FC4 - crash with partial trace included

cwho80 · #1 15th May 2006, 04:48 AM

One of our servers (dual CPU 3GHz Xeon Dell) which run Fedora Core 4 x86_64 (2.6.11-1.1369_FC4smp) crashed mysteriously with something that looks like a kernel trace. We are running in the 64bit configuration. We were able to obtain a partial trace from the screenshot of the server's final state. <Screenshot attached as JPG file> Has anyone encountered such an error before, or have an idea or lead as to the cause of the crash?

The usual applications the server would be running are Tomcat 5.0, MySQL 5.1, ssh and uses NFS.

Any help would be greatly appreciated. Thanks.

ibbo · #2 15th May 2006, 01:39 PM

I have had a few of those too. (it tells me it will continue in N seconds when it happnes "which is perhaps once per week currently"). I reboot and its fine.

Thats said since I upgraded to kernel 2.4.16 from 2.4.15 my hardware started having fits throwing APCI errors the works, so it could be an hardware issues. I relay dont know (I just know your not alone).

Ibbo

ibbo · #3 15th May 2006, 01:40 PM

posted twice ...!!

Jman · #4 16th May 2006, 06:09 AM

The trace and any other useful information should be included in a bug report probably against kernel.

cwho80 · #5 16th May 2006, 11:56 AM

Thanks for the info. However it is one of the primary servers for our application, so monitoring and rebooting frequently would not be an option to us :|

That said, we have been running for almost a month and this is the first time that this has happened. Of course, the CPU and IO usage has been ramping up steadily as we add load. What puzzles me is all the "cdrom" mentions in the trace. We don't even use the cdrom drive on the server - its locked and sitting in a datacenter.

---------------------------------------------
Maybe the resident rodents are dismantling it slowly for parts for their own cheese-based supercomputer

cwho80 · #6 16th May 2006, 12:01 PM

We do have information that it was copying a large file (3+G) to another computer on the network via NFS at that time. The transfer was about 90% complete before the crash. Load on the machine at that time was not high (load average from uptime varying about 1-2).