Re: 2-node LAVC cluster with quorum disk - network disappears - which node CLUEXITs ?

From: Daryl Jones (jones.computer.srv_at_worldnet.att.net)
Date: 02/28/04 

Date: 28 Feb 2004 01:00:27 -0800

Dear Roy Omond:

I am sorry for not clearly reading your question. Anyway, here is
another try.

According to VAXcluster Principle, Preliminary Edition - Fall 1992 us
DECUS, page 7-11:

Since the cluster system manager has given them equal number of votes,
they are considered to be of "equal importance". Thus, it is more or
less random which node is forced out of the cluster. Whichever node
first proposes a new configuration will remain in the cluster, and the
other node will BUGCHECK. If you want a one node to be favored over
the other then change the vote on the favored to be greater than the
other node. Generally by one vote.

>From page 7-12:

The Connection Manager "logically" considers all possible connected
subcluster. For each of these subclusters it computes a "figure of
merit" using the formula

                256+V+N="fiqure of merit"

Where V is the total number of votes in the subcluster, and N is the
total of VAX systems in the subcluster.

The Connection Manager then picks as the "optimal subcluster" the
totally connected subcluster with the highest "figure of merits", and
discards the other possiblities.

>From Page 7-13:

"When multiple subclusters qualify for being optimal, one is
effectively chosen at random, and the others are discarded. In such
situations, the optimal subcluster will generally be the subcluster
containing the system that first noticed the problem."

I hope this helps.

Regards:
Daryl Jones

Roy Omond <Roy.Omond@BlueBubble.UK.Com> wrote in message news:<c1kefd$1jqibc$1@ID-225674.news.uni-berlin.de>...
> Daryl Jones wrote:
>
> > Years ago I ran across a problem where the cluster information was
> > going over the ethernet and not over the CI connection. Therefore,
> > when the ethernet connection was lost there went the cluster until the
> > traffic restablished over the CI. Check to see if the cluster
> > connection was lost over the SCSI or ethernet. There is a sysgen
> > parameter that controls which path is used. I don't recall the sysgen
> > parameter because it was over 30 years ago. Sorry about that.
>
> Cluster communications do *not* go over SCSI (an all too common
> misapprehension).
>
> As stated before, the single and only path for SCS traffic (namely
> the Ethernet) was disconnected.
>
> Roy Omond
> Blue Bubble Ltd.

Relevant Pages

  • Re: 2-node LAVC cluster with quorum disk - network disappears - which node CLUEXITs ?
    ... > going over the ethernet and not over the CI connection. ... Check to see if the cluster ... > connection was lost over the SCSI or ethernet. ...
    (comp.os.vms)
  • Re: 2-node LAVC cluster with quorum disk - network disappears - which node CLUEXITs ?
    ... Years ago I ran across a problem where the cluster information was ... going over the ethernet and not over the CI connection. ... > Is it the last one to access the quorum disk that stays up? ...
    (comp.os.vms)
  • Re: What is the Difference between Shadow and Mirrored disk?
    ... Would it not have to have VMS + Cluster licenses? ... the external controller is the quorum disk. ... If either node loses SCSI or Ethernet, ... If a node loses it's SCSI connection to the HSG, ...
    (comp.os.vms)
  • Re: Clustering w/ Firewire
    ... connection for a cluster if it is faster than normal ... ethernet and it uses tcp/ip protocol. ...
    (microsoft.public.windows.server.clustering)
  • Re: Clustering: switches reliability/redundancy
    ... If NIC A-1 fails, everything still works. ... Each node has a list of legal subclusters, and the general principle is that the transaction coordinator in the cluster, finds agreement on a "survivable" "largest" subcluster with quorum, and your unlucky nodewill be CLUEXITed. ... The general idea is that the list comprises all members, then fewer and fewer members, and at each coordinating step if there is a match of those few members that can maintain quorum, then reconfiguration completes. ...
    (comp.os.vms)