[ovs-discuss] [Solved][Bug] flow deletion before transmission end - transmission stalled

[Ben Pfaff]

I'm glad that you were to avoid to solve the problem.

Was the problem and solution anything worth documenting so that we could
help others to avoid this problem in the future?

Thanks,

Ben.

On Sat, Nov 27, 2010 at 11:36 AM, Christian Fischer
<christian.fischer at easterngraphics.com> wrote:
> I had a day to think about that.
> I had a fragmentation problem. Works now.
>
> On Friday 26 November 2010 20:17:53 Christian Fischer wrote:
>> On Friday 26 November 2010 17:13:40 Christian Fischer wrote:
>> > Same problem with head here.
>>
>> I've tried version 0.99.2, 1.1.0pre2 and head, always the same result.
>>
>> I must correct my statement referring to the vlan settings:
>> - trunks=501 isn't working
>> - tag=501 was never set, this was a setting of the physical switch
>>
>> Fact is that configuring vlan trunks causes trouble im my case as
>> described. It works if vlans are managed by the physical switch.
>>
>> Sorry for the confusion.
>>
>> > On Friday 26 November 2010 15:52:19 Christian Fischer wrote:
>> > > Is there a known bug with 1.1.0pre2 on XCP-0.5 which prevents from e.g.
>> > > copying files bigger than 1.3kb via scp over vlan trunks?
>> > >
>> > > Tried to copy over physical eth0, transmission stalles, trunks=501.
>> > > Tried to copy over physical eth0, transmission ok, tag=501.
>> > >
>> > > No problem with arp requests/replies after sending/receiving a lot of
>> > > acks.
>> >
>> > should mean: after sending but not receiving acks
>> >
>> > > Reproducable: always, stalles after asking for password and printing
>> > >
>> > >  "100% ...0:00"
>> > >
>> > > Fish reports "No route to host", ssh works, ping works, Xen manager
>> > > https works.
>> > >
>> > > Tcpdump can't see any packets on eth0 after connection is stalled but
>> > > the pysical switch connected to eth0 is sending them (tcpdump
>> > > listening on mirror port, lot of acks and arp).
>> > >
>> > > As far I could see ovs deletes the flows before the transmission ends.
>> > > No new calls of process_flow() after deletion for this connection,
>> > > only for new ones.
>> > >
>> > > Comments?
>> > >
>> > > Regards
>> > > Christian
>> > >
>> > > _______________________________________________
>> > > discuss mailing list
>> > > discuss at openvswitch.org
>> > > http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org
>> >
>> > _______________________________________________
>> > discuss mailing list
>> > discuss at openvswitch.org
>> > http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org
>>
>> _______________________________________________
>> discuss mailing list
>> discuss at openvswitch.org
>> http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org
>