[ovs-dev] Question on ovsdb_idl_omit_alert()

[Numan Siddique]

On Wed, Jan 8, 2020 at 1:23 PM Han Zhou <hzhou at ovn.org> wrote:
>
> On Thu, Jan 2, 2020 at 11:51 AM Numan Siddique <numans at ovn.org> wrote:
> >
> > On Fri, Jan 3, 2020 at 1:00 AM Ben Pfaff <blp at ovn.org> wrote:
> > >
> > > On Fri, Jan 03, 2020 at 12:45:06AM +0530, Numan Siddique wrote:
> > > > Hi Ben,
> > > >
> > > > ovn-controller.c has the below code here [1]
> > > >
> > > > ovsdb_idl_omit_alert(ovnsb_idl_loop.idl, &sbrec_chassis_col_nb_cfg);
> > > >
> > > > From what I understand from this comment [2], when ovn-controller
> > > > updates this column, SB ovsdb-server shouldn't send the
> > > > update2/update3 message back to the ovn-controller and to all the
> > > > other ovn-controllers right ?
> > > >
> > > > But I don't see it working as expected. I can see that ovsdb-server
> > > > sends the update2/update3 message to all the connected clients.
> > > >
> > > > Is this a bug ? Or is my understanding wrong ?
> > >
> > > I think that is a misunderstanding.  "Omit alert" does not suppress
> > > messages from ovsdb-server.  Instead, it keeps the ovsdb IDL from
> > > incrementing the sequence number that indicates that something has
> > > changed in the database, that is, it keeps from alerting the IDL's
> > > client that something has changed.
> >
> > Thanks for the reply and clearing the misunderstanding.
> >
> > Numan
> >
>
> Hi Numan,
>
> I had sent a patch to improve this [0], and the last discussion was [1].
>
> Ben, if you remember, you suggested about improving the conditional monitor
> in ovsdb to avoid this problem [1]. However, after checking the conditional
> monitor code I feel the change is not trivial and in fact I didn't find a
> proper way to do it without sacrificing ovsdb performance. So I think the
> simple change in the DB schema [0] would worth it. Please let me know if
> you would like to revisit it.

Thanks Han for sharing this. I guess you have already hit this problem
a long time ago
which we are seeing now in our scale testing.

I am supportive of adding a new chassis private table and I think it's
worth revisiting it.

Thanks
Numan


>
> [0] https://patchwork.ozlabs.org/patch/899608/
> [1]
> https://lists.linuxfoundation.org/pipermail/ovs-dev/2018-April/346206.html
>
> Thanks,
> Han
> _______________________________________________
> dev mailing list
> dev at openvswitch.org
> https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>