We were able to reproduce the issue when using virtual MAC with port
security enabled.

After adding the <VIP, virtual MAC> address pair to the openstack ports
of the keepalived units and restarting the units a test instance was
able to ping the VIP correctly, resolving the VIP to the virtual MAC.

A keepalived failover was triggered by isolating the master unit, shutting down 
its openstack port, but the traffic was sent to the wrong unit by OVN.
- The master M1 unit is isolated.
- The backup unit M2 becomes the master of the cluster and sends the gARP with 
the virtual MAC.
- OVN updates the MAC-chassis binding, selecting as the egress chassis for the 
virtual MAC the chassis of unit M3, the one with the lowest priority. This is 
unexpected, as this unit was never elected as master of the keepalived cluster 
and never sent any gARP request.
- OVN sends the traffic to M3, where the VIP is not configured.
- Service is unavailable.

We were not able to replicate the issue with port security disabled,
after disabling port security and triggering a VIP movement the traffic
flows to the correct master unit.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2166984

Title:
  OVN Mac_Binding table not refreshed after keepalived VIP failover
  between virtual port parents

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/ovn/+bug/2166984/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to