Maestro not sending HELLO back to switch

[Srini Seetharaman]

Hi Zheng
Attached is the wireshark dump of communication between one of our HP
Procurve 5400 switch and the Maestro controller. You can see that the
controller is not responding back with a HELLO. Is this expected? I
think this is preventing the HP from moving into an ACTIVE mode. I am
still unable to ping because of this.

Have you had a chance to test with any of the hardware switches?

Thanks
Srini.

[Zheng Cai]

Hi Srini,

I am sorry that Maestro doesn't work for you now. We only tested
Maestro with netfpga hardware switches, and we didn't get a chance to
try real HP switches. I was just wondering which OPENFLOW version you
switch is running? Because Maestro is running 1.0.0, it won't be able
to support other OpenFlow version right now. So I guess it might be a
version issue.

Thanks,

Zheng

>  maestro-wireshark.PNG
> 87KViewDownload

[Zheng Cai]

Hi Srini,

Actually you were right. After receiving a HELLO message from the
switch, Maestro does not respond with another HELLO message back to
the switch, but with an OFPT_FEATURES_REQUEST message.

Instead, Maestro sends a HELLO message immediately to the switch when
the TCP connection gets created. We implemented Maestro in this way
because in the OpenFlow V 1.0.0 specification, it says "When an
OpenFlow connection is rst established, each side of the connection
must immediately send an OFPT_HELLO message with the version eld set
to the highest OpenFlow protocol version supported by the sender".
Maybe I interpreted it wrong. But at least when we try Maestro with
software OpenFlow switches, it works.

Is it required for the controller to respond to HELLO messages with
another HELLO message? If it is the case, I will prepare a patch to
the existing Maestro code.

Thank you very much for your feedback!

-Zheng

>  maestro-wireshark.PNG
> 87KViewDownload

[Srini Seetharaman]

Hi Zheng
It is possible that the spec is not clear in this regard, and may need
to be brought up on the openflow-spec list.

For now, could you please check the NOX code to see if it does respond
to HELLO messages. If that's cumbersome, can you give me a version of
Maestro that has the patch you mention. Seeing if that patched Maestro
works, will give us a clue on how to proceed.

Thanks
Srini.

[Zheng Cai]

Hi Srini,

I will investigate NOX's code and the OpenFlow more carefully, and try
to come up with a working patch as soon as possible.

Thanks again,

-Zheng

On Jan 5, 9:06 pm, Srini Seetharaman <srini...@gmail.com> wrote:
> Hi Zheng

[Zheng Cai]

Hi Srini,

After I checked NOX code, I think that it does not "respond" to hello
messages from switches, but instead it sends out a hello message when
one connection is created. The first attachment is the wireshark
capture I got for NOX.

I also tried modifying Maestro code to let it actually respond to
hello messages from switches, in addition to the sending of the
initial hello message. Then the switch sent back an error message
which seems to indicate that the second hello message respond was
duplicated and not understood. This can be seen in the second
attachment.

Then I checked again your wireshark capture, and found out that it
said "TCP retransmission of Hello message". Could it be the problem of
the TCP connection? I guess it might be related to the IPv6 not
listening problem in the other thread. I was also just wondering does
NOX work for your HP switch?

Anyway I am also attaching the simple patch which enables Maestro to
respond to hello messages. Put it in the Maestro-0.1.0/drivers
directory and recompile the code, to see whether it will work for you.
Since this patch might be problematic, I am not adding it to the
repository right now.

Thank you very much,

-Zheng