With NSX-T plugin integration, app container network not available after restart

Products

• VMware Tanzu Application Service for VMs

Resolution

Checklist:

With NSX-T Plugin integration, after the app restarts, the app container becomes un-accessible from the outside, both HTTP and SSH cannot reach the container though it's reported to be healthy.

When starting / restarting / stopping an app container, NSX-T Plugin watches over the event on Diego BBS and makes the decision to delete the network for stopped containers and create a network for new containers. 

1. NCP receives a desired_lrp_created event for the new LRP, so it starts creating a network for the new LRP.
2. NCP receives a desired_lrp_removed event for the old LRP, so it starts removing a network for the old LRP.
3. NCP receives a actual_lrp_changed event that the new LRP is ready, it double checks actual-lrp and desired-lrp to synchronize the network status with Diego BBS. 

One issue was identified with NCP + Tanzu Application Service integration at step 3. Due to this issue, NCP makes improper decisions and deletes the network with the new LRP. As a result, the app instance is reported healthy on the TAS side, but no connection could be made to it. 

This issue has been fixed in the NCP v3.1.1 release.