On 21/09/2020 11:17, Numan Siddique wrote:
> On Mon, Sep 21, 2020 at 2:49 PM Mark Gray <[email protected]> wrote:
> 
>> On 14/09/2020 19:12, Numan Siddique wrote:
>>> On Mon, Sep 14, 2020 at 10:18 PM Dumitru Ceara <[email protected]>
>> wrote:
>>>
>>>> On 9/12/20 9:10 AM, Mark Gray wrote:
..
>>>>
>>>> I'm not sure whether this is completely correct.
>>
>> You are right, it is not correct.
>>
>>>>
>>>> What if the sequence of operations is:
>>>>
>>>> ovn-nbctl set logical_switch sw other_config:requested-tnl-key=42
>>>> # ovn-northd writes to datapath_binding.tunnel_key
>>>>
>>>> ovn-sbctl set datapath_binding sw external_ids:foo=bar
>>>>
>>
>> I hadn't considered that case.
>>
>> If I only check for the case when the tunnel key gets updated, then this
>> will force recompute on too many changes. For example, when I tried
>> this, some tests in ovn_performance.at fail (when adding a logical
>> router or logical switch). I don't think this is acceptable?
>>
> 
> I think it should be acceptable given that this use case happens rarely.
> I think it's better to fall back to recompute in rare cases than handle it
> incrementally
> (and in the process make the code more complicated :) if it is not
> straightforward to address it)
> 
> Thanks
> Numan
> 
>

Yeah, I don't want to fall into the trap of optimizing too early. What I
don't know is how often a lr-add or ls-add happens. However, from what
you are suggesting here, it is not too often so I will make the change
as you suggest.

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to