Booting on RAID set failing

[pll at lanminds.com]

In a message dated: Wed, 18 Sep 2002 17:04:09 EDT
bscott at ntisys.com said:

>  Wouldn't this
>
>	4 disks --> RAID 0 --> NBD --> AN --> RAID 0, 1, 5, or 0+1
>
>be more flexible, and use available storage more efficiently, too?

That's what I'm planning on doing, except for the first part:

	4 disks -> NBD -> AN -> RAID5 & RAID1

Pictorially (as I mentioned to you privately some time ago), this is 
what I'm attempting (SN=Storage Node, AN=Access Node):


	 ---------------
SN1	| 1 | 2 | 3 | 4 |
	 ---------------
	 ---------------
SN2	| 1 | 2 | 3 | 4 |
	 ---------------
	 ---------------
SN3	| 1 | 2 | 3 | 4 |
	 ---------------
	 ---------------
SN4	| 1 | 2 | 3 | 4 |
	 ---------------
	 ---------------
SN5	| 1 | 2 | 3 | 4 |
	 --------------- 	

Where each number represents a RAID 5 stripe across nodes; or a RAIN 
set[1].

So, each set could be a RAID5 stripe.  And the neat thing is, that if I 
had twice the number of systems, I could then mirror those RAID5 
stripes across a different set of nodes using RAID11, and I could 
managae all this with something like LVM[2] on the AN.


	 ---------------	   --------------- 
	| 1 | 2 | 3 | 4 |	  | 1'| 2'| 3'| 4'|
	 ---------------	   --------------- 
	 ---------------	   --------------- 
	| 1 | 2 | 3 | 4 |	  | 1'| 2'| 3'| 4'|
	 ---------------	   --------------- 
	 ---------------	   --------------- 
	| 1 | 2 | 3 | 4 |	  | 1'| 2'| 3'| 4'|
	 ---------------	   --------------- 
	 ---------------	   --------------- 
	| 1 | 2 | 3 | 4 |	  | 1'| 2'| 3'| 4'|
	 ---------------	   --------------- 
	 ---------------	   --------------- 
	| 1 | 2 | 3 | 4 |	  | 1'| 2'| 3'| 4'|
	 --------------- 	   --------------- 

[1]  Ahhh, just what we need, another marketing buzz word !
     RAIN  -	 Redundant Array of In(expensive,dependant) Nodes

     Note, I don't think this is original, I've heard it before,
     I just don't know if it's actually a real thing yet.

[2] So, now my single point of failure is the Access node.
    This is where insanity *really* gets to be fun.
    How about using the MCL clustering product (or something similar)
    to cluster the access nodes in a fault-tolerant, HA configuration :)


-- 

Seeya,
Paul
--
	It may look like I'm just sitting here doing nothing,
   but I'm really actively waiting for all my problems to go away.

	 If you're not having fun, you're not doing it right!