Hello, everyone
There is a question.
The Host ip config:
10.1.1.20
255.0.0.0
Gateway:10.1.1.2
and the architecture graph.
The host(10.1.1.20) will ping to switch(192.168.6.162) success.
and switch(192.168.6.162) also ping to controller(192.168.6.189) success.
but the host can't ping through switch.
10.1.1.20 straight to connect 192.168.6.189 is not working.
HP 2920 config is:
ip default-gateway 192.168.6.189
ip routing
snmp-server community "public" unrestricted
openflow
controller-id 1 ip 192.168.6.189 controller-interface vlan 2
instance "control"
listen-port
member vlan 1
controller-id 1
mode passive
enable
exit
enable
exit
oobm
ip address dhcp-bootp
vlan 1
name "DEFAULT_VLAN"
no untagged 3-14,17-24
untagged 1-2
tagged 15-16
ip address 10.1.1.2 255.255.255.0
vlan 2
name "VLAN 2"
untagged 3-24
ip address 192.168.6.162 255.255.255.0
Where is the problem?
Thank you all.
My suspicion is that the ICMP request is being routed by the switch from vlan 1 to vlan 2 and is reaching the controller, based on what you've shown. However, for "ping" to work the ICMP response must also make its way back from the controller (vlan 2) to the host (vlan 1). To do this, the controller will need a route that shows how to reach 10.1.1.20, since that IP isn't on the controller's local subnet.
Two questions:
When thinking in terms of data-plane (vlan 1) and control-plane (vlan 2), you generally don't want your hosts to be able to send traffic on the control-plane, since that bandwidth should be used for control communication between switches and controllers. Given that, even if you get this working I would not recommend it unless you can share a reason why the host needs to ping the controller's control-plane IP.
If you need to manage the controller using the host, then I'd set up a separate management-plane (vlan+subnet) and create a separate interface on the controller. That way it will separate management traffic (host-to-controller) from control traffic (controller-to-switch).
Anwser:
1.
2.
No. controller can't ping to 10.1.1.2
And, why do i use the "ping"?
I order to determine the path alive.
Or are there other ways to determine.
There is my finally architecture.
I need the host connect to the other one.
Is the host traffic through the controller necessary?
or I add two cable between these three switch.
In thinking about an SDN-controlled network, you'll always want to think in terms of management-plane, control-plane, and data-plane:
Traffic should never cross between these three planes. The exception to that statement is when an event on one plane (such as a packet being received on the data-plane) causes another event on another plane (such as an OpenFlow packet-in message sent to the controller on the control-plane).
In your setup, you'll need the links you marked with '?' because they provide data-plane connectivity between end-hosts. The SDN controller never explicitly forwards the packets of these end-hosts, it only uses the OpenFlow (or other) protocol to tell switches how to forward the packets. The decision-making is done by the SDN controller and communicated to the switches over the control-plane, but the action of forwarding packets is always carried out by the switches themselves on the data-plane.
The controller is sending ping replies out its eth0 interface, which is to say in the wrong direction.
The controller needs a route added for the 10.0.0.0/8 network.
I have another question about create a links between the three switches.
and I post a new article :