On 8/14/23 00:31, Han Zhou wrote: > On Fri, Aug 11, 2023 at 11:30 PM Dumitru Ceara <dce...@redhat.com> wrote: >> >> On Saturday, August 12, 2023, Han Zhou <hz...@ovn.org> wrote: >>> >>> >>> >>> On Fri, Aug 11, 2023 at 5:36 AM Dumitru Ceara <dce...@redhat.com> wrote: >>>> >>>> On 8/10/23 17:53, Han Zhou wrote: >>>>> Today the localnet ports are added to local_lports unconditionally. >>>>> This has several side-effects: >>>>> >>>>> - When ovn-monitor-all is true, all localnet ports are added to >>>>> local_lports, including the ones on the non-local logical switches, >>>>> which is undesirable, although there is no direct consequences yet. >>>>> >>>>> - When ovn-monitor-all is false (i.e. conditional monitoring is >>>>> enabled), only localnet ports of local datapaths are supposed to be >>>>> monitored. However, because of the issue discussed at [0], > conditional >>>>> monitoring doesn't take effect initially, and all the localnet > ports >>>>> are monitored and added to local_lports. Later when monitor > condition >>>>> is set for port_bindings, all the localnet ports are added to the >>>>> condition, which causes scale problem e.g. in ovn-k8s environment >>>>> where each node has at least one localnet port, and the SB DB > servers >>>>> are overloaded when handling such huge conditions from every node. >>>>> >>>>> [0] > https://mail.openvswitch.org/pipermail/ovs-dev/2023-July/406201.html >>>>> >>>>> Signed-off-by: Han Zhou <hz...@ovn.org> >>>>> --- >>>> >>>> Looks good to me, thanks! >>>> >>>> Acked-by: Dumitru Ceara <dce...@redhat.com> >>>> >>>> Regards, >>>> Dumitru >>>> >>> >>> Thanks Dumitru. I applied the patch to main. >>> >> >> I didn’t check closely but isn’t this an issue on older branches too. > Don’t we need to backport it there too? > > Yes it is an issue in older branches but it didn't introduce an obvious > user impact until the commit [0]. And now that you have the patches [1] as > the more direct fix for [0], and I assume they will need to be backported, > so I think my patch here doesn't really need a backport. It is more of an > internal improvement. What do you think?
I agree, makes sense. I'll be pushing the [1] patches soon. > > [0] > https://github.com/ovn-org/ovn/commit/1b0dbde940706e5de6e60221be78a278361fa76d > [1] https://patchwork.ozlabs.org/project/ovn/list/?series=368430 > > Best regards, > Han > Regards, Dumitru _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev