Cluster storage config questions...

From: ZoombyWoof (zoombywoofremove_at_thishotmail.com)
Date: 01/17/05 

Date: Mon, 17 Jan 2005 17:53:21 +0100

Hi all. We are trying to setup a RedHat Enterprise (V3) Cluster and use
it for high availability, having database applications failing over and
so forth.
We have 2 servers with FibreChannel interface connected to a NStor
diskarray via a FibreChannel switch. We have made 8 Logical Units (LUNS)
in the array, and so far made 3 of them visible to node A, LUN0, LUN2
and LUN6.
The rh-cluster guide says that by default a linux machine can only see
LUN0. You have to enter "options scsi_mod max_scsi_luns 32" for example,
in /etc/modules.conf. We have done that, but node A only sees 1 disk,
LUN0, which is being mapped to /dev/sda. I can make filesystems on it so
it really works :-) but why cant we see any more LUNS ? If we boot with
DOS diskette with drivers for the FC card, DOS can see the available
LUNS, but not Linux...anyone else had this problem ? Or am I overseeing
something obvious here ?

The second issue is about what device to use as bootdevice. We have
discussed this a lot, and we have 2 opinions on this matter which we
can't aggree on 100%. One opinion is to have local disks in the nodes
and install Linux on them, and Only use the FC connected diskarray för
quorum-partitions and shared data. The other opinion is to, on each
node, install Linux on a diskarray LUN (one Boot-LUN dedicated to each
node)
Anyone here has an opinion on this ? What is usually the best ? Or is it
merely a matter of taste ?

Any advice or opinion on this is very much appreciated.

Thanx

/ZW