On Fri, Dec 15, 2023 at 6:07 PM Lorenzo Bianconi <[email protected]> wrote: > > > On 12/14/23 18:42, Lorenzo Bianconi wrote: > > > When using VLAN backed networks and OVN routers leveraging the > > > 'ovn-chassis-mac-mappings' option for east-west traffic, the eth.src > > > field is > > > replaced by the chassis mac address in order to not expose the router mac > > > address from different nodes and confuse the TOR switch. However doing so > > > the TOR switch is not able to learn the port/mac bindings for routed E/W > > > traffic and it is force to always flood it. Fix this issue adding the > > > capability to configure a given timeout for garp sent by ovn-controller > > > and not disable it after the exponential backoff in order to keep > > > refreshing the entries in TOR swtich fdb table. > > > More into about the issue can be found here [0]. > > > > > > [0] > > > https://mail.openvswitch.org/pipermail/ovs-discuss/2020-September/050678.html > > > Reported-at: https://bugzilla.redhat.com/show_bug.cgi?id=2087779 > > > > > > Signed-off-by: Lorenzo Bianconi <[email protected]> > > > Acked-by: Ales Musil <[email protected]> > > > Signed-off-by: Mark Michelson <[email protected]> > > > --- > > > Changes since v1: > > > - add missing selftest > > > --- > > > > Hi Lorenzo, > > > > Thanks for the backport. > > > > [...] > > > > > diff --git a/tests/ovn.at b/tests/ovn.at > > > index 003bf45d5..987e043df 100644 > > > --- a/tests/ovn.at > > > +++ b/tests/ovn.at > > > @@ -30762,3 +30762,108 @@ check test "$current_id" = "$prev_id" > > > OVN_CLEANUP([hv1],[hv2]) > > > AT_CLEANUP > > > ]) > > > + > > > +OVN_FOR_EACH_NORTHD([ > > > +AT_SETUP([send gratuitous arp for l3gateway only on selected chassis]) > > > > This test is identical to what 77ec310dda51 ("pinctrl.c: Send GARP only > > on chassis atached to l3gw") added to branch 22.03. However, we don't > > have the changes that 77ec310dda51 added to pinctrl.c. > > > > Does that mean this test is flaky and might fail occasionally? Or is it > > just that the test doesn't test the change it was supposed to test? > > > > In any case, why not add a new test, just for garp_max_timeout, for main > > branch? We can backport that one alone to all branches, including 21.12. > > > > Thanks, > > Dumitru > > > > > > Hi Dumitru, >
Hi Lorenzo, > Regarding the self-test for 21.12 backport I posted a standalone test for > garp-max-timeout we can easily backport on 21.12. > > https://patchwork.ozlabs.org/project/ovn/patch/3957de4e1a47075fd8251156c01ec06cad9ba60a.1702659678.git.lorenzo.bianc...@redhat.com/ > Thanks, with that one I also pushed this patch to 21.12. I also had to the bump OVS submodule (cherry picked from 22.03) to make sure our CI is green. Regards, Dumitru _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
