Re: [ovs-discuss] About VXLAN connecting container and physical machine

[i_j_e_...@yahoo.co.jp]

Hi.

Thanks for reply. I added some flow controls but the result is not effective. 
Here's the detail.

To see dump of vxlan packets on the ovs whose port has physical connection to 
34:95:db:28:4a:5e,
it was found that the arp reply arrived to the ovs.

[machine4]
2020-09-07 02:08:34.471064 Out 34:95:db:28:4a:5e ethertype ARP (0x0806）,
  length 44: Request who-has 172.20.0.21 tell 172.20.0.100, length 28

[machien1]
2020-09-07 02:08:34.471727 Out 70:85:c2:44:70:37 ethertype IPv4 (0x0800）,
  length 112: 192.168.1.2.34064 > 192.168.1.30.4789: VXLAN, flags [I] (0x08), 
vni 1025
  34:95:db:28:4a:5e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806）,
  length 60: Request who-has 172.20.0.21 tell 172.20.0.100, length 46
2020-09-07 02:08:34.471749 Out 70:85:c2:44:70:37 ethertype IPv4 (0x0800）,
  length 112: 192.168.1.2.34064 > 192.168.64.1.4789: VXLAN, flags [I] (0x08), 
vni 1025
  34:95:db:28:4a:5e > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806）,
  length 60: Request who-has 172.20.0.21 tell 172.20.0.100, length 46
2020-09-07 02:08:34.477723  In 08:00:27:c2:5a:9f ethertype IPv4 (0x0800）,
  length 94: 192.168.1.30.50525 > 192.168.1.2.4789: VXLAN, flags [I] (0x08), 
vni 1025
  e6:c2:22:f9:1c:41 > 34:95:db:28:4a:5e, ethertype ARP (0x0806）,
  length 42: Reply 172.20.0.21 is-at e6:c2:22:f9:1c:41, length 28

Since it is necessary for the ovs to transfer arp reply to the destination, I 
added the following.

  # ovs-ofctl add-flow ovsbr0 
dl_type=0x0806,nw_proto=2,dl_dst=34:95:db:28:4a:5e,actions=output:15

However, the arp replies are not transferred to machine4.

  # ovs-ofctl dump-flows ovsbr0
  cookie=0x0, duration=172.977s, table=0, n_packets=0, n_bytes=0, 
arp,dl_dst=34:95:db:28:4a:5e,arp_op=2 actions=output:enp1s0f2
  cookie=0x0, duration=253.776s, table=0, n_packets=10, n_bytes=670, priority=0 
actions=NORMAL

Let me know what's wrong with the configuration. 

Regards,

> 2020/08/31 10:50、Heim, Dennis のメール:
> 
> When I have done VXLAN, I did it with a controller. Remember VXLAN defines 
> the data plane, but not the control plane (Mac address, etc).
> 
> -Original Message-
> From: discuss  On Behalf Of 
> i_j_e_...@yahoo.co.jp
> Sent: Sunday, August 30, 2020 1:17 PM
> To: ovs-discuss@openvswitch.org
> Subject: [ovs-discuss] About VXLAN connecting container and physical machine
> 
> Hi.
> 
> I tried to establish L2 network between a docker container and a physical 
> machine by using vxlan of OVS. The container host is different from the 
> physical machine. So far, it is not successful. I'd like to have comments on 
> how to do this.
> 
> Here's the detail.
> 
> As shown in the figure and description below, there are five physical 
> machines, machine1, machine2, machine3, machine4, and machine5. On all of 
> them are running linux. Each of machine1, machine2 and machine3 has a OVS 
> bridge.
> 
> Connection between docker2 and docker3 works fine. Connection between 
> machine4 and machine5 also works fine. When ovsbr1 is down, machine4 and 
> machine5 connection fails. Therefore, it is sure that ovsbr1 is forwarding 
> packets between machine4 and machine5.
> 
> However, connection between docker2 and machine4 is not established. Ping 
> never reaches between these. I dumped packets and found the following.
> 
> Docker2 ping to machine4 did not succeed, during which
>  a) arp requests arriveed on machine1 via the vxlan channel.
>  b) no arp request arrived on machine4.
> Docker2 ping to docker3 succeeded, during which
>  c) arp requests arrived on machine3 via the vxlan channel.
>  d) arp requests arrived on docker3 and docker3 replied.
> 
> To summerize, OVS does not forward arp request comming from vxlan to its 
> connected physical ports in this configuration.
> 
> Please let me know how to establish connection between docker2 and machine4 
> in this configuration.
> 
> Regards,
> 
>   +---+ +---+
>   |docker2| |docker3|
>   +--++ +--++
>  |eth2 |eth3
>  | |
> ovsbr2+--+--+ ovsbr3+--+--+
>  +-+ +-+   +-+
>  | machine1| |   machine2  |   |   machine3  |
>  +---+-+ +++   +++
> ovsbr1+-+--+ || |
> |  | || |
> enp1s0f0+  +enp1s0f1 |enp3s0enp0s3|   enp0s3|
> |   \++-+
> |\ physical switch
> | \
> |eth4  \eth5
>   +-++  +--+
>   | machine4 |  | machine5 |
>   +--+  +--+
> 
> 
> Machine1:
>  Three physical ports, enp1s0f0, enp1s0f1 and enp3s0.
>  A ovs switch named ovsbr1.
>  Enp1s0f[

[ovs-discuss] Can we use LoRa technology with OVS?

[Jordan Sandri]

First thing first, a little bit of context. 
I have 2 raspberry pis with OVS and Docker installed on it. I use ONOS as a 
controller.
On the first one, I have a bridge (BR1) and I created a Docker container (CTN1) 
linked to the bridge thanks to ovs-docker.
On the second, I have a bridge (BR2) and I created a Docker container (CTN2) 
linked to the bridge thanks to ovs-docker as well.
The first goal is that CTN1 pings CTN2 and the other way around of course and 
the second is to link pis with multiple types of communication.

First, I have linked the 2 pis with an ethernet cable, added the eth port to 
each bridge for each pi, and asked CTN1 to ping CTN2 and it worked.
Then I created a wifi ad hoc network between pis and using GRE tunnel with OVS 
I did a wifi connection between pis and finally, CTN1 was able to ping CTN2. 
Now I bought two SX1268 470M LoRa hat. Here is some information about it 
(https://www.waveshare.com/wiki/SX1268_470M_LoRa_HAT#Demo_codes 
) and would 
like to do the same thing again but using LoRa technology. So I plugged hats on 
each pi. However, this time I don’t have a network interface to add to the 
bridge or a way to create a tunnel. Can CTN1 ping CTN2 combining OVS with LoRa 
technology?
I saw this picture showing that it is 
(https://www.mdpi.com/IoT/IoT-01-4/article_deploy/html/images/IoT-01-4-g007.png
 
).
 I know that SDWI stands for Software-Defined Wireless Interfacing but what is 
it exactly? How to do so? Is there an easy way to combine LoRa and OVS?


___
discuss mailing list
disc...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-discuss