Intel server crashes unexpectedly

Hi

Thanks for your reply.

The background is that the server is working normally.  It has now shut down without warning on two occasions. When it shuts down it logs the above events in the ISM log.

The server works normally otherwise.

Thanks

Denis

Hi

Thanks for your input,  I am still processing this.  I have been told the shutdowns may have been caused by Microsoft Software Update services trying to apply an update automatically. This may or may not be the case but I have turned off SUS and am waiting to see if the server shuts down again.

The server is a dual processor.  The server is quite new and has only been in service for a few months. I have sent the logs to Intel to see if they can issue any guidance as to the interpretation of the logs and I am waiting for their feedback.


I basically need to know if the problem is hardware or software related so I can take appropriate action.

regards

Denis

My point wud be to play it safe. No harm done if everything is uptodate on the hardware frnt just in case its a case of hardware/software mismatch resulting in this error.

If its just a one off thing i guess u can feel safe that it wont happen again.

Dan

Both processors then agreed to the shutdown.  What is on "08 BMC LUN #0 " the event trigger?

Read this :

http://www.uwsg.iu.edu/hypermail/linux/kernel/0402.2/1203.html

and be aware that Intel has some problems with indirect addressing with autoincrement, wherein the orginal pointer gets incremented if if points to itself.  There is another problem that a vector during a bus operation can wind up with a randomly generated address and that causes all kinds of problems.

The cause is software and hardware.  The workaround is to write 7 no ops before the call.  This would seem to indicate a combination of hardware timing problems and no software to compensate for it.

The problem is either on the PCI bus or the memory bus, but is reported by the Baseboard Management Controller [BMC].  One patch is in for Linux at :

http://ftp.linux.org.uk/pub/linux/willy/patches/bmc.diff

It seems to be related to "Attached scsi CD-ROM sr0 at scsi0, channel 0, id 0, lun 0"  which is four zeroes in a row, not ideal for a SCSI configuration and seems to lend itself to a bad pointer with offset=0 and hence timing problems for a fast bus capture or usage.

The problem here may be esoteric.  In that a timing problem will be seen if a race condition exists such that the bus is captured too fast or too slow, and the addressing data is not in sync.  The result will be an address of zero, which is not acceptible post boot loader.

Pretty much, it seems that address data may have been stored and locked before it actually arrived [got all zeroes and thought it was valid].  This would indeed get things like segmentation fault, attempt to access protected area, and many others.

This, in turn, would return pointers to the boot area, and any use of those pointers would then corrupt the entire Operating Environment and the take out the Operating System, resulting in a reserved Halt.

This problem seems to be popping up everywhere.

Anyway, it's worth investigating.

Hi

Thanks for your help.  I think I now have a good understanding of the problem.  I have made arrangements to have the system bios upgraded and  I will monitor the server for a few days, and if all is well I will share the points with mysticalDan (Suggested the BIOS update first) and gineric (good info aided my understanding)

Thanks

Denis