Help, my disk array has one dead member

[Chris Green]

On Wed, Mar 29, 2017 at 10:55:28PM -0400, Rashkae wrote:
> On 17-03-27 01:16 PM, Kevin O'Gorman wrote:
> 
> > Fortunately, I think RAID5 is a good fit for my use-case.  Overall,
> > it's compute-intensive, not so much DB-intensive.  My requirement of
> > the database is that it be large and reliable, not so much fast.  I
> > tend to be working on a small part of the data at a time.  I have
> > compute tasks that run for days with no disk activity at all.
> 
> The problem with Raid 5 is not performance, (which I think is kind of
> great,,, CPU's have long since gone past the point where parity calculation
> was an issue.)
> 
> It's reliability.  Generally, there are all kinds of common failures that
> can temporarily take out 2 drives, (example, say, one of your Controller's
> that has 2 drives attached fails or just locks up somehow.)
> 
> the result will be the kernel marking your array as damaged and dead. You
> can force the array to rebuild, but with the huge size of modern drives,
> having to way over 24 hours see if you succeed can be stressful.
> 
> So don't temp Murphy, if using Raid-5, make sure you have good up to date
> backup at all times.
> 
> If you need a system that can be recovered from failure the fastest, stick
> with Raid-1
> 
Stick with automatic, incremental backups to a separate drive.  Better
in almost every way!  :-)

-- 
Chris Green