Thanks for the apt response. Can you help me further -
We have HPA on custom metric for our http service which scales based on number of http requests and latency. Metrics we use to create custom metric are as follows -
It scales up perfectly when load on http endpoint increases but problem arises when all of the requests are processed successfully. The number of pods doesnt come down to min pod configurations. After debugging we got to know that there is no value in
loadbalancing.googleapis.com/https/backend_request_count which leads to failure in sending data to the custom metric timeseries.
Any idea what's being missed here? TIA!