All three nodes are exactly identical. I am able to take the SSH of the VM via floating IP attached to it but the reverse traffic is getting out with the SNAT IP of the router when I put the VM on the gateway chassis. Ammad On Tue, Oct 12, 2021 at 3:26 PM Daniel Alvarez <dalvarez@redhat.com> wrote:

Hi Ammad

On 12 Oct 2021, at 10:33, Ammad Syed <syedammad83@gmail.com> wrote:

 Hi,

I am using openstack wallaby. I have seen an issue not sure if its a bug or configuration related issue. I am using ml2/ovn backend with distributed floating IP enabled. I have made my compute node 1 as a gateway chassis where the routers are scheduled.

I have then created an instance and NATed a public IP. The instance deployed on compute 2. When I see the IP address via curl ipinfo.io it shows the floating IP that I have NATed.

Then I migrated the instance to compute node 1. I had many ping drops for a couple of seconds then its back to normal. I have then seen the IP address via curl ipinfo.io. It showed me the SNAT IP address of router.

Could it be that the compute 1 is not properly configured with a connection on the public network? Provider bridge, correct bridge mappings and so on and then the traffic falls back to centralized?

Then I migrated the instance back to compute node 2, I had ping drops for 20 seconds and then the instance came back. I have seen the IP via curl, it showed the floating IP that I have nated with instance.

Is it the expected behavior ?

Ammad

-- Regards, Syed Ammad Ali