Help, my disk array has one dead member

[Xen]

Karl Auer schreef op 26-03-2017 11:02:
> On Sun, 2017-03-26 at 09:47 +0200, Xen wrote:
>> > http://www.zdnet.com/article/why-raid-5-stops-working-in-2009/
>> What he says is that with high capacity drives there is a high 
>> likelihood of a unrecoverable error while reconstructing a failed
>> drive.
> 
> Yes. But I was not holding him up as being a good source, just as a
> starting point for the discussion, because it was quite an influential
> article, and started a LOT of discussion about the best RAID
> configurations and so on. As you point out, he is slightly confused
> about how probability works.
> 
> The point I was trying to make is that in any sufficiently large
> collection of data, failure becomes a statistical certainty. RAID is
> one approach to that.
> 
> You asked, remember :-)

Yes yes, I just meant to get some clarity on the actual risks for 
personal reasons.

See backups are nice but bit errors, a single sector in some file going 
wrong.

I have video files that used to play well before and suddenly now they 
have file corruption, it seems.

If there is corruption happening on the copy level, that is a matter of 
concern to me.

I have now read 33x 10GB = 3.3TB and there is no discrepancy yet between 
these reads :p.

> I want a chance to fix things without everything stopping while I do -
> and that is what RAID offers.

Aye, but.

I had a RAID 10 system on AMD raid. I shouldn't even mention it here, so 
embarassing.

I had brought the thing down to 2x disk because of reasons. So it ended 
up a simple stripe, no longer a mirror of stripes. No issue. Then I 
added a 3rd disk back to rebuild one of the sides to become a mirror 
again. Rebuild fails.

Wut? The rebuild completes at 99% and then suddenly it says it has 
failed, and it won't add the drive to the array. That's nice.

RAID is pointless if the rebuilding stage is not handled well. On 
Synology NAS devices the biggest problem they have is people not knowing 
on how to do the RAID rebuilding on multi-disk systems and inexplicable 
"drive drops" out of arrays that people just don't understand.

This technology needs to be failsafe to be any use at all.

The amount of issues these devices have with drives suddenly dropping 
out of an array is just remarkable.

On my own RAID 10 system, when it still had 4 drives, regularly a drive 
would suddenly be dropped and I had to rebuild it for no reason.

I know, I know, firmware RAID. Still, you expect something better.

HighPoint RocketRAID is also firmware RAID but you expect something good 
because it is a renowned brand.

However, once you boot Linux the kernel driver says something like:

WARNING: RocketRAID BIOS CORRUPTS DATA. (...). Try not to use the (3rd) 
sector and do not use the last 2GB of the disks. However, the risk of 
corruption is still going to be there.


That's great.

You buy a controller to have RAID functionality and then the controller 
itself is the problem.

Really splendid.