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 recently built a small cluster for running Solr. The cluster consists of 12 Supermicro Blades running E3-1270V2 with 32GB of ram.

11 of these servers are running fine. One of them crashes on me constantly. When the server crashes it typically produces some output on the terminal. The first time it was:

double fault: 0000 [#1]

Hmm... thats pretty cryptic. Since then I've recreated the problem and gottem some more interesting messages.

Here's one example...

Here's another equally cryptic message...

Here's another example

Another interesting wrinkle is that I can fire up sysbench and max out the CPU without it crashing, but it's not until I start Java that it crashes reliably.

I've tried turning off the following CPU features:

  • Turbo Mode
  • C States
  • T States
  • XHCI

Is this just a bad CPU?

Many thanks!

share|improve this question
    
Have you reseated the CPU? Can you check the motherboard socket of the CPUs and check for bent pins? –  ewwhite Mar 10 at 15:24
    
Looks like defective RAM. –  Michael Hampton Mar 10 at 15:30
    
you need to run a memetest –  c4f4t0r Mar 10 at 15:45
    
I'm running memtester now. I've previously run sysbench to test RAM and that didn't generate any faults. Is there something specific that makes you think it's memory related? –  Zorlack Mar 10 at 15:56
    
I wouldn't conclude RAM until the DIMM has been moved to another slot. If the error stays with the slot, it's likely a processor socket issue. If it moves with the DIMM, it's a RAM issue. –  ewwhite Mar 10 at 16:11

1 Answer 1

I have had this type of experience with Nehalem and Westmere CPUs on HP ProLiant servers. In my case, the server would POST properly and recognize all RAM, but would generate machine-check exceptions tied to a particular slot after application load was applied.

If you haven't already, please try isolating the issue to a particular DIMM or DIMM slot to see if it follows to movement of the module. If the error persists and is tied to a specific slot... I'd suggest examining the CPU socket. Check the motherboard socket of the CPU(s) and take note of bent pins.

This is SuperMicro gear, so I don't know the warranty terms. But hopefully this is only RAM, as that's an easier replacement than a system board.

share|improve this answer

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.