On 8/14/23 00:24, Han Zhou wrote:
> On Fri, Aug 11, 2023 at 5:25 AM Dumitru Ceara <dce...@redhat.com> wrote:
>>
>> It's safe to assume that tables that existed in the previous LTS branch
>> first release (currently 22.03.0) can be monitored directly.  Do so and
>> only "optionally" monitor the ones that have been added since.
>>
>> This way we avoid the need for the IDL to expose an API to change the
>> default condition for monitored tables.  It also avoids complex code in
>> ovn-controller because we'd otherwise have to explicitly re-initialize
>> conditions to a non-default (false) value after every SB reconnect.
>>
>> NOTE: In order to make sure that pre-existing L3 and L2 gateways are not
>> initially considered "non-local" we explicitly request for all port
>> bindings of this type to be monitored in the startup stage (before we
>> got the initial contents of the database and our chassis record).
>>
> 
> The commit subject and message had a good introduction about the solution
> but it would be better to at least add a brief description of the problem
> (about the memory spike) before the first paragraph.
> It would also be good to add a Fixes tag for commit 1b0dbde.

Ack, I did that now.

> Otherwise it looks good to me!
> 
> Acked-by: Han Zhou <hz...@ovn.org>
> 

Thanks, applied to main and backported to all branches down to 22.03.

Regards,
Dumitru

_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to