Heartbeat issue: IP never fails over

Hi all,

Have Heartbeat configured using the Linode HA guide for web applications, and it's working like a charm (after ironing out some kinks) except for one problem: IP failover.

When I do a service stop heartbeat on ha1 to test that everything is working, the second ha2 node never picks up the IP address. So even though all the services are running, to the outside world it's as though the server is simply down.

So crm_mon reports

ip1        (ocf::heartbeat:IPaddr2):       Started ha2
ip1arp     (ocf::heartbeat:SendArp):       Started ha2

but the server simply doesn't respond if I try to reach the floating IP address.

AFAIK I have IP failover configured in Linode Manager according to the appropriate guide.

What would cause this? Any ideas? I'm happy to post logs but I thought I would start to see if I'm doing something wrong. Will testing IP failover only work correctly if I actually shut down the server?

Thanks!