Workers
Loco provides the following options for background jobs:
- Redis backed (powered by
sidekiq-rs
) - Postgres backed (own implementation)
- Tokio-async based (same-process, evented thread based background jobs)
You enqueue and perform jobs without knowledge of the actual background queue implementation, similar to Rails' ActiveJob, so you can switch with a simple change of configuration and no code change.
Async vs Queue
When you generated a new app, you might have selected the default async
configuration for workers. This means workers spin off jobs in Tokio's async pool, which gives you proper background processes in the same running server.
You might want to configure jobs to run in a separate process backed by a queue, in order to distribute the load across servers.
First, switch to BackgroundQueue
:
# Worker Configuration
workers:
# specifies the worker mode. Options:
# - BackgroundQueue - Workers operate asynchronously in the background, processing queued.
# - ForegroundBlocking - Workers operate in the foreground and block until tasks are completed.
# - BackgroundAsync - Workers operate asynchronously in the background, processing tasks with async capabilities.
mode: BackgroundQueue
Then, configure a Redis based queue backend:
queue:
kind: Redis
# Redis connection URI
uri: ""
dangerously_flush: false
Or a Postgres based queue backend:
queue:
kind: Postgres
# Redis connection URI
uri: ""
dangerously_flush: false
Running the worker process
You can run in two ways, depending on which setting you chose for background workers:
Usage: demo_app start [OPTIONS]
Options:
-w, --worker start worker
-s, --server-and-worker start same-process server and worker
Choose --worker
when you configured a real Redis queue and you want a process for doing just background jobs. You can use a single process per server. In this case, you can run your main Web or API server using just cargo loco start
.
Choose -s
when you configured async
background workers, and jobs will execute as part of the current running server process.
For example, running --server-and-worker
:
Creating background jobs in code
To use a worker, we mainly think about adding a job to the queue, so you use
the worker and perform later:
// .. in your controller ..
perform_later
.await
Unlike Rails and Ruby, with Rust you can enjoy strongly typed job arguments which gets serialized and pushed into the queue.
Using shared state from a worker
See How to have global state, but generally you use a single shared state by using something like lazy_static
and then simply refer to it from the worker.
If this state can be serializable, strongly prefer to pass it through the WorkerArgs
.
Creating a new worker
Adding a worker meaning coding the background job logic to take the arguments and perform a job. We also need to let loco
know about it and register it into the global job processor.
Add a worker to workers/
:
And register it in app.rs
:
Generate a Worker
To automatically add a worker using loco generate
, execute the following command:
The worker generator creates a worker file associated with your app and generates a test template file, enabling you to verify your worker.
Configuring Workers
In your config/<environment>.yaml
you can specify the worker mode. BackgroundAsync and BackgroundQueue will process jobs in a non-blocking manner, while ForegroundBlocking will process jobs in a blocking manner.
The main difference between BackgroundAsync and BackgroundQueue is that the latter will use Redis to store the jobs, while the former does not require Redis and will use async within the same process.
# Worker Configuration
workers:
# specifies the worker mode. Options:
# - BackgroundQueue - Workers operate asynchronously in the background, processing queued.
# - ForegroundBlocking - Workers operate in the foreground and block until tasks are completed.
# - BackgroundAsync - Workers operate asynchronously in the background, processing tasks with async capabilities.
mode: BackgroundQueue
Testing a Worker
You can easily test your worker background jobs using Loco
. Ensure that your worker is set to the ForegroundBlocking
mode, which blocks the job, ensuring it runs synchronously. When testing the worker, the test will wait until your worker is completed, allowing you to verify if the worker accomplished its intended tasks.
It's recommended to implement tests in the tests/workers
directory to consolidate all your worker tests in one place.
Additionally, you can leverage the worker generator, which automatically creates tests, saving you time on configuring tests in the library.
Here's an example of how the test should be structured:
async