[][src]Crate background_jobs

Background Jobs

This crate provides tooling required to run some processes asynchronously from a usually synchronous application. The standard example of this is Web Services, where certain things need to be processed, but processing them while a user is waiting for their browser to respond might not be the best experience.


Add Background Jobs to your project

background-jobs = "0.3"
failure = "0.1"
futures = "0.1"
tokio = "0.1"

To get started with Background Jobs, first you should define a job.

Jobs are a combination of the data required to perform an operation, and the logic of that operation. They implment the Job, serde::Serialize, and serde::DeserializeOwned.

This example is not tested
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct MyJob {
    some_usize: usize,
    other_usize: usize,

impl MyJob {
    pub fn new(some_usize: usize, other_usize: usize) -> Self {
        MyJob {

impl Job for MyJob {
    fn run(self, _: ()) -> Box<dyn Future<Item = (), Error = Error> + Send> {
        info!("args: {:?}", self);


The run method for a job takes an additional argument, which is the state the job expects to use. The state for all jobs defined in an application must be the same. By default, the state is an empty tuple, but it's likely you'll want to pass in some Actix address, or something else.

Let's re-define the job to care about some application state.

This example is not tested
#[derive(Clone, Debug)]
pub struct MyState {
    pub app_name: String,

impl Job<MyState> for MyJob {
    fn run(self, state: MyState) -> Box<dyn Future<Item = (), Error = Error> + Send> {
        info!("{}: args, {:?}", state.app_name, self);


Next, define a Processor.

Processors are types that define default attributes for jobs, as well as containing some logic used internally to perform the job. Processors must implement Proccessor and Clone.

This example is not tested
#[derive(Clone, Debug)]
pub struct MyProcessor;

impl Processor<MyState> for MyProcessor {
    // The kind of job this processor should execute
    type Job = MyJob;

    // The name of the processor. It is super important that each processor has a unique name,
    // because otherwise one processor will overwrite another processor when they're being
    // registered.
    const NAME: &'static str = "IncrementProcessor";

    // The queue that this processor belongs to
    // Workers have the option to subscribe to specific queues, so this is important to
    // determine which worker will call the processor
    // Jobs can optionally override the queue they're spawned on
    const QUEUE: &'static str = "default";

    // The number of times background-jobs should try to retry a job before giving up
    // Jobs can optionally override this value
    const MAX_RETRIES: MaxRetries = MaxRetries::Count(1);

    // The logic to determine how often to retry this job if it fails
    // Jobs can optionally override this value
    const BACKOFF_STRATEGY: Backoff = Backoff::Exponential(2);

Running jobs

By default, this crate ships with the background-jobs-server feature enabled. This uses the background-jobs-server crate to spin up a Server and Workers, and provides a mechanism for spawning new jobs.

background-jobs-server uses LMDB to keep track of local state. LMDB is a memory-mapped storage mechanism, so the jobs information it keeps track of is all stored locally on-disk. In the future, the storage mechanism may be made generic so implementors can bring their own storage.

background-jobs-server also uses ZeroMQ to transfer data between the spawner, server, and workers. If you plan to run two or more of these pieces from the same process, look at the documentation for the methods new_with_context and init_with_context. It is important that ZeroMQ contexts are shared when possible to avoid spinning up multiple ZeroMQ instances for the same application.

With that out of the way, back to the examples:

Starting the job server
This example is not tested
use background_jobs::ServerConfig;
use failure::Error;
use server_jobs_example::queue_set;

fn main() -> Result<(), Error> {
    // Run our job server

Starting the job worker
This example is not tested
use background_jobs::WorkerConfig;
use failure::Error;
use server_jobs_example::{queue_map, MyProcessor};

fn main() -> Result<(), Error> {
    // Create the worker config
    let mut worker = WorkerConfig::new(
        MyState {
            app_name: "My Example Application".to_owned(),

    // Register our processor

    // Spin up the workers

Queuing jobs
This example is not tested
use background_jobs::SpawnerConfig;
use futures::{future::lazy, Future};
use server_jobs_example::{MyJob, MyProcessor};

fn main() {
    // Create 50 new jobs, each with two consecutive values of the fibonacci sequence
    let (_, _, jobs) = (1..50).fold((0, 1, Vec::new()), |(x, y, mut acc), _| {
        acc.push(MyJob::new(x, y));

        (y, x + y, acc)

    // Create the spawner
    let spawner = SpawnerConfig::new("localhost", 5555);

    // Queue each job
    tokio::run(lazy(move || {
        for job in jobs {
            tokio::spawn(spawner.queue::<MyProcessor>(job).map_err(|_| ()));

Complete Example

For the complete example project, see the examples folder

Using on Windows

background-jobs-server depends by default on tokio-zmq, which only works on unix (and unix-like) systems. This might mean it works on the Windows Subsystem for Linux, but it's untested and hard to say. You can override this behavior by specifying the following in your Cargo.toml

version = "0.1"
default-features = false
features = ["background-jobs-server", "background-jobs-server/futures-zmq"]

futures-zmq Is designed to be a drop-in replacement for tokio-zmq that works on non-unix and non-tokio platforms. The reason why it isn't enabled by default is that it's slower than tokio-zmq, and in all likelihood, the production environment for projects depending on this one will be linux.

Not using a ZeroMQ+LMDB based client/server model

If you want to create your own jobs processor based on this idea, you can depend on the background-jobs-core crate, which provides the LMDB storage, Processor and Job traits, as well as some other useful types for implementing a jobs processor.



The entry point for creating a background-jobs server


SpawnerConfig is the only part of this library required to actually exist in your application.


The entry point for creating a background-jobs worker.





The Job trait defines parameters pertaining to an instance of background job


The Processor trait


The SyncJob trait defines parameters pertaining to a synchronous instance of background job