[fa.info-vax] clusters and V4

info-vax@ucbvax.ARPA (01/08/85)

From: Ken Adelman <KEN@JPL-VLSI.ARPA>

	We have just received V4.0 of VMS to run on 3 750s and 2 780s
connected via CIs. The two 780s and two 750s are "similar" machines,
and we would like at least the 750s to share queues and the like.
The V4.0 documentation implies that machines may or may not share queue
files, but says that a queue name must be unique on the cluster. We
would like the 780s to have theiur own queu file and the 750s their own. Is
this possible?
	Second problem: With just two of the 750s currently running V4.0, 
if one of them crashes, the other blocks process execution.  This
is somewhat less than desirable, and the documentation (Cluster management
guide or something like that) is somewhat less than clear about how to 
prevent this.
	I would appreciate hearing from anyone who has experience (yet!)
with clusters and these problems.

					-Ken Adelman, Caltech
KEN@JPL-VLSI.ARPA
KEN@HAMLET.BITNET
ucbvax!cithep!ken
CITHEX::KEN (hepnet)
------

info-vax@ucbvax.ARPA (01/08/85)

From: "Stephen Tihor" <TIHOR@NYU-CMCL1.ARPA>

	Second problem: With just two of the 750s currently running V4.0,
	if one of them crashes, the other blocks process execution.  This
	is somewhat less than desirable, 

There are two "approved" techniques of maintaining a cluster in a two processor
enviornment.  

	(A) add a quorum disk on a shared device (HSC or DUAL ported disk)

	(B) Make one node the "master" and the other the "slave" by 
		giving the master 2 votes and the slave 0.  Then at least
		the master can keep going.

Is that the sort of information you were asking for?

 \\   Stephen Tihor / CIMS / NYU / 251 Mercer Street  / New York, NY 10012  //
((  DEC Enet: RHEA::DECWRL::"""TIHOR@NYU-CMCL1.ARPA"""  NYUnet: TIHOR.CMCL1  ))
 // ARPAnet: Tihor@NYU-CMCL1   UUCPnet address: ...!ihnp4!cmcl2!cmcl1!tihor \\
 
-------