Neutron/openvswitch not working, how to debug

asked 2014-01-20 05:41:08 -0500

anonymous user

Anonymous

updated 2014-01-22 15:12:38 -0500

anonymous user

Anonymous

Hello, I'm trying to get Neutron to work, but so far, without any luck. Basically I have a test setup with two compute nodes, each hosting one instance. When I log in to one instance, I can't ping the other.

On both nodes, running "ovs-vsctl show" gives something like this:

d5881c89-79c7-47a7-a18f-89d53999071d
    Bridge "br0"
        Port "phy-br0"
            Interface "phy-br0"
        Port "eth0"
            Interface "eth0"
        Port "br0"
            Interface "br0"
                type: internal
    Bridge br-int
        Port "int-br0"
            Interface "int-br0"
        Port br-int
            Interface br-int
                type: internal
        Port "tap87c80479-42"
            tag: 1
            Interface "tap87c80479-42"
    ovs_version: "1.11.0"

I've tried to trace the network traffic using tcpdump. What I've found is that the ARP requests made by the instance that is attempting to ping do pass through the tapXXXX interface, the int-br0 interface, the phy-br0 interface and then out onto the network via the eth0 interface.

On the other machine, I see the ARP requests coming in via eth0 and br0, but phy-br0 shows now traffic at all. So I think this is the point where it breaks. I just don't have a clue why, or how to further analyze this.

Any pointers would be highly appreciated.

(I'm using the latest Havana packages on top of RHEL 6.4.)

Thanks

edit retag flag offensive close delete

Comments

There should be some pointers on the site: use the search and see if any of the suggestions thrown around help you. Keep updating your question adding more details as you dive deeper.

smaffulli ( 2014-01-24 18:43:27 -0500 )edit

Are you using VLANs or GRE?

fetahi ( 2014-01-25 08:27:09 -0500 )edit