😱 Status quo stories: Alan finds dropping database handles is hard.

🚧 Warning: Draft status 🚧

This is a draft "status quo" story submitted as part of the brainstorming period. It is derived from real-life experiences of actual Rust users and is meant to reflect some of the challenges that Async Rust programmers face today.

If you would like to expand on this story, or adjust the answers to the FAQ, feel free to open a PR making edits (but keep in mind that, as they reflect peoples' experiences, status quo stories cannot be wrong, only inaccurate). Alternatively, you may wish to add your own status quo story!

The problem

Alan has been adding an extension to YouBuy that launches a singleton actor which interacts with a Sqlite database using the sqlx crate. The Sqlite database only permits a single active connection at a time, but this is not a problem, because the actor is a singleton, and so there only should be one at a time. He consults the documentation for sqlx and comes up with the following code to create a connection and do the query he needs:

use sqlx::Connection;

#[async_std::main]
async fn main() -> Result<(), sqlx::Error> {
    // Create a connection

    let conn = SqliteConnection::connect("sqlite::memory:").await?;

    // Make a simple query to return the given parameter
    let row: (i64,) = sqlx::query_as("SELECT $1")
        .bind(150_i64)
        .fetch_one(&conn).await?;

    assert_eq!(row.0, 150);

    Ok(())
}

Things seem to be working fairly well but sometimes when he refreshes the page he encounters a panic with the message "Cannot open a new connection: connection is already open". He is flummoxed.

Searching for the Solution

Alan tries to figure out what happened from the logs, but the only information he sees is that a new connection has been received. Alan turns to the documentation for the sqlx crate to see if there are flags that might enable extra instrumentation but he can't find any.

He's a bit confused, because he's accustomed to having things generally be cleaned up automatically when they get dropped (for example, dropping a File will close it). Searching the docs, he sees the close method, but the comments confirm that he shouldn't have to call it explicitly: "This method is not required for safe and consistent operation. However, it is recommended to call it instead of letting a connection drop as the database backend will be faster at cleaning up resources." Still, just in case, he decides to add a call to close into his code. It does seem to help some, but he is still able to reproduce the problem if he refreshes often enough. Feeling confused, he adds a log statement right before calling close to see if it is working:


#![allow(unused)]
fn main() {
use sqlx::Connection;

#[async_std::main]
async fn do_the_thing() -> Result<(), sqlx::Error> {
    // Create a connection
    let conn = SqliteConnection::connect("sqlite::memory:").await?;

    // Make a simple query to return the given parameter
    let row: (i64,) = sqlx::query_as("SELECT $1")
        .bind(150_i64)
        .fetch_one(&conn).await?; // <----- if this await is cancelled, doesn't help

    assert_eq!(row.0, 150);
    
    // he adds this:
    log!("closing the connection");
    conn.close();

    Ok(())
}
}

He observes that in the cases where he has the problem the log statement never executes. He asks Barbara for help and she points him to this gist that explains how await can be canceled, and cancellation will invoke the destructors for things that are in scope. He reads the source for the SqliteConnection destructor and finds that destructor spawns a task to actually close the connection.

He realizes there is a race condition and the task may not have actually closed the connection before do_the_thing is called a second time. At this point, he is feeling pretty frustrated!

Next, Alan seeks verification and validation of his understanding of the source code from the sqlx forum. Someone on the forum explains why the destructor launches a fresh task: Rust doesn't have a way to execute async operations in a destructor.

Finding the Solution

Alan briefly considers rearchitecting his application in more extreme ways to retain use of async, but he gives up and seeks a more straight forward solution. He discovers rusqlite, a synchronous database library and adopts it. This requires some rearchitecting but solves the problem.

🤔 Frequently Asked Questions

What are the morals of the story?

  • Rust's async story is lacking a way of executing async operations in destructors. Spawning is a workaround, but it can have unexpected side-effects.
  • The story demonstrates solid research steps that Alan uses to understand and resolve his problem.
  • Completion of the Cancellation and timeouts docs may have been helpful. It's difficult to know how something absent might have improved the solution search process.

What are the sources for this story?

This specific story describes an actual bug encountered by Sergey Galich at 1Password.

Why did you choose Alan to tell this story?

His experience and understanding of other languages coupled with his desire to apply Rust would likely lead him to try solutions before deeply researching them.

How would this story have played out differently for the other characters?

This story would likely have played out the same for everyone.