Hi Soheil, I can't say I've seen this exact issue, but looks to be something with the OVS flow table based on your description. The best thing to do is file a bug [0] and include as much info as possible so someone can take a look. Thanks, -Brian [0] https://bugs.launchpad.net/neutron/+filebug On 12/6/25 8:49 AM, soheil.bakan@gmail.com wrote:
Hi OpenStack Community,
I am experiencing an issue in a 3-controller HA OpenStack environment related to Neutron L3 HA and Open vSwitch. I would appreciate your guidance or any insights.
### Environment - OpenStack version: 2025.1 - 3 controller nodes with HA setup (kolla-ansible) - Neutron L3 HA enabed - OVS + Keepalived (VRRP) - The router’s VRRP master role is currently on one controller - All controllers share the same external/provider network setup
### Issue Description When I restart the `openvswitch-vswitchd` service on the controller node that currently holds the VRRP Master for an HA router: The following problems occur:
1. ICMP traffic (ping) drops for several minutes. 2. The router’s default route begins to bounce between the other two or three controllers. 3. The VRRP master does not stabilize quickly; it keeps flipping between the remaining controllers. 4. As a result, I see packet loss and routing instability for several seconds.
### What I expect Ideally, when `openvswitch-vswitchd` restarts on the master controller: - VRRP should quickly fail over to one of the standby controllers routers - The default route should converge and remain stable - There should be no (or minimal) packet loss
I should also say what when i restart neutron_l3_agent or in some cases openvswitch_db issue goes away
### Additional Information If needed, I can share: - Logs from OVS, neutron-l3-agent
Thanks in advance for your support. Please let me know if more information is needed.
best regards, soheil