Did you try conducting the same experiment using a different controller? AFAIC, switch based STP and controller provided STP are not the same thing. (Correct me if I'm wrong.) Instead of relying on the controller provided STP, you might try enabling STP on the switches.
On Thu, Nov 1, 2012 at 2:50 AM, Weiyang Mo <[email protected]> wrote: > Hi, > > I tried to use Flowvisor to run with multiple OpenFlow controllers. > However I found it will have problems when I do a routing in fully mesh > networks.(with spanning_tree protocol(STP)) . > > Following is the my exepriment and result under Mininet: > > Without FlowVisor, the mininet network is a 4sw-4host fully meshed > networks and each host connects to a switch. To do the shoretest path and > avoid loop, in my controller ( POX) I run l2_multi and spanning_tree > components. Finally we pingpair the hosts and they work. > > Now, I want to insert flowvisor between Mininet and POX controller. In > this case, Mininet points to flowwvisor and flovisor has a slice which > points to the POX controller. This time the pingpair won't work because it > seems that the STP cannot work properly. (But as far as I know, inerting a > flowvisor won't affect anything, it shouldn't have the influence on the > controller's applications). > > Until now, I only have problems in fully mesh networks which must have > STP. For other Mininet networks( e.g. linear topology), inserting > flowvisor between the controller and Mininet works fine. > > Note: Assume my controller running STP and l2_multi is on > 192.168.56.102:6789 > I have the following settings: > fvctl createSlice slice1 tcp:192.168.56.102:6789 [email protected] (point > to controller) > fvctl addFlowSpace any 10 any Slice:slice1=7 > > Then run mininet which points to Flowvisor. > > Any advice? > > Thanks very much. > > Weiyang > _______________________________________________ > openflow-discuss mailing list > [email protected] > https://mailman.stanford.edu/mailman/listinfo/openflow-discuss >
_______________________________________________ openflow-discuss mailing list [email protected] https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
