Zero downtime with AWS NLB/ALB

[billyshambrook]

Description:

To achieve zero downtime deploys with AWS NLB/ALB, it's required to include a sleep preStop - https://aws.github.io/aws-eks-best-practices/networking/loadbalancing/loadbalancing/#ensure-pods-are-deregistered-from-load-balancers-before-termination

Seems like the best place is to include the sleep in the shutdown manager lifecycle hook, something like this

lifecycle:
  preStop:
    exec:
      command:
      - /bin/sh
      - -c
      - "sleep 180 && envoy-gateway envoy shutdown"

However, this isn't possible as the docker image doesn't include the sleep command.

An alternative would be to include a --delay argument, something like envoy gateway shutdown --delay 180s.