Copilot-cli: Support HealthCheckGracePeriodSeconds for Load Balanced Web Services

[mikelhamer]

In my manifest I have defined image.healthcheck.start_period with a value of 120. But when I look in my ECS service in the details tab it shows this

Is this the same setting, or maybe it's just not working?

[efekarakus]

Hi @mikelhamer ! Can you try with 120s as an input? (adding the unit at the end)

[mikelhamer]

Hey @efekarakus thanks for the suggestion! Unfortunately this didn't seem to have any effect.

[efekarakus]

Hi @mikelhamer, I think this is happening due to a confusion around which healthcheck we are talking about 😄

What's happening now?

The image healthcheck field configures the container healthcheck. For example:

image:
  build: frontend/Dockerfile
  healthcheck:
    command: ["CMD-SHELL", "curl -f http://localhost:8080 || exit 1"]
    start_period: 120s
  port: 8080

And I can see the container healtcheck being configured accordingly:

From navigating to the task definition:

What do we want?

I believe what you are looking for is configuring HealthCheckGracePeriodSeconds in the ECS service which we don't surface at the moment 🙇

This would be a feature request for us where we add a start_period field under the http section where the default is 60s.

http:
  healthcheck:
     start_period: 120s # new field

[efekarakus]

One way of mitigating this concern I think for now until we surface this field is modifying the application code to have two healthcheck handlers:

1. For the target group healthcheck that always just responds OK
2. For container healthchecks that has the current logic with a start_period of 120s.

Then once this field is supported, the healthchecks can be combined and the container healthcheck can be removed

[mikelhamer]

Thanks as always for the great explanation and possible solution!