[ovs-discuss] Issue when using local_ip with VXLAN tunnels in OVS

Siva Teja ARETI siva_teja.areti at nuagenetworks.net
Wed Nov 28 23:15:48 UTC 2018

Hi Greg,

Please find the answers inline below.

On Tue, Nov 27, 2018 at 1:35 PM Gregory Rose <gvrose8192 at gmail.com> wrote:

> Siva,
> You have a routing issue.
> See interalia
> https://github.com/OpenNebula/one/issues/2161
> http://wwwaem.brocade.com/content/html/en/brocade-validated-design/brocade-vcs-fabric-ip-storage-bvd/GUID-CB5BFC4D-B2BE-4E9C-BA91-7E7E9BD35FCC.html
> http://blog.arunsriraman.com/2017/02/how-to-setting-up-gre-or-vxlan-tunnel.html
> For this to work you must be able to ping from the local IP to the remote
> IP *through* the remote IP address.As we have seen that doesn't work.

Did you mean to be able to ping using remote interface? I am able to get
this to work when I connect the two bridges using a veth pair.

[root at vm1 ~]# ping -I eth2
PING ( from eth2: 56(84) bytes of data.
64 bytes from icmp_seq=1 ttl=64 time=0.655 ms
64 bytes from icmp_seq=2 ttl=64 time=0.574 ms
64 bytes from icmp_seq=3 ttl=64 time=0.600 ms
64 bytes from icmp_seq=4 ttl=64 time=0.604 ms
64 bytes from icmp_seq=5 ttl=64 time=0.607 ms
64 bytes from icmp_seq=6 ttl=64 time=0.620 ms
64 bytes from icmp_seq=7 ttl=64 time=0.466 ms
64 bytes from icmp_seq=8 ttl=64 time=0.623 ms
--- ping statistics ---
8 packets transmitted, 8 received, 0% packet loss, time 7000ms
rtt min/avg/max/mdev = 0.466/0.593/0.655/0.059 ms

Even with this routing setup, the local_ip option with vxlan tunnels does
not seem to work and GRE tunnels work.

As an aside, why do you have two bridges to the same VMs?  Your
> configuration makes it impossible to
> set a route because  you have two sets of IP addresses and routes all on
> two bridges going into the same
> VMs.  In that configuration the local ip option makes  no sense.  You
> don't need it - you're already bridged.

I was to trying to mimic a use case with two hypervisors and each
hypervisor is connected to two different underlay networks. So, used linux
bridges when imitated the topology with VMs. Please advice if this is not
the right approach.

I understand that you have seen the gre configuration work and I'm not sure
> why because it has the same
> requirements for the local ip to be routable through the remote ip.  And
> again, there is no point to the
> local ip option because the ip addresses do not need to be routed to reach
> each other.
> In any case, I'm going to set up a valid configuration and then make sure
> that the local ip option does work
> or not.  I'll report back when I'm done.
I will look out for your conclusions.

> Thanks,
> - Greg
> On 11/20/2018 10:13 AM, Gregory Rose wrote:
> On 11/20/2018 10:03 AM, Siva Teja ARETI wrote:
> On Tue, Nov 20, 2018 at 12:59 PM Gregory Rose <gvrose8192 at gmail.com>
> wrote:
>> On 11/19/2018 6:30 PM, Siva Teja ARETI wrote:
>> [user at hyp-1] ip route
>> default via A.B.C.D dev enp5s0  proto static  metric 100
>> dev testbr0  proto kernel  scope link  src
>> linkdown
>> dev testbr1  proto kernel  scope link  src
>> dev testbr2  proto kernel  scope link  src
>> Hi Siva,
>> I'm curious about these bridges.  Are they Linux bridges or OVS bridges?
>> If they are Linux bridges please provide the output of 'brctl show'.
>> If they are OVS bridges then please provide the output of 'ovs-vsctl
>> show'.
>> Thanks!
>> - Greg
> Hi Greg,
> These are linux bridges.
> [user at hyp1 ] brctl show
> bridge name bridge id STP enabled interfaces
> docker0 8000.02428928dba5 no veth6079ee7
> testbr0 8000.000000000000 yes
> testbr1 8000.fe540005937c yes vnet2
> vnet5
> testbr2 8000.fe540079ef92 yes vnet1
> vnet4
> virbr0 8000.fe54000ad370 yes vnet0
> vnet3
>  Siva Teja.
> Thanks Siva!  I'll follow up when I have more questions and/or results.
> - Greg
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openvswitch.org/pipermail/ovs-discuss/attachments/20181128/d4ec38e9/attachment-0001.html>

More information about the discuss mailing list