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

My Windows Server 2008 R2 x64 machine has blue-screened twice in the last month. In both occasions, the reported STOP cause was a MEMORY_MANAGEMENT error, and the parameter 1 value was 41790 ("An unknown memory management error occurred", according to the official documentation).

This is a brand new machine with 4 GB RAM running IIS and SQL Server 2008. Workload is very light.

Any ideas of what could be wrong? What are the most common causes of this kind of fault? What diagnostic strategies would you employ?

Thanks!

Edit. I have run some memory tests on the machine as suggested. All tests passed and no issues were detected whatsoever. We have recently made some adjustments to the boot delay in the BIOS so that the RAID controller in the machine has time to initialise all the disks before the server tries to find an OS, and we haven't seen a blue screen since. Maybe this was it?

share|improve this question
    
That's a new one on me but if it's worked then that's worth remembering. I've never seen a drive spin up issue cause a problem precisely like this but it certainly seems plausible. –  Helvick Feb 9 '10 at 14:59

3 Answers 3

up vote 2 down vote accepted
+75

This can be caused by faulty memory sticks, motherboard faults, and driver issues.

Let us know the make & model of your hardware and we may be able to offer specific tools and methods for you to troubleshoot.

First off I would verify your drivers are all in order. If your server is an HP, and you build it from the smartstart CD, the drivers should be kosher and can be (tentatively) ruled out.

On the other hand if you downloaded and installed all the drivers manually, or are coasting on the default windows drivers, update them.

Once you've run out of easy tests on the drivers, run a stress test on the RAM. Again, your vendor may provide tools for this (HP's are on the CDs that shipped with the server). Some vendors have a tester built into the BIOS menus.

If both of these solutions come up blank then post up some more info about the hardware and build method and we can take it from there.

share|improve this answer
    
Thanks Chris. This server is at a colo facility and I will need to ask about the make and model. I will definitely ask technical support to check on the drivers. –  CesarGon Feb 6 '10 at 0:11

I would run memtest to test your RAM fully for hardware issues.

share|improve this answer
    
Wouldn't the Windows Memory Diagnostic tool that commes with Windows do the same thing? –  CesarGon Feb 1 '10 at 9:39
1  
Weird. Just today I had the Windows tool pass my memory, and then a subsequent run of memtest found a problem almost immediately. Explained why the machine would freeze up every so often. So I would say that no, they're not necessarily the same thing. Couldn't hurt to try memtest. –  Boden Feb 5 '10 at 4:35
2  
I believe memtest is more "torturous" than Windows Memory Diagnostic Tool. –  deddebme Feb 5 '10 at 13:28
    
I see. Thank you. –  CesarGon Feb 6 '10 at 0:08

This definitely looks like a faulty module, a poorly seated module or a manufacturing defect. It might even be an indication that there is a cooling problem with your server although that generally causes other things to fail rather than RAM. There is also a possibility that someone has misconfigured BIOS settings so it is worth checking those against the Server manufacturer's recommendations for the type and quantity of RAM that is installed just to be certain.

I install a lot of servers and I always make sure that there is a long Memtest86+ burn in (24 hours or more) before I sign off on any of them. In the couple of years I've seen failures in two separate batches of servers (out of about 50) both of which were manufacturing quality control issues (IMO) so I wouldn't be at all surprised if you were seeing this as a result of a defect.

Oh and with modern hardware make sure you are using Memtest86+. The original Memtest86 is still out there but isn't up to snuff for testing current generation hardware with lots of RAM.

share|improve this answer
    
Thanks, Helvick. I am adding an edit now to my original post. –  CesarGon Feb 7 '10 at 21:54

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.