I am wondering how long I need to run memtest in order to determine if my ram has any errors. My server had been locking up and freezing just about every 24 hours so I set it up to reboot daily which usually reduced the problem to once or twice a week. When I was able to catch the logs it was basically telling me that the ECC ram was not correcting the errors. Since then I've purchased 128GB of ECC ram and it has been running in the exact same server for a week straight with no reboots and no issues. 

 

I've been running memtest for over 140 hours now and it is on pass 5. No errors detected by it. I've read that if the ram does have problems 99% of the time the test finds an error within the first pass or so. I've also seen that the longer the test takes before finding an error the less likely an error would cause my server to have an issue and that if it doesn't find an error then it most likely is not the rams problem. 

 

But since I have replaced the ram resulting in the issue resolved, and the ram has no errors in memtest, what could the issue have been? Or does the test need to run longer? 

 

Assuming you're using ECC RAM that memtest won't detect corrected errors, check the board's system event log or system log.

Just now, johnnie.black said:

Assuming you're using ECC RAM that memtest won't detect corrected errors, check the board's system event log or system log.

I have checked the boards log and it does correspond to the errors I received in my unraid logs so I guess if I stop this test and check I may see them for this. 

 

Didn't know memtest did not detect ECC errors. I'm assuming this test hasn't froze like my server would due to how low intensive it is? 

Single bit error will be corrected by ECC (and nothing will show on memtest), double bit errors will halt the server to prevent data corruption.