prevent failing requests for app during deployment rollout

[n1ru4l]

Sometimes it happens that the app is not reachable for a short time period during a deployment. This should not happen.

• Requests should be routed to the old pod, while new pods getting ready
• Once new pods are ready, new requests should be routed there
• Once the old pods have no open connections (or after some grace period); the old pods should be killed