Description of problem: Customer who updated from 16.1.x to 16.2.3 is facing a problem for pre-existing octavia's LB (amphora driver). Both vms (lb clients) and lb are using an external vlan network Basically, after the upgrade to 16.2.3, customer founds that clients were not able to reach existing lbs vip as the arp table was point to the vip's port mac address instead of the master's amphora vrrp port hw addr. Failovering the loadbalancer is useful to temporary fix the issue, but after a while the issue comes back. As i wrote all is running on top of a provider network, and virtual machines outside openstack connected to the same network never faced the issue. New Loadbalancers created after the upgrade seems working as expected. Version-Release number of selected component (if applicable): Openstack 16.2.3 OVN with DVR How reproducible: failover a lb and wait Steps to Reproduce: 1. failover a lb, vms will start reaching out vip correctly 2. wait some time 3. clients will not be able anymore to reach the vip, as the arp table will be populated with the vip's port hw address Actual results: openstack vms arp table is pupulated with vip port's mac addr. Expected results: openstack vms arp table to be pupulated with master amphora vrrp port's mac addr.
Hi, Thank you for the information. This issue is a duplicate of BZ 2126055. You can attach your customer case there. The issue will be fixed in 16.2z4, which will get released on Dec 7th. We will close this BZ as a duplicate.
*** This bug has been marked as a duplicate of bug 2126055 ***