UCS blade adding linux bonded sub interface to external bridge problems

asked 2015-04-15 10:11:40 -0600

visualne gravatar image

updated 2015-04-15 10:13:38 -0600


I recently have installed a three node non-ha icehouse cluster on a UCS chassis. I am using the openvswitch plugin with a vxlan overlay. The installation went fine. However, I am seeing some strange behavior once the bonded sub interface is added to the external bridge on the network node. Connectivity appears to be intermittent... In other words connectivity to the ip associated to br-ex is intermittent. And of course connectivity to the floating ips is intermittent as well.

Not quite sure why this is.... I have seen mention in this link https://answers.launchpad.net/openstack-cisco/+question/245039 (https://answers.launchpad.net/opensta...) , response number two that the person has seen intermittent problems with adding a traditional linux bond to an external bridge.

The way the the NICs are presented to the network node is via a traditional port-channel on two physical interfaces that is trunking four vlans. To accommodate this setup, I setup a traditional linux bond and then created sub interfaces corresponding to each vlan. ex) bond0.200, bond0.300, bond0.400.

I then added bond0.200 to the external bridge because that is where our floating ips will be hosted. When I do this I am seeing connectivity flakiness inbound to the vms and outbound from the vms.

misc The NIC driver is: 'Ethernet controller: Cisco Systems Inc VIC Ethernet NIC'

Topology - http://pastebin.com/VVaqjfnU

^ -- The above e0 and e1 ports are on the network node that is hosted on a UCS blade.

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2015-04-15 22:28:05 -0600

dodi gravatar image

Have you tried using the OVS Bond instead of the traditional Linux bond?

http://blog.scottlowe.org/2012/10/19/link-aggregation-and-lacp-with-open-vswitch/ (http://blog.scottlowe.org/2012/10/19/...)

edit flag offensive delete link more


Yes. However, I have had trouble with it. I have created "fake" interfaces in the same bridge that the bond was created in. And then put those interfaces in the corresponding vlans on the trunk. However, I am unable to achieve connectivity to of these "fake" interfaces.

visualne gravatar imagevisualne ( 2015-04-20 07:25:16 -0600 )edit

I can't see arp traffic going over the bond. But it's never being forwarded to ports in the same vlan on the bridge.

visualne gravatar imagevisualne ( 2015-04-20 07:26:11 -0600 )edit

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2015-04-15 10:11:40 -0600

Seen: 788 times

Last updated: Apr 15 '15