Heartbeat + drbd

[morph]

Hello

I have a problem with heartbeat and drbd. I'm testing right now a
zimbra cluster.
My configuration is:
2x VMware mashines with CentOS system installed
on both:
- zimbra
- drbd
- heartbeat

Everything works fine, when I shutdown, reboot mashine or kill ha
proces on one node, second takes his work. After first wake up drbd
disks make sync and the zimbra system is all the time avalible. The
problem makes on simulated network error. When I unplug cable on node1
(primary) the second node2 takes his job, but if I restore network on
node1 everything makes bad. On both nodes are working services and
both are primary now, the disks are out of sync and cant see one
another.

How can I resolve this problem?

Sorry for ma english :)