I would expect so.  One use that people have brought in the past is that you 
could setup backup routes at a lower priority, with the higher ones timing out 
or being forcibly removed.  I don't know how useful that actually is, but 
people have discussed doing that before with OpenFlow.

In general, Ben and I have been fairly negative in the OpenFlow discussions 
about putting in "sanity" checks at flow insertion time due to the complexity 
and expense of enforcing them.  There was quite a lengthy discussion a year or 
so back about preventing users from being able to put overlapping flows at the 
same priority--we lost that one.  :-)

--Justin


On Feb 21, 2011, at 10:52 PM, Derek Cormier wrote:

> I expected that since the flow with the lower priority will never be matched 
> over the higher one, it has no purpose and would not be added.
> 
> Hmm, on second glance, the protocol says that if you add an identical flow 
> that also has the same priority then the flow will just be replaced.
> But, I don't think it says anything about this case. Sorry, I should have 
> confirmed that before posting.
> 
> That said, do you think these identical flows should be allowed?
> 
> - Derek
> 
> On 02/22/2011 03:43 PM, Justin Pettit wrote:
>> I would expect that you can.  What did you expect?  What did you see?  It's 
>> not clear from this output if it's working or not, since neither counter 
>> shows a value.  Are you running traffic?
>> 
>> --Justin
>> 
>> 
>> On Feb 21, 2011, at 10:35 PM, Derek Cormier wrote:
>> 
>>> Should I be able to add two identical wildcarded flows with different 
>>> priorities? I was talking with KK and he thinks this might be a bug. I 
>>> always assumed this was just the way it works. Here is my ovs-ofctl flow 
>>> dump:
>>> 
>>> cookie=0x0, duration=94.608s, table_id=0, n_packets=0, n_bytes=0, 
>>> priority=5,dl_type=0x0005 actions=drop
>>> cookie=0x0, duration=98.109s, table_id=0, n_packets=0, n_bytes=0, 
>>> priority=1,dl_type=0x0005 actions=drop
>>> 
>>> Thanks,
>>> - Derek
>>> 
>>> _______________________________________________
>>> dev mailing list
>>> [email protected]
>>> http://openvswitch.org/mailman/listinfo/dev_openvswitch.org
>> 
>> 
> 


_______________________________________________
dev mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/dev_openvswitch.org

Reply via email to