A powerful PostgreSQL-backed job queue for Rust applications, based on Graphile Worker. This is a complete Rust rewrite that offers excellent performance, reliability, and a convenient API.
Graphile Worker RS allows you to run jobs (such as sending emails, performing calculations, generating PDFs) in the background, so your HTTP responses and application code remain fast and responsive. It's ideal for any PostgreSQL-backed Rust application.
Key highlights:
- High performance: Uses PostgreSQL's
SKIP LOCKED
for efficient job fetching - Low latency: Typically under 3ms from task schedule to execution using
LISTEN
/NOTIFY
- Reliable: Automatically retries failed jobs with exponential backoff
- Flexible: Supports scheduled jobs, task queues, cron-like recurring tasks, and more
- Type-safe: Uses Rust's type system to ensure job payloads match their handlers
This port is mostly compatible with the original Graphile Worker, meaning you can run it side by side with the Node.js version. The key differences are:
- No support for batch jobs yet (create an issue if you need this feature)
- In the Node.js version, each process has its own worker_id. In the Rust version, there is only one worker_id, and jobs are processed in your async runtime thread
Add the library to your project:
cargo add graphile_worker
A task consists of a struct that implements the TaskHandler
trait. Each task has:
- A struct with
Serialize
/Deserialize
for the payload - A unique identifier string
- An async
run
method that contains the task's logic
use serde::{Deserialize, Serialize};
use graphile_worker::{WorkerContext, TaskHandler, IntoTaskHandlerResult};
#[derive(Deserialize, Serialize)]
struct SendEmail {
to: String,
subject: String,
body: String,
}
impl TaskHandler for SendEmail {
const IDENTIFIER: &'static str = "send_email";
async fn run(self, _ctx: WorkerContext) -> impl IntoTaskHandlerResult {
println!("Sending email to {} with subject '{}'", self.to, self.subject);
// Email sending logic would go here
Ok::<(), String>(())
}
}
Set up the worker with your configuration options and run it:
#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
// Create a PostgreSQL connection pool
let pg_pool = sqlx::postgres::PgPoolOptions::new()
.max_connections(5)
.connect("postgres://postgres:password@localhost/mydb")
.await?;
// Initialize and run the worker
graphile_worker::WorkerOptions::default()
.concurrency(5) // Process up to 5 jobs concurrently
.schema("graphile_worker") // Use this PostgreSQL schema
.define_job::<SendEmail>() // Register the task handler
.pg_pool(pg_pool) // Provide the database connection
.init() // Initialize the worker
.await?
.run() // Start processing jobs
.await?;
Ok(())
}
Connect to your database and run the following SQL:
SELECT graphile_worker.add_job(
'send_email',
json_build_object(
'to', '[email protected]',
'subject', 'Welcome to our app!',
'body', 'Thanks for signing up.'
)
);
// Get a WorkerUtils instance to manage jobs
let utils = worker.create_utils();
// Type-safe method (recommended):
utils.add_job(
SendEmail {
to: "[email protected]".to_string(),
subject: "Welcome to our app!".to_string(),
body: "Thanks for signing up.".to_string(),
},
Default::default(), // Use default job options
).await?;
// Or use the raw method when type isn't available:
utils.add_raw_job(
"send_email",
serde_json::json!({
"to": "[email protected]",
"subject": "Welcome to our app!",
"body": "Thanks for signing up."
