RaZ wrote:
>
> Hi there,
>
> I was wondering if anyone can help me with a little configuration problem. I
> have a setup with two nodes and four service groups. In normal production
> there will be two services running on each node. Each node of my system has
> got two ge interfaces. My plan was to make a MILTINIC type in one of my
> service groups, and then point to the MULTINIC via a proxy from the other
> group. But what will happen when you do a fail-over of the group with the
> MULTINIC????? How do you make the best and most flexible solution on this
> setup?
>
> Hope somebody has got some input J
>
> Kindest regards
>
> RaZ


The MultiNICA resource is of type "none" this means that VCS will only
monitor the resource. Since VCS does not do an online or offline of the
MultiNICA it will not be effected by a fail-over. The configuration
your describing is a common one since having two MultiNICA resources
monitoring the same set of interfaces does not work. Since the proxy
agent will replicate the state of another resource on a per node basis
it will cause a failover of what ever group it's in on that system. For
example, node A and node B share service groups alpha and beta, group
alpha has a MultiNICA resource that controls the two public interfaces
for nodes A and B. In group beta a proxy has been setup to watch the
MultiNICA resource in group alpha. If all the network public network
cables are pulled on node B then the MultiNICA resource in group alpha
will fault as will the proxy resource in group beta. The MultiNICA and
proxy resources will still report as online on node A for both groups
alpha and beta. When the network connections are restored on node B the
MultiNICA resource will return to online automatically as will the proxy
resource for groups alpha and beta on node B.

This interaction is the second most complicated behavior in VCS

Bill Borsari

** Any opinions expressed are my own and not of my employer **

Hi William,

I was just wondering, what is the most complicated behavior in VCS then??
:-)

Kindest regards

RaZ
"William Borsari" wrote in message
news:3B0AC9BF.6508EAC3@veritas.com...
> RaZ wrote:
> >
> > Hi there,
> >
> > I was wondering if anyone can help me with a little configuration
problem. I
> > have a setup with two nodes and four service groups. In normal
production
> > there will be two services running on each node. Each node of my system
has
> > got two ge interfaces. My plan was to make a MILTINIC type in one of my
> > service groups, and then point to the MULTINIC via a proxy from the
other
> > group. But what will happen when you do a fail-over of the group with
the
> > MULTINIC????? How do you make the best and most flexible solution on
this
> > setup?
> >
> > Hope somebody has got some input J
> >
> > Kindest regards
> >
> > RaZ
>
>
> The MultiNICA resource is of type "none" this means that VCS will only
> monitor the resource. Since VCS does not do an online or offline of the
> MultiNICA it will not be effected by a fail-over. The configuration
> your describing is a common one since having two MultiNICA resources
> monitoring the same set of interfaces does not work. Since the proxy
> agent will replicate the state of another resource on a per node basis
> it will cause a failover of what ever group it's in on that system. For
> example, node A and node B share service groups alpha and beta, group
> alpha has a MultiNICA resource that controls the two public interfaces
> for nodes A and B. In group beta a proxy has been setup to watch the
> MultiNICA resource in group alpha. If all the network public network
> cables are pulled on node B then the MultiNICA resource in group alpha
> will fault as will the proxy resource in group beta. The MultiNICA and
> proxy resources will still report as online on node A for both groups
> alpha and beta. When the network connections are restored on node B the
> MultiNICA resource will return to online automatically as will the proxy
> resource for groups alpha and beta on node B.
>
> This interaction is the second most complicated behavior in VCS
>
> Bill Borsari
>
> ** Any opinions expressed are my own and not of my employer **

Service group dependanies.

Bill Borsari

** Any opinions expressed are my own and not of my employer **

RaZ wrote:
>
> Hi William,
>
> I was just wondering, what is the most complicated behavior in VCS then??
> :-)
>
> Kindest regards
>
> RaZ
> "William Borsari" wrote in message
> news:3B0AC9BF.6508EAC3@veritas.com...
> > RaZ wrote:
> > >
> > > Hi there,
> > >
> > > I was wondering if anyone can help me with a little configuration
> problem. I
> > > have a setup with two nodes and four service groups. In normal
> production
> > > there will be two services running on each node. Each node of my system
> has
> > > got two ge interfaces. My plan was to make a MILTINIC type in one of my
> > > service groups, and then point to the MULTINIC via a proxy from the
> other
> > > group. But what will happen when you do a fail-over of the group with
> the
> > > MULTINIC????? How do you make the best and most flexible solution on
> this
> > > setup?
> > >
> > > Hope somebody has got some input J
> > >
> > > Kindest regards
> > >
> > > RaZ
> >
> >
> > The MultiNICA resource is of type "none" this means that VCS will only
> > monitor the resource. Since VCS does not do an online or offline of the
> > MultiNICA it will not be effected by a fail-over. The configuration
> > your describing is a common one since having two MultiNICA resources
> > monitoring the same set of interfaces does not work. Since the proxy
> > agent will replicate the state of another resource on a per node basis
> > it will cause a failover of what ever group it's in on that system. For
> > example, node A and node B share service groups alpha and beta, group
> > alpha has a MultiNICA resource that controls the two public interfaces
> > for nodes A and B. In group beta a proxy has been setup to watch the
> > MultiNICA resource in group alpha. If all the network public network
> > cables are pulled on node B then the MultiNICA resource in group alpha
> > will fault as will the proxy resource in group beta. The MultiNICA and
> > proxy resources will still report as online on node A for both groups
> > alpha and beta. When the network connections are restored on node B the
> > MultiNICA resource will return to online automatically as will the proxy
> > resource for groups alpha and beta on node B.
> >
> > This interaction is the second most complicated behavior in VCS
> >
> > Bill Borsari
> >
> > ** Any opinions expressed are my own and not of my employer **

One trick I have found when working with multiNic is to put the
mulit-nic in its own service group, and throw a phantom resource
on top of it. (Note, the IPMultiNic resource should not be in
this group, it should be with the real service groups of your
applications)
This way

1: All your real service groups use a proxy, so no one service
group really owns the multi-nic resouce.
2: For the truly sadistic, you can use a service group dependancy
to make all your other service groups dependant on the multi-nic
service group. You can then test failover of multiple groups
by downing the multinic service group on a particular node.
Any groups on that node should fail over. (I thought about doing
this myself, but never actually did it. As the man said, service
group depeancies are waaaaay too complex.)

You really have to do this for a round-robin style of failover,
as you cant guarentee that each node will be a member of each
service group. Also keeps things sinple, all the nic resouces
in one place.

-Andrew


"RaZ" wrote:
>
>Hi there,
>
>
>
>I was wondering if anyone can help me with a little configuration problem.
I
>have a setup with two nodes and four service groups. In normal production
>there will be two services running on each node. Each node of my system
has
>got two ge interfaces. My plan was to make a MILTINIC type in one of my
>service groups, and then point to the MULTINIC via a proxy from the other
>group. But what will happen when you do a fail-over of the group with the
>MULTINIC????? How do you make the best and most flexible solution on this
>setup?
>
>
>
>Hope somebody has got some input J
>
>
>
>Kindest regards
>
>
>
>RaZ
>
>