[pox-dev] l2_learning_tutorial flow rules (strange behavior?)
Murphy McCauley
murphy.mccauley at gmail.com
Mon Oct 1 11:57:57 PDT 2012
Yes, that's always going to happen with l2_learning_tutorial. This is, in fact, an example of how NOT to try to write a learning switch, but without proper context. I deleted it in development fork/branch a while back, but looks like I never did on the master branch. I have done so now.
The problem you're seeing here was actually just discussed on this list a few days ago. The issue is that the "learned" rules result in hiding other packets from the controller, so it never learns all the destinations.
If you want a good example of a switch for a tutorial, look at forwarding.l2_pairs or samples.of_tutorial. l2_pairs gets around the issue of hiding other sources by only installing rules for each PAIR of learned addresses. These may only be available in the betta (development) branch.
-- Murphy
On Oct 1, 2012, at 11:38 AM, Tmusic wrote:
> Hi,
>
> I'm trying to learn how to establish flow rules by studying l2_learning_tutorial, but I'm experiencing rather strange behavior.
>
> On a VM with mininet (one switch, three hosts) and POX installed I first start up mininet and then the l2_learning_tutorial (the behavior is easier to see this way).
> I added one line after the flow installing to see what happens: print 'added route to', packet.src,'(packet destined to',packet.dst,')'
>
> When I perform pingall in mininet (one or more), the same routes are added over and over again (see print statement above).
> Dpctl shows that only two entries are being added (I expected three):
>
> cookie=0, duration_sec=268s, duration_nsec=529000000s, table_id=0, priority=32768, n_packets=16, n_bytes=1344, idle_timeout=0,hard_timeout=0,
> dl_dst=00:00:00:00:00:02,actions=output:1
> cookie=0, duration_sec=268s, duration_nsec=463000000s, table_id=0, priority=32768, n_packets=11, n_bytes=910, idle_timeout=0,hard_timeout=0,dl_dst=00:00:00:00:00:03,actions=output:2
>
> The packets seem to keep going to the controller.
>
> Any ideas?
>
> Thank you in advance!
>
> Tim
>
>
>
>
>
>
More information about the pox-dev
mailing list