[BlueOnyx:07116] Re: raid array

[Michael Stauber]

Hi Ken,

> Does anyone know how large a drive the BlueOnyx and Aventurine could read?
> 4TB ?

Chris pretty much answered the question already and I may not have much new 
info to add.

The boot manager GRUB has problems booting off disks larger than 2TB of size. 
So that's currently the limiting factor for BlueOnyx and Aventurin{e}, but 
also for other distributions that use GRUB.

Typically when you use multiple really large disks, you're often better off 
using a good dedicated RAID controller in conjunction with them, as they 
usually offer a better performance than using any onboard controller.

Use the dedicated controller to set up two (or more) RAID arrays. Make the 
first array smaller than 2TB and install just the OS on it. For just the OS a 
couple of gigabytes usually make do, but if you got plenty of space available, 
I wouldn't be greedy there and would probably give it 50 or even 100 Gig. Just 
to be flexible. 

Then use the remaining (larger) array for storing the data. 

When installing BlueOnyx or Aventurin{e}, you can type "self" (or "linux 
self") at the CD boot prompt to launch into the diskdruid wizard that allows 
you to define your own custom partitioning scheme. 

There is only one gotcha there that you must keep in mind: Disable all but the 
primary array when you install the OS. Otherwise the BlueOnyx or Aventurin{e} 
installer will see the various volumes that the (hardware) RAID provides as 
individual disks and will install a software RAID on top of the hardware RAID. 
;o)

Simply turn the offlined arrays back on after the OS install and then format 
and mount them manually.

If you really add plenty of diskspace to a server, then I'd even advise to 
create multiple separate RAID arrays for more redundancy and flexibility, but 
also for faster recovery from the usual problems. Think of RAID resyncs (which 
take ages on really large arrays), or the dreaded integrity checks when a 
partition is mounted. Having FSCK run on startup on a 4TB volume after a 
server crash is no fun and will take all night and then some. 

Unfortunately CentOS doesn't really support the XFS filesystem that well out 
of the box, as it would be really ideal for excessively large volumes. RedHat 
just doesn't want us to know that IBM did get something right when they 
created XFS. ;o)

-- 
With best regards

Michael Stauber