On Sat, Sep 5, 2020 at 12:46 AM Numan Siddique <num...@ovn.org> wrote:

>
>
> On Sat, Sep 5, 2020, 12:04 AM Han Zhou <hz...@ovn.org> wrote:
>
>> On Fri, Sep 4, 2020 at 11:18 AM Numan Siddique <num...@ovn.org> wrote:
>> >
>> >
>> >
>> > On Wed, Sep 2, 2020 at 4:57 AM Han Zhou <hz...@ovn.org> wrote:
>> >>
>> >> This option is added to avoid the flooding of nb_cfg updates from
>> >> a large nubmer of hypervisors interfere with the original SB DB
>> >> data distribution and handling, so that the e2e control plane latency
>> >> can be measured more accurately during scale testing.
>> >>
>> >> Signed-off-by: Han Zhou <hz...@ovn.org>
>>
>
Hi Han,

If I understand correctly, if the option is configured, ovn-controller will
sleep before it updates
the OVN Southbound database. Can you please explain a bit on how it helps
in latency measurement.

Also I don't think this option will be used during live deployments. So
instead of adding this as configuration
in the ovs db, I would suggest to add an ovn-appctl command -
debug/set-delay-nb-cfg-report (or  a more
appropriate name) so that a control plane latency test suite can set this
option -
   ovn-appctl -t ovn-controller  debug/set-delay-nb-cfg-report 3 (for
example).

Thanks
Numan



>> ---
>> >>  controller/ovn-controller.8.xml | 12 ++++++++++++
>> >>  controller/ovn-controller.c     |  9 +++++++++
>> >>  2 files changed, 21 insertions(+)
>> >>
>> >> diff --git a/controller/ovn-controller.8.xml
>> b/controller/ovn-controller.8.xml
>> >> index 6687731..4cb3f41 100644
>> >> --- a/controller/ovn-controller.8.xml
>> >> +++ b/controller/ovn-controller.8.xml
>> >> @@ -233,6 +233,18 @@
>> >>          The boolean flag indicates if the chassis is used as an
>> >>          interconnection gateway.
>> >>        </dd>
>> >> +
>> >> +      <dt><code>external_ids:ovn-delay-nb-cfg-report</code></dt>
>> >> +      <dd>
>> >> +        An integer that represents the time (in seconds) to sleep
>> before
>> >> +        <code>ovn-controller</code> updating <code>nb_cfg</code>
>> column
>> >> +        of <code>Chassis_Private</code> table after it completes
>> processing
>> >> +        the corresponding changes. By default it is 0 (meaning do not
>> sleep).
>> >> +        This value can be configured when doing control plane latency
>> testing,
>> >> +        to make sure the updates of <code>nb_cfg</code> to the
>> Southbound DB
>> >> +        from a large number of hypervisors do not interfere with the
>> original
>> >> +        Southbound DB data processing.
>> >> +      </dd>
>> >>      </dl>
>> >>
>> >>      <p>
>> >> diff --git a/controller/ovn-controller.c b/controller/ovn-controller.c
>> >> index 89f10a6..5ab5a13 100644
>> >> --- a/controller/ovn-controller.c
>> >> +++ b/controller/ovn-controller.c
>> >> @@ -2568,6 +2568,15 @@ main(int argc, char *argv[])
>> >>                      sbrec_chassis_private_set_nb_cfg(chassis_private,
>> cur_cfg);
>> >>                      sbrec_chassis_private_set_nb_cfg_timestamp(
>> >>                          chassis_private, time_wall_msec());
>> >> +                    const struct ovsrec_open_vswitch *cfg =
>> >> +                        ovsrec_open_vswitch_first(ovs_idl_loop.idl);
>> >> +                    if (cfg) {
>> >> +                        int delay_cfg_report =
>> smap_get_int(&cfg->external_ids,
>> >> +
>> "ovn-delay-nb-cfg-report", 0);
>> >> +                        if (delay_cfg_report) {
>> >> +                            xsleep(delay_cfg_report);
>> >> +                        }
>> >> +                    }
>> >
>> >
>> > Hi Han,
>> >
>> > Maybe we can also sync this option from the ovs config to chassis
>> other_config in south db ?
>> >
>> > What do you think ?
>> >
>>
>> Hi Numan,
>>
>> I am not sure what's the purpose of syncing this to SB? It is used only by
>> ovn-controller, something similar to ovn-remote-probe-interval, etc.
>>
>
> We do copy some of the configs from ovs db to chassis other_config. But I
> agree this config doesnt need to go to sb and probably CMS wouldn't care
> much know about it.
>
> Thanks
> Numan
>
>
>> Thanks,
>> Han
>>
>> > Thanks
>> > Numan
>> >
>> >>
>> >>                  }
>> >>              }
>> >>
>> >> --
>> >> 2.1.0
>> >>
>> >> _______________________________________________
>> >> dev mailing list
>> >> d...@openvswitch.org
>> >> https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>> >>
>> _______________________________________________
>> dev mailing list
>> d...@openvswitch.org
>> https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>>
>>
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to