How do the switches crash? With an assertion about insufficient buffer headroom or something? Or about a bad buffer ID (are you applying multiple actions?)? Or do they just segfault?
Feel free to send along a capture of the OpenFlow traffic between NOX and the switch (along with the NOX and switch logs if possible), and we can take a look. Though I'd be surprised if it was NOX sending something invalid that was causing your problem. If you can develop a very simple application that crashes the switches, that would be valuable, and you might try sending it to one of the OpenFlow mailing lists too. Any which way, the simplest answer is probably to use Open vSwitch instead of the OpenFlow reference implementation. -- Murphy On Nov 8, 2011, at 1:31 AM, andrea simeoni wrote: > Hi guys, > > I'm running a test with 2 hosts, 2 OF switches and 1 controller. The switches > and hosts are on 4 different ubuntu virtual machines and the controller is on > the host machine. For switches I use the user space OF implementation. > I modified the simplerouting code to install a OFPP_CONTROLLER action in both > switches when a flow needs to be routed. To install actions I use the simple > API provided by routeinstaller. It happens that both switches crash and close > connection to controller. > I tried the SAME controller on mininet (also with the same network topology) > and things seem be working. > Any suggestions? > -- > Andrea Simeoni > _______________________________________________ > nox-dev mailing list > nox-dev@noxrepo.org > http://noxrepo.org/mailman/listinfo/nox-dev _______________________________________________ nox-dev mailing list nox-dev@noxrepo.org http://noxrepo.org/mailman/listinfo/nox-dev
