Re: RAID6 - CPU At 100% Usage After Reassembly

[Wols Lists <antlists@youngman.org.uk>]

On 11/10/16 04:53, Brad Campbell wrote:
> On 07/10/16 19:23, Francisco Parada wrote:
> 
>> issue with the model drives that I'm using.  I've got WD 3TB Green
>> EZRX drives, which I recently found out via the RAID Wiki, that they
>> didn't have error correction (after I spent over a thousand dollars on
>> the drives in 3 years) ... Had I known better, I would have opted for
>> different models.
> 
> That's not entirely true. EZRX drives have just as much "error
> correction" as any other drive. What they don't have is the ability to
> rapidly terminate their error correction (WD call it Time Limited Error
> Recovery or TLER). Nothing wrong with that provided your OS storage
> stack timeouts are set appropriately, as without that the drives will
> take longer to process the error than the OS will wait and things get
> ugly resulting in drives kicked from arrays.
> 
> Sure, they're not the best drives for the job, and buying new I wouldn't
> recommend them; but disable periodic head parking and spindown, set the
> timeouts appropriately and they'll do the job until they need replacing.
> 
I've put your timeout fix script on the wiki :-)

It strikes me this would be a good addition too - I could probably
manage something of the sort, but do you or anyone have the code to
detect a green, and do that "disable parking and spindown"? It's a pain,
but having a script that deals with all known and common issues with
drives is probably a good thing.

I ought to go through that script and comment it heavily - I know the
experts will say "Why?" :-) but (although I may be a guru elsewhere)
this is an area I'm learning - hopefully fast - and comments are helpful
when reading code :-) They can always be deleted in production.

Cheers,
Wol