[ovs-discuss] Lookup and Match count behavior

Jesse Gross jesse at nicira.com
Thu May 23 15:21:42 UTC 2013


Please still check the actual packets that are being sent using
tcpdump. If that shows what you expect, then post that output plus the
exact messages that you are seeing.

On Wed, May 22, 2013 at 9:25 AM, love you <thunder.love07 at gmail.com> wrote:
> I am sending the non matching packet from the ostinato and my physical
> interface is down, I also didn't have any virtual machine on my setup.
>
> One more thing here is that the lookup count is incremented by 2 only.
>
>
> On Wednesday, May 22, 2013, Jesse Gross <jesse at nicira.com> wrote:
>> On Wed, May 22, 2013 at 6:31 AM, love you <thunder.love07 at gmail.com>
>> wrote:
>>> sorry the behavior was observed for non-matching packets, let me explain
>>> again
>>>
>>>
>>> I have following observations regarding the dump-table stats, I think the
>>> actual behavior is incorrect, please confirm.
>>>
>>> Table stats should increment the lookup and matched count by one for
>>> every
>>> non-matching packet. When we are sending two non-maching packet for a
>>> flow
>>> entry, the match count is incremented by three.
>>>
>>> For even count of packets, odd number of matched counts are incremented.
>>>
>>> However as per my understanding the match count for non matching packet
>>> are
>>> incremented because of hidden flows of the switch.
>>>
>>>
>>> Steps:-
>>> 1) Add flow entry with valid ether-type 0x0800 with action forward
>>> 2) show table statistics
>>> 3) send two matching packets with ether type as 0x0800 for flow entries
>>> and
>>> check statistics
>>>
>>> Expected Result:-
>>> step 2 will show zero packet lookups and zero packet matches
>>> after step 3, packet lookups will be 2 and packet matches will be 2
>>>
>>> Actual Result
>>> after step 3, packet lookups will be 2 and packet matches will be 3
>>
>> ARP?
>>



More information about the discuss mailing list