Connecting to unassigned floating IP port causing 3-5 minute downtime in FIP subnet

Issue

• We discovered an issue where in a new domain created with new external networks assigned to the new domains, if we follow the below steps:

1. Create a FIP-1 in Floating Subnet A
2. Create a port-1 in Internal Subnet B
3. Associate FIP-1 with port-1 in Internal Subnet B
4. Create FIP-2 in Internal Subnet A
5. Create port-2 in Internal Subnet B
6. Associate FIP-2 with port-2 in Internal Subnet B
7. Assign port-2 to an instance-1
8. SSH to instance-1
9. telnet from instance-1 to FIP-1 in Floating Subnet A
10. Repeat step 9 a few times until you lose connectivity to instance-1
11. Notice that for 3-5 minutes no IP's in Floating Subnet A are accessible from outside of OpenStack

• Also we found that in step 9 if we telnet FIP-1 from outside OpenStack the issue in step 11 does not happen. We also only appear to have this issue when the environment is deployed using Terraform, and only in the networks assigned to a particular new domain and its projects. If we:

• Use the same terraform code to deploy the same system in another domain/project/network the issue does not occur

• Deploy a similar environment to what is being deployed by Terraform but rather using OpenStack tools in the new domain/project/network, then the issue does not occur