Hi Nathan,

If you just want eth1 to have the traffic between nodes, you just need to set CONFIG_NEUTRON_OVS_TUNNEL_IF=eth1 and leave the bridge configuration to use eth0. The bridge will then be used to provide external network access to your instances, while the vxlan tunnel will be using eth1.

Also, I think reconfiguring the bridge mappings after a first install is not well tested, so that might have caused the error you found.

Regards,
Javier



I've been experimenting with a particular configuration from a commercial course on OpenStack, but I've hit an error in packstack pike.

Essentially the idea is to have an eth0 as an external interface out to the wider internet, and eth1 for all the internal traffic between nodes. Both eth0 and eth1 have been given valid IP addresses(eth0: 10.0.2.9, eth1: 192.168.2.9), and the hostname of the host has been put into /etc/hosts with the IP address of the external interface.

I tested my answer file to see that it worked properly as an all-in-one - which it did, but when I make these OVS changes, packstack doesn't finish up properly

CONFIG_NEUTRON_OVS_BRIDGE_MAPPINGS=physnet1:br-eth1
CONFIG_NEUTRON_OVS_BRIDGE_IFACES=br-eth1:eth1
CONFIG_NEUTRON_OVS_TUNNEL_IF=eth1

The error I see happens during the compute.pp part of applying the puppet manifests.

ERROR : Error appeared during Puppet run: 10.0.2.9_compute.pp
Error: Evaluation Error: Error while evaluating a Function Call, no address for  at /var/tmp/packstack/<code>/modules/packstack/manifests/neutron/ovs_agent.pp:50:27 on node rdotest.gateway

I've tried to dig through the trace log, but no smoking gun is showing up. 
I can post log files and the answer file if needed.

Thanks,


_______________________________________________
users mailing list
users@lists.rdoproject.org
http://lists.rdoproject.org/mailman/listinfo/users

To unsubscribe: users-unsubscribe@lists.rdoproject.org