Re: kernel: EDAC k8 MC0: GART TLB errorr: transaction type(generic), cache level(generic) [message #16443 is a reply to message #16438] |
Thu, 06 September 2007 15:40   |
khorenko
Messages: 533 Registered: January 2006 Location: Moscow, Russia
|
Senior Member |
|
|
sspt wrote on Thu, 06 September 2007 18:42 | MC0 refers to a DRAM controller but all the memory has been tested for about 16 hours without any error.
Since there's usually a quota issue i guess there's some I/O at the moment, is it possible that's related with the HDD and not the RAM controller?
|
Hi,
well, i personally would not suspect the HDD in this situation: 5 minutes passed between EDAC report and VZQUOTA exceed issue, too long to suspect their direct dependence. At the same time i suppose both these messages mean that node's load raised at that time.
i suggest you to test again the memory or just replace it if possible (may be temporarily - just to test.) What tool do you use? If you used a userspace memory test, then it certain makes sense to retest with memtest86+ (yes, it requires the long node downtime, but it's much more reliable than userspace memory tests).
If it won't show the problem again, then may be cpu or even motherboard is guilty. It's harder to check but makes sense to run several 'cpuburn's along with userspace memory tests - they will give a pretty heavy load. (BTW, is there a new BIOS firmware available for the MB?)
One more possible thing is a power abnormals - if somebody comes to the work and switch on some powerful device somewhere near - this might give such a result...
Hope this helps.
If your problem is solved - please, report it!
It's even more important than reporting the problem itself...
|
|
|