[ovs-dev] Q on ovs-dpdk + Vxlan user space tunneling

Ravi Kerur rkerur at gmail.com
Wed Oct 17 17:00:07 UTC 2018


Hello Daniel,

When I tried to configure vxlan routes I had mentioned I was seeing errors,
when I checked log files under /var/log/openvswitch there is no
information...

-rw-r----- 1 root adm       0 Oct 17 06:25 ovs-vswitchd.log
-rw-r----- 1 root adm     536 Oct 17 12:32 ovsdb-server.log

cat ovsdb-server.log
2018-10-17T12:26:46.865Z|00001|vlog|INFO|opened log file
/var/log/openvswitch/ovsdb-server.log
2018-10-17T12:26:47.185Z|00002|ovsdb_server|INFO|ovsdb-server (Open
vSwitch) 2.10.0
2018-10-17T12:26:47.334Z|00003|fatal_signal|WARN|terminating with signal 15
(Terminated)
2018-10-17T12:32:25.971Z|00001|vlog|INFO|opened log file
/var/log/openvswitch/ovsdb-server.log
2018-10-17T12:32:26.663Z|00002|ovsdb_server|INFO|ovsdb-server (Open
vSwitch) 2.10.0
2018-10-17T12:32:26.783Z|00003|fatal_signal|WARN|terminating with signal 15
(Terminated)
~

Link http://docs.openvswitch.org/en/latest/howto/userspace-tunneling/
doesn't mention which version of OVS was tested, looks like it is part of
2.10.90? Should I download latest code for this?

Thanks

On Sun, Oct 14, 2018 at 7:51 AM Ravi Kerur <rkerur at gmail.com> wrote:

> Hello Daniel,
>
> Thank you for your quick response and explanation on br-phy to br-int. I
> will gather debug logs and update.
>
> Thanks.
> Ravi
>
> On Sun, Oct 14, 2018 at 6:19 AM benli ye <danielbenliye at gmail.com> wrote:
>
>> Hi Ravi,
>>
>> Let me try to answer your question.
>> (1) For the first one: I believe NORMAL action can lead the packet to
>> tunnel port.
>> (2) For how packets flow from br-phy to br-int, You can search code
>> “terminate_native_tunnel()” in ofproto-dpif-xlate.c.
>>      When it generates datapath flow for an incoming tunnel packet, a
>> OVS_ACTION_ATTR_TUNNEL_POP is inserted in the actions.
>>      When the encapsulated packet hit OVS_ACTION_ATTR_TUNNEL_POP action,
>> it will be handled to tunnel port on br-int as a re-injected packet.
>>      Then tunnel headers can be split and delivered to corresponding
>> interface in br-int. When packet sent out from br-int to br-phy, it will
>> hit the tnl_push action which
>>      is added when computing output action (see native_tunnel_output()),
>> then it will use the route you added to send the encapsulated packet from
>> br-phy interface.
>> (3) That’s a historic issue as it copied from an original guide. I
>> believe your assumption is right.
>> (4) I don’t try this command, I think more log information in
>> /var/log/openvswitch may be helpful. Hope someone have this answer.
>>
>> Thanks,
>> Daniel Benli Ye
>>
>> > On Oct 14, 2018, at 8:55 PM, Ravi Kerur <rkerur at gmail.com> wrote:
>> >
>> > Hi,
>> >
>> > I am going through following link and setting up the testbed
>> >
>> > http://docs.openvswitch.org/en/latest/howto/userspace-tunneling/
>> >
>> > and it is very confusing to me at least. Can someone please clarify
>> > following questions?
>> >
>> > (1) I believe flows needs to be programmed on br-int bridge with
>> > in_port=vm_port0, out_port=vxlan0, vice-versa. When bridge is created it
>> > has a default flow with NORMAL action, i don't think this suffices?
>> >
>> > (2) How do packets flow from br-int to br-phy and br-phy to br-int?
>> >
>> > (3) Link says add following route
>> >
>> > "ovs-appctl ovs/route/add 172.168.1.1/24 br-eth1"
>> >
>> > What is br-eth1? I don't see anything about it in the link?
>> >
>> > (4) Assuming br-eth1 is br-phy, I tried to add route on the testbed I
>> > am configuring I get
>> >
>> > ovs-appctl ovs/route/add 11.11.11.4/32 br1-phy
>> > Error while inserting route.
>> > ovs-appctl: ovs-vswitchd: server returned an error
>> >
>> > Thanks.
>> > _______________________________________________
>> > dev mailing list
>> > dev at openvswitch.org
>> > https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>>
>>


More information about the dev mailing list