Hi Lorenzo, On 11/11/24 15:38, Lorenzo Bianconi wrote: > [...] >>> +AT_CHECK([ovs-appctl dpctl/dump-conntrack | FORMAT_CT(fd13::a) | \ >>> +sed -e 's/zone=[[0-9]]*/zone=<cleared>/' | >>> +sed -e 's/mark=[[0-9]]*/mark=<cleared>/' | sort], [0], [dnl >>> +]) >>> + >>> +OVS_APP_EXIT_AND_WAIT([ovn-controller]) >>> + >>> +as ovn-sb >>> +OVS_APP_EXIT_AND_WAIT([ovsdb-server]) >>> + >>> +as ovn-nb >>> +OVS_APP_EXIT_AND_WAIT([ovsdb-server]) >>> + >>> +as northd >>> +OVS_APP_EXIT_AND_WAIT([ovn-northd]) >>> + >>> +as >>> +OVS_TRAFFIC_VSWITCHD_STOP(["/.*error receiving.*/d >>> +/.*terminating with signal 15.*/d"]) >>> +AT_CLEANUP >>> +]) >> >> We don't have any test for the case when two OVN routers have routes >> configured via the same next-hop IP. We should check what happens when >> we remove one of the routes. > > so far when we have a given next-hop mac shared by multiple ct entries > we flush it just when we remove the last reference to this mac address > (so we do not introduce any unexpected flush). Does it work for you? >
If we have two independent routers that happen to use the same next-hop IP (and mac) when removing the route from one of them we should flush all conntrack entries that correspond to the next hop on that router (but not on the other one). We don't cover that case AFAIU. Thanks, Dumitru _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
