Current Kourier gateway has uses healthcheck/fail before stop:

          lifecycle:
            preStop:
              exec:
                command: ["/bin/sh","-c","curl -X POST --unix /tmp/envoy.admin http://localhost/healthcheck/fail; sleep 15"]

However, there is another option "drain_listeners", which is used by Istio. We should evaluate if it improves the shutdown behavior or not. Especially we want to make sure that longer-running requests will not be ungracefully terminated during scaling down the gateways.

• Slack thread: https://redhat-internal.slack.com/archives/CF5ANN61F/p1697632968730869
• Upstream issue: https://github.com/knative-extensions/net-kourier/issues/1118