Fix container healthcheck

[feelingsonice]

Previous container creation logic had the following 2 issues afaik:

1. Health check was too quick. Increasing this from 50 to 500 fixed the problem for me on M2 Max.
2. Container setup logic didn't do cleanup after failure. Resulted in error messages like Conflict. The container name \"/cornucopia_postgres\" is already in use by container on subsequent startup.

Misc:

• Fix the build script for auto_build example.

This fixes #244

[Gerald-Hu]

Why too quick is a trouble? And do we really want to 10x it?

[feelingsonice]

This health check is pinging the Postgres container as it starts up, because the container needs sometime to start, ping it too quickly means the container won't be ready yet.

[heksesang]

But it retries the pings, why would it be a problem that the container isn't ready yet?

[feelingsonice]

I think the health checker is erroring out before it could do any retry:

fn healthcheck(podman: bool, max_retries: u64, ms_per_retry: u64) -> Result<(), Error> {
    let slow_threshold = 10 + max_retries / 10;
    let mut nb_retries = 0;
    while !is_postgres_healthy(podman)? { // <------ this won't execute for me
       
    }
    // ...
}

I haven't confirmed this but a quick Look suggests that is_postgres_healthy is calling a command on the cornucopia_postgres container before it's even created:

/// Checks if Cornucopia's container reports healthy
fn is_postgres_healthy(podman: bool) -> Result<bool, Error> {
    Ok(cmd(
        podman,
        &["exec", "cornucopia_postgres", "pg_isready"],
        "check container health",
    )
    .is_ok())
}

[heksesang]

But is_postgres_healthy can only return an Ok(true) or Ok(false), so it cannot actually error out?

[Gerald-Hu]

Can barely understand this file and I hate lifecycle lol... Is this a generated file?

[feelingsonice]

Its auto generated.