Hi Sylar,
First thing that is perhaps a detail but it seems that the DPIDs your switches
have are from mininet (00:00:00:00:00:00:00:01, 00:00:00:00:00:00:00:04)
although you may have set the dpid in the NEC and Cisco switch. But just in
case you have a mininet running somewhere that is creating havoc in your setup.
Another thing, that is odd is the BufferUnderflowException which is usually
caused when we read from the input buffer to FV with a too large length value.
Could you please sniff the connection between the RouteFlow and FlowVisor and
send the pcap?
Cheers.
--
Ali
On 26 Feb 2014, at 06:44, Sylar Shen <kimul...@gmail.com> wrote:
> Dear experts:
> I have a simple test environment in a LAN which contains...
> Controller(Routeflow) * 1
> Flowvisor(CentOS 6.5, version 1.4) * 1
> NEC PF5240 openflow switch *1
> Cisco WRT54GC openflow switch *1
>
> Flowvisor manages the two openflow switches and have a slice named
> "routelfow" for Routeflow.
> The flowspace(the name is myflowspace) for Routeflow has no constraints which
> means Routeflow can use any switch and the rule is like.....
> {"force-enqueue": -1, "name": "myflowspace", "slice-action": [{"slice-name":
> "routeflow", "permission": 7}], "queues": [], "priority": 1, "dpid":
> "all_dpids", "id": 10, "match": {"wildcards": 4194303}}
>
> After setting up the environment and running, everything seemed to be OK.
> The Routeflow worked fine and the Flowvisor was OK as well.
> But....after maybe 3 minutes later, the connections started to be unstable.
> Therefore, I got many error and warning information that came from the NEC
> switch constantly.
> And a lot of warning information from both switches.
> Here are some examples from the NEC switch....
> Error log
> 1. flowvisor: ERROR classifier-dpid=00:00:00:00:00:00:00:01 : STARVING:
> handling event took 26ms: org.flowvisor.events.FVIOEvent@4f853093
> 2. flowvisor: ERROR slicer_routeflow_dpid=00:00:00:00:00:00:00:01 : STARVING:
> handling event took 24ms: org.flowvisor.events.FVIOEvent@638d0d04
> 3. flowvisor: ERROR slicer_routeflow_dpid=00:00:00:00:00:00:00:01 : got
> unknown error; tearing down and reconnecting:
> java.nio.BufferUnderflowException
> Warning log
> 1. flowvisor: WARN none : End of LLDPDU is missing <-- (This showed a lot...)
> 2. flowvisor: WARN classifier-dpid=00:00:00:00:00:00:00:01 : inserting drop
> (hard=0,idle=1) rule for
> FlowEntry[dpid=[all_dpids],ruleMatch=[OFMatch[]],actionsList=[Slice:routeflow=7],id=[10],priority=[1],]
> 3. flowvisor: WARN slicer_routeflow_dpid=00:00:00:00:00:00:00:01 :
> flowrewriteDB: tried to remove non-existent flow FVFlowRemoved
> [match=OFMatch[in_port=1,dl_dst=02:a1:a1:a1:a1:a1,dl_src=00:17:5a:ea:cc:6b,dl_type=0x800,dl_vlan=0xffff,dl_vpcp=0,nw_dst=20.0.0.2,nw_src=20.0.0.1,nw_proto=6,nw_tos=48,tp_dst=42930,tp_src=179]]
>
> Here are the warning information that came from both switches.
> Both NEC switch and Cisco WRT switch show this warning...
> flowvisor: WARN slicer_routeflow_dpid=00:00:00:00:00:00:00:04 : Verifying
> Slice is not over its flow rule limit
>
> However, if I removed Flowvisor and let the switches connect to Routeflow
> directly, everything was back to normal and worked fine.
> I've tried everything I can imagine, for example, make Routeflow send hello
> message from 1 second to 10 seconds, but still, the error and warning
> messages kept showing.
> Hope someone who could help me with this problem. I will appreciate it.
> Thank you very very much....^^
> --
> Best Regards,
> Sylar Shen
> _______________________________________________
> openflow-discuss mailing list
> openflow-discuss@lists.stanford.edu
> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
_______________________________________________
openflow-discuss mailing list
openflow-discuss@lists.stanford.edu
https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
