Please see my response below.
On Apr 8, 2016, at 2:36 PM, Justin Pettit <jpet...@ovn.org> wrote: > > >> On Apr 8, 2016, at 1:33 PM, Scott Lowe <scott.l...@scottlowe.org> wrote: >> >> Please see my responses inline, prefixed by [SL]. >> >> >>> On Apr 8, 2016, at 9:33 AM, Eugene L. Vorokov <v...@pidarasy.org> wrote: >>> >>> Hi Scott, >>> >>> Thank you for your reply. >>> According to my understanding of how STP works (please correct me if >>> I'm wrong), it will just take down one of the links to break the >>> circle. This way things will work, but that wouldn't be a working full >>> mesh. For instance, STP may take down link between 2 and 3, and then >>> packet from 2 to 3 will go through 1, which is exactly what I'm trying >>> to avoid. >> >> >> [SL] I have no direct experience running any sizable OVS implementation >> using STP, but feedback that I received from another organization using LXC >> with OVS, tunnels, and STP indicated that STP recalculated every time a >> container was attached or detached from OVS (on any of the hosts), and this >> resulted in "pauses" of traffic. > > I don't know if it would help, but OVS has supported RSTP since 2.4, so it's > possible it could have improved if they just had used plain STP. Ah, thanks Justin. This was a while ago (pretty positive it was pre-2.4), so they almost certainly would have been using plain STP. -- Scott _______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss
