I keep this PR open for the unhealthy state.
See #5372 for enum

Conflicts :/

Don't worry, i'll handle it.

Solved and up to date and ready (to be tested anyway...)

I have podman on my machine and I can't launch the container... I can't test it and I don't have time to debug...

I have podman on my machine and I can't launch the container...

Why may I ask? Do you run into an issue here?

Health checks may need to be adjusted, probably reduce the interval during startup to speed up startup of all containers.

start period provides initialization time for containers that need time to bootstrap. Probe failure during that period will not be counted towards the maximum number of retries. However, if a health check succeeds during the start period, the container is considered started and all consecutive failures will be counted towards the maximum number of retries.

start interval is the time between health checks during the start period. This option requires Docker Engine version 25.0 or later.

See: https://docs.docker.com/reference/dockerfile/#healthcheck

This should allow for better dependency management.
Can be used for helm chart, I guess.

Thanks for the idea! However our health checks are currently built in a way that they never fail after a specific time. See for example https://github.com/nextcloud/all-in-one/blob/main/Containers/apache/healthcheck.sh. for the rest the defaults are good enough imho

Thanks for the idea! However our health checks are currently built in a way that they never fail after a specific time. See for example https://github.com/nextcloud/all-in-one/blob/main/Containers/apache/healthcheck.sh. for the rest the defaults are good enough imho

The problem with doing this is that docker considers the container ready... This should immediately take the container out of the starting state.

This is the only reliable way to check if the container is really up and running.

A fix would be to fail the container if nextcloud:9000 is not reachable and add a startup period (e.g. 10 minutes)

#!/bin/bash
- nc -z "$NEXTCLOUD_HOST" 9000 || exit 0
+ nc -z "$NEXTCLOUD_HOST" 9000 || exit 0
nc -z 127.0.0.1 8000 || exit 1
nc -z 127.0.0.1 "$APACHE_PORT" || exit 1
if ! nc -z "$NC_DOMAIN" 443; then
 echo "Could not reach $NC_DOMAIN on port 443."
 exit 1
fi

- HEALTHCHECK CMD /healthcheck.sh
+ HEALTHCHECK --start-period=10m CMD /healthcheck.sh

A fix would be to fail the container if nextcloud:9000 is not reachable and add a startup period (e.g. 10 minutes)

This is the problem: we cannot spwcify a time here as is depending on the overall setup like installed apps and features, amount of users, given hardware and else especially during upgrades

This is the problem: we cannot spwcify a time here as is depending on the overall setup like installed apps and features, amount of users, given hardware and else especially during upgrades

Yes, I see...

Would it be better to manage dependencies like docker compose?

The solution for detecting the ready state of a service is to use the condition attribute with one of the following options:

• service_started
• service_healthy. This specifies that a dependency is expected to be "healthy", which is defined with healthcheck, before starting a dependent service.

https://docs.docker.com/compose/how-tos/startup-order/

I fear this is going to have the same problem afaics, no?

@docjyJ LGTM now 😊

Can you please rebase the PR and squash the commits?
Afterwards we can merge this :)