#task #channel #worker #resilient #reusable #programs #long-lived

sisyphus-tasks

Reusable utils for long-lived, worker-oriented rust programs

2 releases

0.1.1 Dec 10, 2023
0.1.0 Dec 7, 2023

#2 in #resilient

31 downloads per month

Apache-2.0 OR MIT

58KB
951 lines

Sisyphus Tasks

Utilities for long-running, resilient tasks.

This library contains code I wrote, found useful, and want to keep using. It aims to provide systems

The general idiom is focused on spawning long-lived worker loops, which use channels to communicate.

Understanding this crate

High Level Example

pub struct MyWorker{
  pipe: Pipe<WorkItem>,
}

impl Boulder for MyWorker {
  fn spawn(self) -> JoinHandle<Fall<Self>> {
    tokio::spawn(async move {
      // pipe gives refs to items
      while let Some(item) = self.pipe.next().await {
        // worker does work on each item as it becomes available
        self.do_work_on(item);
        // do some async work too :)
        self.async_work_as_well(item).await;
      }
    });
  }
}

let task = MyWorker::new(a_pipe).run_forever();

Current Utils:

  • Sisyphus

    • A scaffolding system for spawning long-lived, recoverable tasks
    • A Boulder is a looping, fallible task
    • Boulder logic is defined in a Boulder::spawn() method that returns a JoinHandle
    • A Fall is an error in that task.
      • Fall::Recoverable - errors that the task believes it can recover
      • Fall::Unrecoverable - errors that the task believes it cannot recover
    • The Boulder::run_until_panic method handles restarting on recoverable Fall s, and reporting unrecoverable errors
    • Boulder s may define custom recovery or cleanup logic
    • Boulder s may also panic. In that case, no Fall is generated, and the
      • panic is propagated upward
    • Sisyphus manages a Boulder loop. He exposes an interface to observe its status and abort the work
  • Pipe

    • An inbound and an outbound channel
    • Enforce process-once semantics
    • Prevents data loss on worker error
    • Designed for relatively linear data-processing pipelines
      • e.g. retrieval -> metrics -> indexing -> other handling
    • Convenience methods for running synchronous and asynchronous for_each on channel contents

Future Utils:

  • Abstraction layers for instantiating complex pipes from lists of Sisyphuses
  • Pipes should allow sync & async transforms (inbound T, outbound U)

Some code descends from utilities written for Nomad. It is used and reproduced under its license terms.

Dependencies

~11–24MB
~323K SLoC