Linux heartbeat losing packets

With Pacemaker you don't have 'an active node' and 'a passive node'. All cluster nodes (there may be more than 2) can run services equally and the rules in the CIB database tell CRM (Pacemaker's resources manager) what node can run what services.

If you configured the service to run as a single instance that can run on both nodes with no constraints, then you cannot tell which node will run it. If you have two such services then you may end with one running on one node and the other on the other node. For the first service node1 will be active and for the second – node2.

You define actual preference by declaring some constants. Example: 'run service 1 on a node where service 2' is running or 'always prefer node 1 for both services'.

Usually you have a service which defines 'a logical master' – it can be an IP address or DRBD volume in the primary state – then all other services depend on this and you choose the 'master' by setting preferences for the primary service.

See the Pacemaker documentation for details on setting the constraints.

Heartbeat itself, when used with Pacemaker, doesn't make any decisions on master/slave states or on what resources are running.

In haresources, try specifying the interface you want to have the shared address:

loadb1  192.168.1.100/24/eth0 httpd

or /eth1 as appropriate (the /24/ specifies the netmask, change that if appropriate). If you do that, restart both HA processes, and it's still not working, can we get the output of ifconfig -a on both servers?

Edit: ah, that's probably not such a good idea. This page says that

Note that it is absolutely mandatory for the haresources files to be identical on both machines.

so you mustn't say /eth0 on one machine, and /eth1 on the other. May we ask why one machine uses eth0 and the other eth1, and if there's any way of changing it? And can we get the output of ifconfig -a, please?