}),
Default::default(),
).await?;
You can provide shared state to all your tasks using extensions:
use serde::{Deserialize, Serialize};
use graphile_worker::{WorkerContext, TaskHandler, IntoTaskHandlerResult};
use std::sync::{Arc, atomic::{AtomicUsize, Ordering::SeqCst}};
// Define your shared state
#[derive(Clone, Debug)]
struct AppState {
db_client: Arc<DatabaseClient>,
api_key: String,
counter: Arc<AtomicUsize>,
}
// Example database client (just for demonstration)
struct DatabaseClient;
impl DatabaseClient {
fn new() -> Self { Self }
async fn find_user(&self, _user_id: &str) -> Result<(), String> { Ok(()) }
}
#[derive(Deserialize, Serialize)]
struct ProcessUserTask {
user_id: String,
}
impl TaskHandler for ProcessUserTask {
const IDENTIFIER: &'static str = "process_user";
async fn run(self, ctx: WorkerContext) -> impl IntoTaskHandlerResult {
// Access the shared state in your task
let app_state = ctx.get_ext::<AppState>().unwrap();
let count = app_state.counter.fetch_add(1, SeqCst);
// Use shared resources
app_state.db_client.find_user(&self.user_id).await?;
println!("Processed user {}, task count: {}", self.user_id, count);
Ok::<(), String>(())
}
}
// Add the extension when configuring the worker
let app_state = AppState {
db_client: Arc::new(DatabaseClient::new()),
api_key: "secret_key".to_string(),
counter: Arc::new(AtomicUsize::new(0)),
};
graphile_worker::WorkerOptions::default()
.add_extension(app_state)
.define_job::<ProcessUserTask>()
// ... other configuration
.init()
.await?;
You can customize how and when jobs run with the JobSpec
builder:
use graphile_worker::{JobSpecBuilder, JobKeyMode};
use chrono::Utc;
// Schedule a job to run after 5 minutes with high priority
let job_spec = JobSpecBuilder::new()
.run_at(Utc::now() + chrono::Duration::minutes(5))
.priority(10)
.job_key("welcome_email_user_123") // Unique identifier for deduplication
.job_key_mode(JobKeyMode::Replace) // Replace existing jobs with this key
.max_attempts(5) // Max retry attempts (default is 25)
.build();
utils.add_job(SendEmail { /* ... */ }, job_spec).await?;
Jobs with the same queue name run in series (one after another) rather than in parallel:
// These jobs will run one after another, not concurrently
let spec1 = JobSpecBuilder::new()
.queue_name("user_123_operations")
.build();
let spec2 = JobSpecBuilder::new()
.queue_name("user_123_operations")
.build();
utils.add_job(UpdateProfile { /* ... */ }, spec1).await?;
utils.add_job(SendEmail { /* ... */ }, spec2).await?;
You can schedule recurring jobs using crontab syntax:
// Run a task daily at 8:00 AM
let worker = WorkerOptions::default()
.with_crontab("0 8 * * * send_daily_report")?
.define_job::<SendDailyReport>()
// ... other configuration
.init()
.await?;
The WorkerUtils
class provides methods for managing jobs:
// Get a WorkerUtils instance
let utils = worker.create_utils();
// Remove a job by its key
utils.remove_job("job_key_123").await?;
// Mark jobs as completed
utils.complete_jobs(&[job_id1, job_id2]).await?;
// Permanently fail jobs with a reason
utils.permanently_fail_jobs(&[job_id3, job_id4], "Invalid data").await?;
// Reschedule jobs
let options = RescheduleJobOptions {
run_at: Some(Utc::now() + chrono::Duration::minutes(60)),
priority: Some(5),
max_attempts: Some(3),
..Default::default()
};
utils.reschedule_jobs(&[job_id5, job_id6], options).await?;
// Run database cleanup tasks
utils.cleanup(&[
CleanupTask::DeletePermenantlyFailedJobs,
CleanupTask::GcTaskIdentifiers,
CleanupTask::GcJobQueues,
]).await?;
- Flexible deployment: Run standalone or embedded in your application
- Multi-language support: Use from Rust, SQL or alongside the Node.js version
- Performance optimized:
- Low latency job execution (typically under 3ms)
- PostgreSQL
LISTEN
/NOTIFY
for immediate job notifications SKIP LOCKED
for efficient job fetching
- Robust job processing:
- Parallel processing with customizable concurrency
- Serialized execution via named queues
- Automatic retries with exponential backoff
- Customizable retry counts (default: 25 attempts over ~3 days)
- Scheduling features:
- Delayed execution with
run_at
- Job prioritization
- Crontab-like recurring tasks
- Task deduplication via
job_key
- Delayed execution with
- Type safety: End-to-end type checking of job payloads
- Minimal overhead: Direct serialization of task payloads
- PostgreSQL 12+
- Required for the
generated always as (expression)
feature - May work with older versions but has not been tested
- Required for the
Production ready but the API may continue to evolve. If you encounter any issues or have feature requests, please open an issue on GitHub.
This library is a Rust port of the excellent Graphile Worker by Benjie Gillam. If you find this library useful, please consider sponsoring Benjie's work, as all the research and architecture design was done by him.
MIT License - See LICENSE.md