Shorten nginx keepalive timeout
To help address the issues discovered in #1358 (closed) where we are sending traffic to pods long after they are terminated, we think shortening the nginx keepalive timeout will help shorten the time-window which needs to be shorter than
blackoutSeconds, currently set to 240 seconds. During the blackout window, puma can still respond successfully, but after that window expires requests will respond with a 502.
This can be set in the nginx configmap
data: keep-alive-requests: "xxx" upstream-keepalive-requests: "xxx"