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.

We've got two Dell R900 servers deployed in a well known managed hosting provider in the US. One of the Dell R900 servers has had its 128 GB memory (32x 4 GB) swapped out 6 times now. Every time the server chassis has reported the memory ECC fault at a different location to Dell OpenManage 6.5.

We've swapped out the complete chassis (including processors) twice and sent both into Dell for diagnostics and they claim to not find a problem.

Has anyone out there experienced anything along these lines and possibly know why the chassis display and OpenManage can't agree on the failure memory bank location?

share|improve this question
    
What operating system do you use? I had similar cases on R310, 510, 710 using older linux distributions. –  Nils Aug 2 '11 at 20:15
    
The machine having the issues is running Windows Server 2008 R2 SP1. The machine not having the issues is running Windows Server 2008 SP2. Both x64. –  CraigN Aug 3 '11 at 7:46
    
With elder linux-distributions there is a service called "microcode_ctl" that applies Intel-Microcode-Updates. This KB-article hints that there is a similar mechanism in MS-OSses. So what is the date of your last BIOS-update? –  Nils Aug 3 '11 at 19:10
    
How relevant is a processor microcode update in this situation? Surely this is more like a BMC BIOS issue with regards to memory bank layout? –  CraigN Aug 8 '11 at 9:13
    
I got a couple of cases where we had ECC errors with R3x0 servers (both Core i3 and XEON processors). The explanation was that they used a newer memory layout that couples CPUs with memory banks more closely. If the CPU gets some timings wrong ECC errors will result. The fix is to update the CPU-microcode... I`ve been told that HP servers with that layout suffer from the same problem (with the same solution). The Dell PowerEdge R90x series has the same memory layout... –  Nils Aug 9 '11 at 21:12
add comment

3 Answers

Current BIOS should be 1.2.0 This sounds as if W2K8SP2 contains a newer Intel-Microcode-Update than your current BIOS - or W2K8R2SP1 downgrades to a faulty microcode... Dell is not aware that the OS might change this.

share|improve this answer
    
Yes the BIOS is 1.2.0. I like to keep up-to-date. –  CraigN Aug 4 '11 at 16:45
    
At Intel I found a microcode update from 04/28/2011 - that`s newer than the one provided by the 1.2.0 BIOS. This is just a guess. Perhaps someone who is more deeply into W2K8 can elaborate here... –  Nils Aug 4 '11 at 20:20
    
@CraigN: Perhaps I`ve got a new trace now. We`ve got a kernel-crash on an R510. Dell says this might be due to a single core that went into sleep state. At what times do your ECCs occour? When there is almost nothing to do for the server? –  Nils Aug 11 '11 at 19:34
add comment

This sounds like an issue I have ran into with PE 1950's in the past. Sometimes when one updates the BIOS of a given server, it changes the supported memory list within the BIOS. Are you getting any multi bit ECC errors by any chance?

share|improve this answer
    
What do you mean by "supported memory list"? –  Nils Aug 4 '11 at 20:21
    
As far as I'm aware only Dell supplied memory is being used. The errors are single bit ECC errors. –  CraigN Aug 8 '11 at 9:13
add comment

For the most part enterprise level BIOS's have a list within them that support certain manufacturer PN's. I have seen update's knock out a certain PN and you have weird memory issues thereafter.

share|improve this answer
    
What do you mean "knock out"? Do you refer to the list F10 shows on a iDRAC6? –  Nils Aug 11 '11 at 19:36
add comment

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.