[SunHELP] How Do I......

[Phil Stracchino]

On Tue, Jun 03, 2003 at 02:18:54PM -0400, Michael A. Turner wrote:
> 	I have a new sun server as some may know. I am about ready to give
> up on Raid. $2K was just not budgetted for. So here is what I want to do. I
> want to make a file system that spans the 4 disks in my D2 array. This array
> consists of 4 36 GB drives. I would like to mount a /newdir on the system
> where /newdir will span all four disks making the volume size of /newdir be
> around 144GB. Can I do this? what is the command/tool for it? If not I will
> have to mount the four disks seperatly.	


OK.  There are several questions you need to ask yourself about this.
Which is most important to you -- space, speed, or redundancy? Can you
spare any of that 144GB? Is cost the only thing stopping you from using
RAID?

Once you've thought about those, you're now ready to remember that
Solaris ships with very good software RAID tools (there's even a
graphical manager for them, if you're willing to endure the horror that 
is CDE), and ask yourself whether you want to:
   (A) just join the disks end-to-end into one big logical disk?
   (B) distribute data across all four disks in parallel?
or (C) set up RAID5 across the four disks?

Once you've chosen one, it's time to talk about how to do it.  (For all
of these examples, I'm going to assume your disks are targets 1,2,3 and
4 on controller 1, each partitioned as a single 36GB slice.  I'm also
going to arbitrarily pick metadevice d10, though you can pick anything
from d0 to d127.)


(A) is called a concat, and it's generally not a good idea because it's
an inferior solution to all the other options.  It has the vulnerability
to failure of striping, without the speed.  If you wanted to do it,
you'd do it like this:

metainit d10 4 1 c1t1d0s0 1 c1t2d0s0 1 c1t3d0s0 1 c1t4d0s0

You would then have a metadevice, /dev/md/dsk/d10, which you can newfs
and mount.


(B) is called a stripe.  It's fast because all four disks are sharing
reads and writes, but like a concat, it has no redundancy.  Failure of
any single disk will result in loss of the entire metadevice.  If you
wanted to do it this way, you'd do it like this:

metainit d10 1 4 c1t1d0s0 c1t2d0s0 c1t3d0s0 c1t4d0s0 -i 128k


(C) is not as fast as a plain stripe due to the overhead of the parity
calculations, but gives you the benefit of redundancy.  You can survive
the failure of any single disk without loss of data or performance, but
one full disk's worth of space* (out of however many disks you use) is
used for parity, so your working data set would be 108GB instead of
144GB.  If you wanted to do it this way, you would do this:

metainit d10 -r c1t1d0s0 c1t2d0s0 c1t3d0s0 c1t4d0s0 -i 128k

* A single disk's worth of space, but not a single disk -- parity data
  is distributed across all of the disks.


Before you can do any of these, you first need to create metadbs on your
boot devices.  I suggest starting out by reading the man pages for
metadb, metainit, and everything they refer you to.



-- 
 .*********  Fight Back!  It may not be just YOUR life at risk.  *********.
 : phil stracchino : unix ronin : renaissance man : mystic zen biker geek :
 :  alaric at caerllewys.net : alaric-ruthven at earthlink.net : phil at latt.net  :
 :   2000 CBR929RR, 1991 VFR750F3 (foully murdered), 1986 VF500F (sold)   :
 :    Linux Now!   ...Because friends don't let friends use Microsoft.    :