Stupid RAID question

[Lennart Sorensen]

On Tue, Oct 04, 2011 at 06:52:42PM -0400, William Muriithi wrote:
> Hmm, good explaination. I was not aware about this stuff. In short, they
> sell the less intelligent drive more expensive to discourage consumers from
> buying them.

Well that's my guess on the reason for the price difference.  I really
don't get the impression that the hardware is different in any way.
Of course non raid drives ship in vastly larger quantities too.  So the
economies of scale thing probably matters too.  But if they really cared
about that, they would have a jumper for it instead and let people choose
to change the jumper if they know what they are doing.  Of course jumpers
cost money too.

> One follow up question, does this too apply to linux software RAID? I mean,
> would software RAID eject a drive for taking too long to respond?

Yes.  The idea after all is that if you have raid a bad sector is no
big deal, just recognize it as bad, tell the OS "sorry I couldn't read
that", and then the raid will rewrite the sector after reading it from
the other disk and when it is written the drive can remap the bad sector
to a spare sector.  So waiting multiple minutes potentially while the
drive tries to fix the problem itself through brute force and stuberness,
the raid could have solved the problem in a few miliseconds.  So a raid
edition drive has a time limit on correction attempts, and if it can't
correct the error in that amount of time it will just return "read error".

To the raid controller (or software) a drive taking more than 30 seconds
or so to respond to a request might be totally dead.  No way to know,
so to get on with life the drive is kicked out of the raid even though
all it has is one bad sector.  Not really worth having to do a complete
resync for.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists