Re: RAIDs and JBOD?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]



On Wed, Nov 11, 2009 at 12:07:21AM +0000, James Bensley wrote:
> 
> Also I was initially going to get a PCI-E SATA card to connect up all
> these drives and use mdadm to make a software RAID, for this
> particular setup is that ill advised or do people think this will
> suffice (simple because my budget is low and hardware RAID controller
> cards are more expansive, in my experience but if you know of a good
> bargain I'm all ears!).

The 3ware 9550SXU-4LP is ~US$300.  (The -8LP is almost $500.)  If you
only need to support four disks in your RAID this seems cheap (but your
low budget may be different from mine).  A bunch of other people on the
list swear by Areca, which I believe are slightly less expensive than
3ware.

> Just quickly, this brings me back to the issue
> of a hot swappable drive. Up time isn't critical as its a home server
> so I don't believe a hotspare is needed, in the event of a drive
> failure I can shutdown the server and fire up single user mode and
> have the RAID file system dismount and then replace the failed drive
> and rebuild the array, is this correct?

Theoretically, yes.  The hot spare saves you from being unable to
perform the shutdown before another drive fails, killing the array.
If your motherboard supports SATA, you could put four disks on your new
PCI SATA card, and two disks on the motherboard, to accomodate a
five-disk RAID5 plus a hot spare.  (Or run a four-disk RAID5 plus hot
spare if you don't want to add more disks.)  You don't need a hotswap
bay for this; simply let the RAID array rebuild with the hot spare,
then shut down and remove/replace the dead drive.  The hotswap bay is
what helps uptime; the hot spare helps protect reliability.

One other suggestion would be to run five disks in a RAID6 with no hot
spare.  When one disk dies, you get (more or less) equivalent
performance to RAID5, and still have a redundant disk; when you replace
a dead disk you can rebuild the RAID6 and have two redundant disks
again.  It's slightly more fault tolerant than RAID5 with the drawback
that all the disks are getting used, instead of having one disk that's
not used till needed.  (I haven't benchmarked RAID5 vs. RAID6, so
perhaps someone else can chime in.)

If you are really trying to go on the supercheap, then you can skip the
hot spare and go RAID5, but be sure to have a computer store close by in
case you do lose a drive, because if you lose another drive before you
get to the store say bye-bye data.  (Don't forget the backups!)

You probably don't even need single user to rebuild the array with the
new disk; just make sure your disks are marked clearly, so you know
which one has failed.

--keith

-- 
kkeller@xxxxxxxxxxxxx

Attachment: pgpmj06n55C43.pgp
Description: PGP signature

_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos

[Index of Archives]     [CentOS]     [CentOS Announce]     [CentOS Development]     [CentOS ARM Devel]     [CentOS Docs]     [CentOS Virtualization]     [Carrier Grade Linux]     [Linux Media]     [Asterisk]     [DCCP]     [Netdev]     [Xorg]     [Linux USB]
  Powered by Linux