[Linux-cluster] Failover with multiple interfaces

[Lon Hohberger]

On Wed, 2007-11-07 at 14:19 +0000, James Fidell wrote:
> So, I have my NFS cluster all set up, with NFS services and shared
> storage all managed over a "private" network.
> 
> My intention is to join several more nodes to the cluster using the
> same storage, but providing services other than NFS.  Is it possible
> at the same time to provide services on a "public" network and have
> floating public IP addresses which are also migrated across the
> failover domain when the cluster fences a node?
> 
> Is it just a case of creating a second IP address service in the
> failover domain?  Will the resource manager just "do the right thing"
> and bind the floating address to the correct network interface?

Yes, it binds IPs to NICs based on netmask + network.

So...
  eth0 address/mask = 10.1.1.1/24
  eth1 address/mask = 192.168.1.1/16
  
  * All VIPs in the network 10.1.1.0/24 would get bound to eth0
  * All VIPs in the network 192.168.0.0/16 would get bound to eth1

It doesn't matter what the cluster's using for its internal
communications.

-- Lon