

[Diego Castelli]

Which O.S.?
IF WIN {
What about the power settings in Windows?
What in the DEVMGMT.msc? any yellow bang?
}

[Wolfgang Ailec]

No, we are running a linux OS, unbuntu 8.10 server without ILO capabilities.
I installed freeipmi 0.7.11 and configure it to read all the values. Everything seems to be normal. Also the HP offline insight diagnostics can't find something.

Wolfgang

[Diego Castelli]

i'd take a look at the power supplies revisions.

[Wolfgang Ailec]

OK, it takes a lot of time to get one of the
problem servers.
I take a look at the power supply, here are the informations on the label:

Model: PS-6361-4HF1
S/N: 00703949
CT: 591820AM7WX33Q
Rev: 0A
R0851

Could this be a power-supply from a series with troubles?
Thanks a lot,
Regards
Wolfgang

[Diego Castelli · Accepted solution]

You should check if the problematic servers have the same revision of PS and, as a cross-check, if the non-problematic servers have a different one.

No minidump at all at the prob. servers??
Another thing to keep in mind is that STOP 9C is often logged in the minidumps if the problem is kinda hardware, but you have to check the O.S. settings about using the minimal (64k) dump image: it must be set so, otherwise the server probably cannot dump the whole memory 'cause it doesn't have the time (and you remain wihtout logs).

Does perhaps the servers have a E200 smart array controller?
if so have a look here:
http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=c00715130〈=en&cc=us&taskId=135&prodSeriesId=1157688&prodTypeId=329290

Strange problem, anyway.. it seems you discovered something HP should be informed...

[Diego Castelli]

sorry i can't give you more info about configuring the memory dumps in linux...

[Wolfgang Ailec]

Hello Diego,
thanks a lot for your help, I will give this a try. It seems to me, that you are 100% wright pointing me to the power-supply, because if I exchange the power supplies, the problem moves.
Thanks a lot!
Regards,
Wolfgang