Help, my disk array has one dead member

[Xen]

Kevin O'Gorman schreef op 24-03-2017 3:26:
> Well, okay, this should be my "real" last word.
> 
> I had no idea what to expect.  But I put my 3 drives into USB3 drive
> docks (my 3rd one arrived today).  mdadm is now projecting RAID5 will
> be finished in 4.6 days.  Egad!
> I guess it makes sense.  7.4 Tib to read, 3.7Tib of parity to write.

In general opinions are not great about RAID 5.

I once ran a RAID 10 but it was on a crappy AMD controller with 
more/less crappy software.

But there are some that prefer RAID 10 over RAID 5 because of simplicity 
and speed.

The extra disk cost is worth the advantage to them.

> I think it made 512-byte blocks.  Would anything go faster if I
> changed that or any of the other defaults?

I really don't know but that would not be very good. A stripe should be 
between ... well what do I know. I would suspect you'd want at least 4k, 
but these can cause rewrites... I don't know. Maybe someone else can 
answer. Sector per sector writes are just awfully slow you know...

I wonder how many times you'd actually write less than 4k...

Regardless.

Ideally RAID 10 has the throughput of a stripe and the access times of a 
mirror. It's not extremely failsafe but if the "right" drives fail you 
can handle 2 drive failures.

Well anyway.

Not something you can do right now right :p.