Single SQLx Postgres pool connection OR multiple pool connections?

[iodapson]

Should I use just one PgPool / PgPoolOptions` connection to my Postgres across my entire application like this:

use anyhow::{Context, Result};
use sqlx::{postgres::PgPoolOptions, Pool, Postgres};
use std::sync::Arc;

//struct DatabaseConnectionPool(Arc<Pool<Postgres>>);

pub async fn run_database(db_url: String) -> Result<Arc<Pool<Postgres>>> {
    // create a PgPoolOptions connection to DATABASE_URL
    let db_pool_options = PgPoolOptions::new()
        .max_connections(50)
        .connect(&db_url.as_str())
        .await
        .with_context(|| "Connection to Database failed")?;

    // Correctly embed into this application database migrations
    // ...performed by command `$ sqlx db setup`
    // ...You can use this command each time you add to dir migrations
    sqlx::migrate!()
        .run(&db_pool_options)
        .await
        .with_context(|| "SQLx DB migration failed")?;

    Ok(Arc::new(db_pool_options))
}

Or create multiple PgPool or PgPoolOptions connections inside each handler function that needs to CRUD to database?

Both approaches work, but I really want to follow the one that is of best practice. Thanks in advance.

[davidpdrsn]

Create one PgPool and share that.

[iodapson]

Okay, I will. Thank you very much!

[jplatte]

By the way, there is no point in putting PgPool from sqlx inside an Arc as it is already a cheap to clone reference counted type (probably just Arc internally but I'm not entirely certain).