Hello World,

goyald@gmail.com wrote:
> Hello,

> I intend to create a configuratin where host A and host B host two
> independent applications. I want hort A to failover to host C,
> and I want host B to also failover to host C.
> basically
> A <-> C
> and
> B <-> C
> Is it possible using HACMP or using Veritas clustering?

I don't know enough about Veritas clustering to comment on that, but it
can be done with HACMP. You create one cluster consisting of the three
nodes, A, B and C. You create two resource groups, one of which
contains nodes A and C, the other contains nodes B and C.

Another thing you may want to consider is to create a mutual takeover
cluster. Eliminate C, but have machine A take over application B and
vice versa. One machine less to keep in sync, and in the case of LPARs,
more resources per machine.

I've found that a machine that sits there quietly waiting for one of
the two others to fail has a certain attraction for types who want to
misuse it as a test/development server, which leads to all kinds of
complications.

> If necessary I can use three LPARs instead of physical machines.

That is possible, of course, but it leaves a Single Point Of Failure
i.e. the machine, unless of course the LPARs are on different machines.

Cheers,
Menno