Open Stack

> > This is what i have configured on my compute node. I have two NIC and
> > both nic connected to same cisco switch but [eth1] is trunk port.
> >
> > [eth0]   - Management  (vlan20)
> >
> > [eth1]   - (br-vlan bridge) - Data traffic (vlan 20,10)    - Trunk to cisco
> >
[Kaustubh] (I hope I got the VLANs correct). Try to imagine what will happen if the switch port connecting to eth0 gets a broadcast packet (an ARP request, for example). Because the other port is a trunk port, the switch will forward the broadcast packet out to eth1, which in turn will forward to br-int and eventually eth0 thereafter. See the loop? 

I believe, l2_population and arp_responder mechanisms were designed to alleviate this problem. You could enable them and see if this solves the problem.


-Kaustubh