🚀 go-pugleaf

In article <1993Jan19.154520.20736@nessie.mcc.ac.uk>, mzagsaa@v4.cgu.mcc.ac.uk (Tony Arnold) writes:
> 
> I run a LAVC where two of the nodes are a 3600 and a 3400. These nodes each
> have a CMD CQD223TM SCSI adapter on them and there is a 1Gb disk on each
> SCSI bus.
> 
> The CMD controllers support multi-hosting, so it is possible to effectively
> connect both controllers on the same SCSI bus. The disks would then be
> accessible from both nodes which would reduce the impact of one or other
> machines going down.
> 
> One of the disks is currently the system disk for the cluster. My question,
> for anyone who may have done this, is it possible to boot both nodes direct
> from the disk (using separate roots, of course) or do I still have to boot
> one from the disk and the other remotely via a MOP request?

Both systems can boot directly as long as you can set BFLAGS on one node to
select an alternate root.  The hardware you have may or may not support this.

Now all sounds peachy, but there are several significant caveats:

	- SCSI cable length is an issue be sure to keep well within the 5
	  meters, and don't try to fill up the SCSI bus with 6 devices.
	- SCSI termination is an issue which implies that even if one node 
	  goes down you can't simply turn it off without removing power
	  on the terminating resistor on one end of the bus.  Doing this will
	  most likely hang the scsi bus or at least make data corruption VERY 
	  likely.
	- If you had more powerful CPU's and had more disks on the bus you'd
	  also have to worry about controller/qbus bandwidth being a serious 
	  limitation (especially if you try to do shadowing).

> Tony Arnold,
> Manchester Computing Centre,
> OpenVMS Systems Manager.
-- 
--
Mark Pizzolato - INFO COMM Computer Consulting, Redwood City, Ca
PHONE:	(415)369-9366	UUCP:  decwrl!infopiz!mark or uunet!lupine!infopiz!mark
DOMAIN:	mark@infocomm.com