Thanks for getting back to me.

It would be good to try both if at all possible. 2nd patch is probably
closer to what I plan to post for the 'net' tree.

On Fri, Oct 23, 2015 at 9:05 AM, Flavio Leitner <f...@redhat.com> wrote:
>
> Hi Andy,
>
> I built a test package with your previous patch to Joe, but
> we need to schedule an appropriate time to test it, so it's
> a bit slow on that front.
>
> Having said that, which patch do you think is better to try first?
>
> Thanks,
> fbl
>
> On Thu, Oct 22, 2015 at 04:36:20PM -0700, Andy Zhou wrote:
>> Hi, Joe,
>>
>> This is potentially a better fix that I'd like to propose for the
>> 'net' tree. It would be great if you can test it in your set up to
>> see if it solved the issue you are facing with.
>>
>> Thanks,
>>
>> andy
>>
>> diff --git a/datapath/actions.c b/datapath/actions.c
>> index c529bbb..208eb30 100644
>> --- a/datapath/actions.c
>> +++ b/datapath/actions.c
>> @@ -1003,11 +1003,11 @@ int ovs_execute_actions(struct datapath *dp,
>> struct sk_buff *skb,
>>         err = do_execute_actions(dp, skb, key,
>>                                  acts->actions, acts->actions_len);
>>
>> -       if (!level)
>> -               process_deferred_actions(dp);
>> -
>>         this_cpu_dec(exec_actions_level);
>>
>> +       if (level <= 1)
>> +               process_deferred_actions(dp);
>> +
>>         /* This return status currently does not reflect the errors
>>          * encounted during deferred actions execution. Probably needs to
>>          * be fixed in the future.
>>
>>
>>
>> On Tue, Oct 20, 2015 at 8:55 PM, Andy Zhou <az...@nicira.com> wrote:
>> > On Tue, Oct 20, 2015 at 1:42 PM, Joe Talerico <jtale...@redhat.com> wrote:
>> >> When using balance-tcp bonding with OVS we were seeing ARP issues when we
>> >> reached ~ 100 guests. I Tracked as much as possible here :
>> >> https://bugzilla.redhat.com/show_bug.cgi?id=1267291
>> >>
>> >> Has anyone seen this behavior before?
>> >>
>> >> Switching to active/backup resolves the issue.
>> >
>> > Thanks for reporting the issue with lots of relevant information.  I
>> > have not seen this issue before.
>> > One the error messages in the bugzilla report suggested that you may
>> > ran into the deferred action fifo limit.
>> >
>> > Would you please try to increase its size with the following patch,
>> > and report back how much it helped. This changes OVS kernel module, so
>> > you will have to recompile the kernel module.
>> >
>> > This is not like the right fix, but should help to confirm if there
>> > are any other issues in the way to get your set up working.
>> >
>> > diff --git a/datapath/actions.c b/datapath/actions.c
>> > index c529bbb..a4afecb 100644
>> > --- a/datapath/actions.c
>> > +++ b/datapath/actions.c
>> > @@ -53,7 +53,7 @@ struct deferred_action {
>> >         struct sw_flow_key pkt_key;
>> >  };
>> >
>> > -#define DEFERRED_ACTION_FIFO_SIZE 10
>> > +#define DEFERRED_ACTION_FIFO_SIZE 100
>> >  struct action_fifo {
>> >         int head;
>> >         int tail;
>>
_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss

Reply via email to