OOB control doesn't help unfortunately (I forget that HP doesn't support
inband control). We need to be able to slice the control channel so
that control packets are prioritized I think.
btw this is starting to look like a common issue as it's been reported
a few times in the past. I guess out-of-band control or control packet
priority would alleviate this
On Mon, Mar 1, 2010 at 10:20 PM, Guanyao Huang <[email protected]
<mailto:[email protected]>> wrote:
I now use smaller traces and no such problem again. I think it is
because of heavy load. ...
Thanks.
On Mon, Mar 1, 2010 at 10:12 PM, Martin Casado <[email protected]
<mailto:[email protected]>> wrote:
> Actually nevermind, I don't think the HP switch support inband
control so
> you must be using out of band.
>
> Another option is to turn off the echo timer in Nox ...
>
>
>> Sorry, my English is poor, what do you mean by "out of band
control"? :)
>>
>> On Mon, Mar 1, 2010 at 10:07 PM, Martin Casado
<[email protected] <mailto:[email protected]>> wrote:
>>
>>>
>>> I wonder if this is a problem with lost echo requests under
load ...
>>>
>>> Can you set up your environment to use out of band control?
>>>
>>>
>>>>
>>>> My program runs about 15 mins, it processes some traces. Datapath
>>>> leaves at around 12 mins.
>>>> Current CPU load is around 30%, and hit 35% occasionally.
>>>>
>>>> On Mon, Mar 1, 2010 at 9:59 PM, kk yap <[email protected]
<mailto:[email protected]>> wrote:
>>>>
>>>>
>>>>>
>>>>> Just to be blunt. What is the CPU load of your HP?
>>>>>
>>>>> Regards
>>>>> KK
>>>>>
>>>>> On 1 March 2010 21:42, Guanyao Huang <[email protected]
<mailto:[email protected]>> wrote:
>>>>>
>>>>>
>>>>>>
>>>>>> Which part defines the time interval for this echo request?
I guess
>>>>>> enlarger it will temporarily handle my problem.
>>>>>>
>>>>>> On Mon, Mar 1, 2010 at 9:38 PM, kk yap <[email protected]
<mailto:[email protected]>> wrote:
>>>>>>
>>>>>>
>>>>>>>
>>>>>>> Hi Guanyao,
>>>>>>>
>>>>>>> NOX periodic checks if the switch is still alive by
sending an echo
>>>>>>> request. Basically, your switch did not reply. You
should look at
>>>>>>> the tcpdump to see if you can find the echo reply. If
yes, bug
>>>>>>> people
>>>>>>> in this list. If no, debug your switch.
>>>>>>>
>>>>>>> Regards
>>>>>>> KK
>>>>>>>
>>>>>>> On 1 March 2010 21:34, Guanyao Huang <[email protected]
<mailto:[email protected]>> wrote:
>>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>> Hi
>>>>>>>> Sorry to bother others.
>>>>>>>> In my program the datapath leaves network again. This
time it seems
>>>>>>>> a
>>>>>>>> little different:
>>>>>>>>
>>>>>>>> 00341|openflow|WARN:stream: no response to inactivity
probe after 15
>>>>>>>> seconds, disconnecting
>>>>>>>> 00342|nox|WARN:stream: disconnected (Protocol error)
>>>>>>>> Datapath leave, f7:ca:e4:00:01:90
>>>>>>>> 00343|bindings_storage|ERR:NDB error on get_locnames_cb
(ok if
>>>>>>>> transient): Can't find specified row
>>>>>>>> 00344|openflow|WARN:stream: no response to inactivity
probe after 15
>>>>>>>> seconds, disconnecting
>>>>>>>> 00345|nox|WARN:stream: disconnected (Protocol error)
>>>>>>>> Datapath leave, f7:ca:e4:00:01:2c
>>>>>>>> 00346|openflow|WARN:stream: no response to inactivity
probe after 15
>>>>>>>> seconds, disconnecting
>>>>>>>> 00347|nox|WARN:stream: disconnected (Protocol error)
>>>>>>>> Datapath leave, f7:ca:e4:00:00:c8
>>>>>>>>
>>>>>>>>
>>>>>>>> I am wondering what is this 15 second probe.
>>>>>>>>
>>>>>>>> Thanks.
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> nox-dev mailing list
>>>>>>>> [email protected] <mailto:[email protected]>
>>>>>>>> http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>
>>>> _______________________________________________
>>>> nox-dev mailing list
>>>> [email protected] <mailto:[email protected]>
>>>> http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org
>>>>
>>>>
>>>
>>>
>
>
_______________________________________________
nox-dev mailing list
[email protected] <mailto:[email protected]>
http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org
_______________________________________________
nox-dev mailing list
[email protected]
http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org
