Shrinking pool size when not in heavy use

[nyurik]

As reported by user of the Martin tile server, the number of connections to their Postgres database keeps growing until it reaches the pool size, and it never shrinks back, even when the service is mostly not in use. Is this the expected behavior of the deadpool, or is there something deadpool users should do to handle it? Apparently they tried to set keepalives=0, but that didn't change anything. Thx!

[bikeshedder]

This is expected as there is no active background process in deadpool.

This is related to those two issues:

• Ahead of time object creation/recycling #167
• Add min_size configuration option #245

At the moment deadpool just cares about the max_size and creates objects as needed. If you need to remove objects from the pool after a given time you can implement your own scheduler which removes objects via the Pool::retain method.

I should add the example from #178 (comment) somewhere in the docs:

let interval = Duration::from_secs(30);
let max_age = Duration::from_secs(60);
tokio::spawn(async move {
    loop {
        tokio::time::sleep(interval).await;
        pool.retain(|_, metrics| metrics.last_used() < max_age);
    }
});

I believe every connection pool is different and baking any kind of behavior into the pool is bound to be wrong for one project or another. So I provide various hooks and introspection methods to roll your own.

I haven't spent much time on this, yet, but I'd like to offer a plugin interface in the future which allows easy addition of such features without changing the core implementation of the pool.