OT: ECC Memory - WAS: Re: [Samba] Possible Filesystem Corruption with Samba 3.0.25a (with XFS and LVM)

[Charles Marcus]

John Drescher, on 6/26/2007 4:50 PM, said the following:
>> I recall reading somewhere that ECC memory was considerably slower than
>> non-ECC, and its benefits was mostly sales hype - ie, its ECC was not
>> precisely reliable...

> Neither are true.

>> Anyone know of an authoritative answer to this question?

> I use ECC memory in all of my servers and the main benefit is that if
> you ever get a single correction you probably have a hardware
> problem. On these systems machine check exceptions are logged in the
> bios for each memory problem seen. If you see more than one of these
> it is highly likely that your hardware is the problem.  If you have
> the correct hardware and your kernel has the CONFIG_X86_MCE and
> CONFIG_X86_MCE_INTEL or CONFIG_X86_MCE_AMD you should be able to view
> these errors on a running system by executing mcelog.

Hi John,

Thanks for the response...

This is what I was actually suspecting, but had no clue about these 
errors being logged in the BIOS - and thanks for the pointer to being 
able to read these on a running system - I'm off to read and test/implement.

-- 

Best regards,

Charles