[Info-vax] Looking to remote cluster my two OVMS machines. Questions. Need help!!!

[Hans Vlems]

On 11 okt, 13:15, Paul Sture <nos... at sture.ch> wrote:
> In article
> <f1ac15f3c2b4c4325eaebe4b30fb0... at msgid.frell.theremailer.net>,
>  Fritz Wuehler <fr... at spamexpire-201210.rodent.frell.theremailer.net>
>
>  wrote:
> > Rob Brown <br... at libra.gmcl.internal> wrote:
>
> > > On 2012-10-10, Len Whitwer <l... at psds.com> wrote:
> > > >     The server is not very busy at all however I do need
> > > >     continuious availability if main server goes down.
>
> > Can't you run a few SIMH instances on cheap WinTel boxes as nodes? Not that
> > they don't go down...but the more the merrier.
>
> That reminds me of a customer with 2 Alphas in buildings a mile apart.
> What we really wanted was a third voting member which could see both,
> preferably in a third location, so that if either building lost power,
> the other Alpha would carry on.
>
> DEC suggested that a lowly VAX 2000 would be up to the task.  Nowadays a
> SimH instance would do that fine.
>
> --
> Paul Sture


A solution that works fine for me is to give each node in a two node
NI cluster its own quorum disk.
The quorum disk is local to that node. Each disk and each host have
one vote. If both nodes are running
then the quorumdisks do not participate in the voting process. As soon
as one node leaves the luster the
remaining node discovers its quorum disk, adds one vote and keeps on
running.
This is not a supported configuration though. An NI might loose its
network connection, especially on large LANs.
If that happens then both nodes will decide that the other one crashed
and recover as described.
Unfortunately when the LAN comes on-line then the hosts will, or may
not, recognize each other. At this "rediscovery"stage
the cluster status is undefined and all kinds of things (especially
datastructures) may fail.
For simple applications that just share files this set up works fine.
Hans