[CentOS] RAIDs and JBOD?

[Robert Heller]

At Tue, 10 Nov 2009 21:26:37 +0000 CentOS mailing list <centos at centos.org> wrote:

> 
> Hey Guys,
> 
> I have some questions regarding a new home server I am going to build
> in the hopefully very near future (ASAP, I just need to finish
> planning everything and this is the penultimate hurdle), I will be
> creating a software RAID...
> 
> Lets say I have three drives "knocking" around which are all 1TB SATA
> II drives but each made by a different manufacturer. I am going to
> guess that these couldn't be used in a RAID 5? Or could they?

They probably could be.  The RAID system would use the size of the
'smallest' disk as the base size for each disk. That is if you drives
were *actually* 1.02TB, .985TB and 1.12TB, the RAID system would use
.985TB of each disk, fully utilizing the .985TB disk, and leaving a
'small' amount of unused space on each of the 1.02TB and 1.12TB.  Oh,
the sizes will be much closer -- that is you might be lossing only a few
sectors here and there.

> 
> However could a similar result of 2TBs of data with redundancy be
> achieved with JBOD?
> 
> Also regarding RAID 5, three drives of data to one for parity is the
> max ratio I believe? I.e. to expand this by adding another data drive,
> the original parity drive would no longer cover this and another would
> be required, is this correct?

No.  You can use as many disks as you like for RAID 5.  The 'parity' is
not actually 1 bit.  The capacity of a N disk RAID 5 (where N >= 3), is
(N-1)*sizeof(one disk).

> 
> One more question about hot swappable drives, I understand that you
> can create RAID arrays with and without hot swappable drives but I am
> confused by this concept. I'm my experience with RAIDs I have only
> every delt with a RAID 1 that has degraded. I simply set the drive as
> offline, replaced it, set it to online and the RAID rebuilt itself all
> without restart the server and operation was never interrupted. So we
> can presume the server had hot swappable drives enabled yes? (It was a
> hardware RAID). With a software RAID is this still achievable?

Hardware RAID system almost always had hot swappable drives, esp. SCSI
ones.  And many of the old hardware RAID SCSI server boxes were equiped
with hot swappable drive bays.

For software RAID, it depends on the controller and what the driver for
that controller supports.  It also depends on how the drives are
mounted.  You *can* 'hot' [un]plug conventually mounted drives (eg
remove the cover of the *running* machine and reach in and pull the
data and power plugs off the disk in the *correct* order), but it is
tricky (and not really recomended).  It is far easier to get a hot-swap
chassis.  If the controller supports hot swapping AND the driver
supports the controller's hot swapping, yes, you can hot swap with
software RAID.

With RAID 5, what you want is a 'hot' spare: an additional disk that is
not part of the array, but is associated with it.  What you do is 'fail'
the drive you want to pull.  The system will then 'rebuild' itself using
the 'hot' spare.  You then 'remove' the 'failed' disk from the RAID set,
spin it down (requires the proper incantation, such as with sg_start
from sg3_utils), pull the drive, insert a new drive, scan for the new
drive (eg sg_scan), spin it up (sg_start), partition it (if necessary),
and then add it as a hot spare.

> 
> Thank you for reading.
> 
> Regards,
> James ;)
> 
> Charles de Gaulle  - "The better I get to know men, the more I find
> myself loving dogs."
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> http://lists.centos.org/mailman/listinfo/centos
> 
>                                                                         

-- 
Robert Heller             -- 978-544-6933
Deepwoods Software        -- Download the Model Railroad System
http://www.deepsoft.com/  -- Binaries for Linux and MS-Windows
heller at deepsoft.com       -- http://www.deepsoft.com/ModelRailroadSystem/