compose health check with higher rate during start

Description

When checking the health_check to create a dependency on container start you want to have a high rate check (like every second) to start the service as soon as possible. But once the containers are started you don't want to overload your service with health_checks.

Without this proposal, you need to wait interval time to discover your container is healthy and start the dependent containers even if the service was healthy a few seconds after being up.

I propose to have a new parameter like start_interval to override the interval period during the start_period. Let take this config exemple:

  db:
    health check:
	test: ...
	   interval: 30s
	   timeout: 3s
           retry:2
	   start_period: 20s
           start_interval: 1s

This will perform a health check on every 1s during the start_period (from start to healthy stat, retry limit is ignored), then perform health check on every 30s with timeout & retry applied.

This is for reducing the downtime during stack restart and avoid to overload the health test after being started.

Apr 12 '23 09:04 disk91

This sounds like a nice improvement! Compose's health check support is actually a wrapper around the functionality in Docker Engine, so this would need to exist there before Compose can support it.

I would open a feature request at https://github.com/moby/moby - if/when it's included in an Engine release, we can look at extending <svc>.healthcheck in YAML accordingly.