Ben:
Thank you for your quick reponse, now I can communicate 2 VMs with GRE
tunnel using just one bridge!
The original purpose for connecting two bridges is that I want to develop
QINQ in ovs, and I thought the QINQ topology should be like this:
+-------+
| VM |
+--------+
|
| p0(access)
+---------------------+
| br0 |
+---------------------+
|p1(trunk)
|
|p2(qinq)
+---------------------+
| br1 |------eth0-------->to public internet
+----------------------+
p0: br0, access, tag=100
p1: br0, trunk, trunks=100,200
p2:br1, qinq, tag=2
eth0:br1, trunk, trunks=2,3
send:
packet from p1, add tag 100; then packet come into p2, add tag 2;
recv:
packet from eth0, with outer tag 2 and inner tag 100; packet out from p2,
strip tag 2; packet out from p1, strip tag 100;
so now two bridges connected via port p1 and p2, that's similar to
traditional network
1. under this condition, do I still need patch port to connect two bridges?
2. for developing QINQ in ovs, does this topology seems rational?
At 2014-09-29 11:11:28, "Ben Pfaff" <b...@nicira.com> wrote:
>On Mon, Sep 29, 2014 at 09:39:21AM +0800, ychen wrote:
>> Thanks very much, the information really helps.
>> What I am confused is about GRE tunnel
>> http://blog.scottlowe.org/2013/05/07/using-gre-tunnels-with-open-vswitch/
>> In the above blog, each host uses two bridges, one for GRE, one for
>> physical port
>> but there seems no patch port between 2 bridges.
>
>I think that the presence of two bridges in this configuration is a red
>herring. The way that packets make it from the GRE tunnel into the
>second bridge in this scenario is the same way that packets make it from
>a GRE tunnel into a bridge when there is only one bridge. If you are
>setting up a GRE tunnel with Open vSwitch for the first time, then it is
>probably easiest to just use one bridge (and keep the physical device
>off that bridge).
_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss
