I don't think it's possible to diagnose your issue from the information that you've sent.
However, one important question is: are you sure you're taking care of loops in the network? If not, it's easy to create a flood of packet-ins so large that it can't possibly be dealt with. See "I'm seeing many packet_in messages and forwarding isn't working; what gives?" and the following question in the FAQ. -- Murphy On May 5, 2014, at 2:31 AM, Shiyao Ma <i...@introo.me> wrote: > Hi. > > I am using a k=4 fat-tree topology. When the flow numbers become large (and > also, importantly, the communication with controller become heavier), I often > got keyError. > > core.openflow.connections[dpid] > Seemingly, the dpid or to say the switch lost the connection with the switch. > > So when I send something back to the switch, I got error. I wonder, what > results that > > connection loss? > > And is there a way to avoid it ? > > And I'd like to know during the execution of my own code, what other events > are happening under the hood? Do I need any explicit lock? (Though I think it > shouldn't apply here). > > I am using fatree with k=4, and the flow number is about 150 among all hosts. > I think > it's a relatively small scale, but the result error is too dissatisfying. > > Btw, when I lower done the number of flows, like, lower it to 30, the problem > won't occur. > > > > Thanks. > > > > > > > -- > > 吾輩は猫である。ホームーページはhttp://introo.me。
