In the composer.yaml file, we have the definition for the backend service. The healthcheck still needs to be implemented.

# todo: implement healthcheck
#healthcheck:
#  test:
#    [
#      "CMD-SHELL",
#      "cargo run healthcheck"
#    ]
#  interval: 10s
#  retries: 5
#  start_period: 10s
#  timeout: 3s

For that reason, workflows take longer than they should. In the script to run E2E tests, we wait 20 seconds instead of waiting only until the container is healthy.

wait_for_container_to_be_healthy torrust-mysql-1 10 3
# todo: implement healthchecks for tracker and backend and wait until they are healthy
#wait_for_container torrust-tracker-1 10 3
#wait_for_container torrust-idx-back-1 10 3
sleep 20s

See: https://github.com/torrust/torrust-index-backend/blob/develop/docker/bin/run-e2e-tests.sh#L47-L51

Proposal 1

Add a new endpoint tho the API: GET /status that should return a 200 code is the app is ready to process requests. It could contains a json body with more verbose report like:

{ 
  "status": "Ok",
  "database": "Ok",
  "tracker": "Ok"
}

When something is wrong could return something like:

{ 
  "status": "Error",
  "error": "Cannot connect to database",
}

We can define this later. For now we just check that return a 200 code.

The healthcheck could be something like

healthcheck:
  test: ["CMD", "curl", "-f", "http://localhost:3030/health"]
  interval: 30s
  timeout: 10s
  retries: 3

Links

• Docker Healthchecks: Why Not To Use curl or iwr
• https://stackoverflow.com/a/68822241/3012842