Environment:

• Kubernetes distribution (e.g. AKS, ACS Engine, GKE, EKS, etc.): Oracle Kubernetes Engine
• Kubernetes version (use kubectl version): v1.10.11
• Osiris version (specify by SHA or semver): master branch
• Install method (specify exact helm install command used): helm

What happened?
When zero scaler is terminating pods after reaching metrics interval threshold. During termination process, if a new request comes in simultaneously. We are running into 502 errors because either it is possible the traffic is reaching the pods being terminated or PODs being activated is taking time to bring up the containers.
What you expected to happen?
The request should never fail even when zero scaler is terminating pods i.e. there should be some retry or timeout which can help to recover the 502 errors
How to reproduce it (as minimally and precisely as possible):
This is happening with nginx ingress controller lbr endpoint when we configure domain name using annotation in service yaml.
Have a nginx ingress controller based lbr with ingress configured for service end point.
Wait for zero scaler PODs termination to kick off after metrics interval threshold is reached. When the pods are being terminated submit a request from lbr you will see 502 errros. It is easily reproducible.
Anything else that we need to know?
Are there any kubernetes ingress level annotations which can avoid 502 errors and still be able to retry and timeout for this scenario?