pgjobs, a dead simple postgres job queueing mechanism

This project aims to be a blueprint for your own job queue solution with Go and PostgreSQL. It is recommended to fork this project and adjust the job queue with features to your own needs.

By using Postgres SKIP LOCKED feature, this allows to make a performant, non-blocking and independent queue. The technique is described here and here.

Performat: Non-blocking queue mechanism
Robust: Jobs will be 'freed' again when a worker crashes
Failed jobs will be retried until MAX_RETRIES, current attempt is passed as argument
Schedule jobs for later execution with EnqueueAt(ctx, job, "queuename", timeAt)
Support for multiple queues
Zero dependency

Example usage

A complete, runnable examples can be found under ./example/.

First define a pgjobs.Job, for example in ./jobs/emailUser.go:

package jobs

import (
	"encoding/json"
	"log"

	"github.com/rverton/pgjobs"
)

type EmailUser struct {
	Email string
}

func NewEmailUser(email string) *EmailUser {
	return &EmailUser{
		Email: email,
	}
}

// the action which should be executed
func (e EmailUser) Perform(attempt int32) error {
	log.Printf("emailing %v, attempt=%v", e.Email, attempt)
	return nil
}

// this is boilerplate code and does not need to be modified
func (e EmailUser) Load(data string) (pgjobs.Job, error) {
	var n EmailUser
	err := json.Unmarshal([]byte(data), &n)
	return n, err
}

You can then setup a queue, (optionally) enforce the jobs table schema, and work on queued jobs.

package main

import (
	"context"
	"database/sql"
	"example/jobs"
	"log"
	"os"

	_ "github.com/jackc/pgx/v5/stdlib"
	"github.com/rverton/pgjobs"
)

func main() {
	db, err := sql.Open("postgres", os.Getenv("DB_URL"))
	if err != nil {
		log.Fatal(err)
	}

	ctx := context.Background()

	// initiate queue with postgres connection
	queue := pgjobs.NewQueue(db)
	if err := queue.SetupSchema(ctx); err != nil {
		panic(err)
	}

	job := jobs.NewEmailUser("foo@example.com")
    
	// enequeue an example job for immediate execution
	if err = queue.Enqueue(context.Background(), job, "default"); err != nil {
		log.Printf("error enqueueing: %v", err)
	}
    
	// enequeue an example job for execution in 10s+
	if err = queue.EnqueueAt(context.Background(), job, "default", time.Now().Add(10*time.Second)); err != nil {
		log.Printf("error enqueueing: %v", err)
	}

	// start worker and pass all processable jobs
    // note: this worker will only process the passed jobs
	queues := []string{"default"}
	if err := queue.Worker(ctx, queues, &jobs.EmailUser{}); err != nil {
		log.Println(err)
	}
}

Configuration

The polling interval for workers can be adjusted via pgjobs.PollInterval.
For all other configuration, it is currently recommended to create a fork and adjust as needed.

ToDo

Make job processing more robust by using a transaction
Implement attempt handling
Add error handling and retries?
Add scheduled execution
Remove github.com/lib/pq dependency
Add more tests (dequeing, reflection)
Add priority queuing