On 27/07/2010 6:54 AM, John Almberg wrote:
John Almberg wrote:
If you have hardware controller with RAID capabilities, using native
RAID is better, otherwise look towards gvinum or maybe ccd; see also:
I've just been reading up on RAID in my Absolute FreeBSD book, and it
occurs to me that my client has a SCSI RAID drive chassis that he is
using stupidly...
It's a 14 bay drive, and he's currently got seven 32G drives stuck in
it, configured with RAID-0. This is the original 200G drive I was
talking about. It's a few years old.
Over the next few years, this guy is going to need lots of storage
for his videos.
After a bit of reading, I'm wondering if the best idea might be to
toss out those 32G drives and replace them with 3 big (say, 300G)
drives configured with RAID-5. It sounds to me like a RAID-5 array
can be expanded by adding new drives.
QUESTION: is expansion normally a matter of just plugging in a new
drive? Is the new drive automatically grafted onto the old drives? Or
do you have to go through a process like, backing up the data,
plugging in the new drive, reformatting the expanded array of drives,
and restoring the data.
I don't know the brand/model of the RAID drive chassis, but the
client thinks it can be switched to use RAID 5. I'm waiting for the
technical details, but assuming it can handle RAID-5 for now.
Answering my own question...
So its a HP 6402 / 128 RAID controller. From a quick skim of the
manual, it looks like the controller has to go through an 'expansion'
process when adding a new drive. This sounds time consuming, but more
or less automatic -- i.e., handled by the controller.
Sounds like this might be the best way to go.
It's been a while since I dealt with HP SCSI RAID, but ISTR that you'd
need to install and configure the 3 disks as a RAID 5 set, copy the data
from the 7x36GB array to the new array, (using a temporary mount point,
generally, and dump | restore) switch the mount points across so that
the /videos tree is the new copy, then remove the RAID0 set from the
controller.
You may or may not find that the RAID controller changes LUN IDs after a
cold start too, so LUN 1 (new RAID 5) suddenly becomes LUN 0 on the cold
start after the old RAID set is decommissioned and pulled. This is often
accompanied by a heart attack on the part of the person restarting the
server.
After that, though, expansion is a cinch - but it will be quite slow
since it needs to read and write the entire content of all disks. I'd
therefore go as many spindles as you can - 3 disks, 5 disks and 9 disks
are what I recall as being optimal groups for RAID 5.
Also consider that you can supplement the RAIDs with the BSD tools
previously mentioned. Today is 3 x 300GB. Tomorrow add another 3 x 300
(assuming IOPS is OK) and concatenate them to be a 1.8TB "disk" - 2D+P +
2D+P.
Dave.
--
David Rawling
PD Consulting And Security
Mob: +61 412 135 513
Email: [email protected]
_______________________________________________
[email protected] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "[email protected]"
