[ale] The perpetual question: best current HDD?

[Phil Turmel]

On 01/08/2013 11:20 AM, Derek Atkins wrote:
> Hey all,
> 
> I'm looking to replace some 1TB HDDs in a s/w RAID-10 array with some
> 2TB models.  The existing drives have been running flawlessly for a few
> years, so they are due to get swapped out anyways.  I did have one disk
> fail a year or so ago so it was swapped out, and I bought a cold spare
> at the same time so I have one more spare (of the same type/model as the
> replacement drive).  So I'm looking for another pair of drives that I
> can use as the mirrors (so each mirror has one of type/batch-A and one
> of the yet-to-be-bought set of drives).
> 
> Of course, when I bought the drives warranties were 3 or 5 years, not
> the '1 or 2' years they are now.  So I'm looking for the "best value"
> 2TB drives available today -- lowest price for highest quality + good
> warranty.  It looks like I can pretty much only choose between WD and
> Seagate nowadays -- I guess lots of consolidation in the market?  (My
> existing drives were Hitachi, which in my experience were always great
> drives).
> 
> What's the current going theories and best practices?  Any concrete
> suggestions (links to NewEgg or some other vendor would be appreciated).

First, you might want to browse the archives of the linux-raid list, and
see how many casual raid users get burned by non-enterprise drives.

That said, if you just can't afford enterprise drives, you *must* do one
of two things:

1) Buy drives w/ scterc support, and use a boot script to make sure it's
set to less than 30 seconds on each drive.  Hitachi Deskstar 7k3000
family and WD Red are the only ones on the market at the moment, I believe.

2) Use a boot script to set 120 second controller timeouts on all
devices in raid.  Yes, that'll play havoc with some services, but
otherwise the first real read error encountered can take out your entire
array.

And no matter what kind of drives you get, scrub the arrays on a regular
schedule from a crontab.

Phil