Hi, Our topology consists of 2 VM on one host machine running KVM, every VM has 2 interfaces.
An openvswitch instance running on the host, which has 2 logical separate bridges, one for management and the other for internal usage between VM's. We wanted, through one of the bridges, to use it for management of all VM's as well as the host. So a trunk (VLAN 100) arrives at the host and gets to the first bridge in openvswitch. Then after, we configure an internal port (tagged 100) for the host, and 2 additional ports, each one connected to each VM and also tagged 100. Until here it is all ok, connectivity works just fine, and from vlan 100 we can get to all machines (be them VM's or the host, for management purposes). Now we wanted to have separate completly internal networks which aim to generate internal trafic between the VMs. We configured an additional bridge connected to the VM's as well in another 2 interfaces (1 for each VM) to create another network just for those 2 VM's (this connection is separated VLAN tagged 150). Problem: The very same moment one of the machines send an ARP request in the internal network, the ARP gets replicated even on the other network (VLAN 100) to the point the very same host hungs up having 2 VM's running at 100%. In the beggining, we have thought vlan's would solve the problem and help segment the two networks and avoid the arp storm. Then we tried to make the hosts not to announce or answer back ARP requests that were not directly related to their interface. Nothing has worked. We tried this very same thing setting a controller properly installed, but still doesen't explain why without controller the ARP storm gets across the the 2 VLANS. Any help on the case would be much appreciated. Thanks in advance.
_______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss
