Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I had a server lock up this morning. Here is a screen shot from the console:

enter image description here

None of the messages from the screen shot mean anything to me. I have a feeling that the important stuff probably scrolled off the console. I can not find any of the messages from the above screen capture in the syslog, message, dmesg, debug logs or anything logged at all at the time of the crash. Shouldn't this stuff have been logged?

This is a Debian box running Proxmox. uname output:

2.6.32-4-pve #1 SMP Mon May 9 12:59:57 CEST 2011 x86_64 GNU/Linux

The server has been online for about a year with no other crashes and it booted up again just fine.

I would love to figure out what the issue might have been so that we can prevent it from occurring again in the future. But, from the evidence I have so far, I don't even know if this was a hardware or software issue. Ideas?

share|improve this question

2 Answers 2

up vote 5 down vote accepted

Exactly which Debian kernel release do you run? You can see the full version and revision numbers if you do "dpkg -l | grep linux-image".

It looks like you're hitting a fairly prevalent bug that I've seen strike numerous times: In kernels before 3.2 mainline, before 2.6.32.50 stable and before Debian's 2.6.32-45 (based on 2.6.32.50 stable), there's a clock overflow that will strike after ~208 days of uptime, which will in turn enable the potential of crashing. I don't know exactly what can cause the crash after that time; the patch itself is pretty vague about it too:

Although we may still have enough bits to store the value of ns,
in some cases, we may not have enough bits to store cycles * cyc2ns_scale,
leading to an incorrect result.

I've seen upwards of hundred crashes due to this issue, before it was determined what caused it and the patch was deployed.

The bug was discussed at length in the lkml at the end of 2011. There could be a possible link to this divide by zero bug, but I haven't found any conclusion.

TL;DR: The likely fix is to upgrade to Debian's linux-image version 2.6.32-45 or later.

share|improve this answer
    
This server is running Proxmox 1.8 which is based on Debian Lenny 5.0.8... however, I believe the kernel is supplied by OpenVZ: pve.proxmox.com/wiki/Roadmap#Proxmox_VE_1.8 ... I will check with the Proxmox guys to see about getting an updated kernel. Thanks! –  Curtis Aug 14 '12 at 18:17
    
Forgot to mention... "dpkg -l | grep linux-image" returned no results (probably because of this being a Proxmox server). –  Curtis Aug 14 '12 at 18:26
    
Since this crash has only happened once, it's hard to say if this is really the cause, but based on the evidence you present here, it seems the most likely cause (need to upgrade the kernel), so I'm marking your answer as being most helpful. Thanks! :-) –  Curtis Aug 16 '12 at 12:44

This is a screenshot of a kernel panic. The traceback is printed inside out, so whatever function finally killed the kernel is off the top of the screen, but starting from the top is a call to divide_error() in hpet_msi_next_event() divide_error() is defined in the kernel as a trap for FPE_INTDIV, so something in hpet_msi_next_event() attempted to divide by zero.

Unfortunately, the cause of that could be either hardware, software, or even just a transient bit flip error. (Are you using ECC ram?)

share|improve this answer
    
I'm not sure if this server has ECC or not... I know our server vendor only uses ECC in situations where we request it or the motherboard/cpu combination requires it. I don't see anything in my notes on this server requesting it, but here is what the invoice says: ASUS RS100-E6/PI2 1U Xeon X3430, 2.4G/QC/8M/2.5GTs 16GB RAM -(2x8GB) D3-1333 REx8Q Kingston RAM –  Curtis Aug 14 '12 at 18:03

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.