On Jun 7, 2012, at 1:47 PM, shreya pandita wrote:
> Jan 01 04:44:18|00009|rconn|INFO|tcp:10.0.0.5:6633: connected
> Jan 01 04:44:48|00010|rconn|ERR|tcp:10.0.0.5:6633: no response to inactivity
> probe after 15 seconds, disconnecting
> Jan 01 04:44:48|00011|rconn|INFO|tcp:10.0.0.5:6633: connection dropped
> Jan 01 04:44:49|00012|rconn|INFO|tcp:10.0.0.5:6633: connecting...
>
>
> The controller and the openwrt switch are connected , for sometime and then
> they lose connection . Is this a normal behaviour ??. Is there any file or
> something where I can increase the inactivity period.
Hi Shreya,
I have seen this behavior as well. It's not a serious problem for two reasons
1) it immediately re-connects after dropping the connection due to inactivity,
and 2) if you have any OF messages being exchanged between the switch and the
controller, these heartbeat messages won't be responsible for keeping the
connection alive, and you won't see the disconnect/reconnect cycle.
I've looked into why this is happening and it seems to be some sort of timing
issue. The switch receives the OF ECHO_REPLY message from the controller, but
doesn't process it. However, if you add the option "--verbose=vconn:file:dbg"
to ofprotocol, then it *does* process the ECHO_REPLY message. That seems to be
enough to cause the packet to be processed.
I recently tried re-compiling the kernel for the TP-LINK to use 1000 Hz timers
instead of the default 100 Hz, but that causes the ethernet drivers not to
load, and you'll need to re-flash the firmware with a serial cable. :-) I'm
going to try 250 Hz timers when I get a chance, but I'm still speculating on
the root cause.
Happy to share more details with you off list.
cheers,
Andrew
_______________________________________________
openflow-discuss mailing list
openflow-discuss@lists.stanford.edu
https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
