[ovs-discuss] How to configure OVS to add multiple physical interfaces that are on different networks?

Pavel Popa pashinho1990 at gmail.com
Wed Jan 18 15:46:25 UTC 2017


Hi OVS community,

I successfully extended the OVS kernel module with a tunneling protocol in
a "native" manner, i.e. extending the flow key definition "struct
sw_flow_key" with a new flow key matching the new tunnel parameters. At
this point I would like to perform a performance test in an environment
looking as follows:

               Network A               Performance Testing Machine
(PTM)             Network B
[Machine A]------------------[(eth2)  VM with extended OVS kernel module
(eth4)]-----------------[Machine B]

Machine A sends packets encapsulated in the new tunneling protocol, call it
"new cool tunneling protocol" (NCTP), where the tunnel destination IP
address is that of the PTM's eth2.
Machine B sends normal user packets, non-NCTP encapsulated, destined to the
PTM with destination IP address of the eth4. The PTM should:

    capture the NCTP-encapsulated packets sent from Machine A to eth2,
decapsulate them, and forward to Machine B via eth4;
    capture the normal (not encapsulated) packets sent from Machine B to
eth4, NCTP-encapsulate them, and forward to Machine A via eth2.

The extended OVS kernel module knows how to encap into/decap from this new
tunneling protocol. The only problem I'm having is what should be the
correct configuration for OVS to forward: the encapsulated packets to
Machine A, and the decapsulated ones to Machine B.
I tried to add both the physical NICs (eth2 and eth4) into an OVS bridge
called "mybridge" with: "ovs-vsctl add-port mybridge eth2" and "ovs-vsctl
add-port mybridge eth4". However, this doesn't capture any packet received
from eth2 and neither from eth4. Any ideas for how to properly configure
OVS resulting in such environment?

Thank you
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openvswitch.org/pipermail/ovs-discuss/attachments/20170118/07a9faa4/attachment.html>


More information about the discuss mailing list