[kwlug-disc] DRDY errors on drive that passes smartcontrol

[B.S.]

----- Original Message -----
> From: Paul Nijjar
> To: kwlug-disc at kwlug.org
> Sent: Monday, December 28, 2015 7:46 PM
> Subject: [kwlug-disc] DRDY errors on drive that passes smartcontrol
> 
> but when I network boot the machine into a Live CD and run
> gsmartcontrol, both the short and extended tests pass. I can also fsck
> the file systems successfully. 
> 
> I no longer trust the drive and do not intend to use it further. But I
> do not understand why I cannot find evidence that the drive is bad. 
> 
> This makes me worried that the problem is not the drive but the
> server. 
> 
> Do these symptoms make sense to anybody on the list?


So ... it might not be the drive. (But probably is.)

Does smartctl -a | grep realloc show anything other than 0?
(Your surface tests would have caused flaky sectors to get remapped, and if still within the slack space, wouldn't report an error.)

If you remount the drive ro and do a fsck, do you get anything interesting?

Put your memory through a memtest86 pass. (The DMA error caught my eye, there.)

I just had a stick of memory go bad and I didn't pick up on it for some time.

But btrfs started throwing off errors. (Which is why I didn't pick up on memory.)

Not fun.

fs checks showed fine but internal consistency, I expect for space to be allocated next, shot to heck.

btrfs send | receive showed fine, but compares against backups showed missing files. i.e. Even if your fsck shows fine ... doesn't mean you didn't lose stuff some days ago and by now such don't get fsck'ed to give errors.

If you can, clone what you can off to a spare drive? If you want to investigate further ... better to not be on a live filesystem.

Won't hurt to put the drive on a different port, see if errors follow the port.

I suppose the real problem is ... if you don't find / verify the problem ... how much of any of it, or which parts, can you trust going forwards.