Queues

Queues are the final aspect of Payload's Jobs Queue and deal with how to run your jobs. Up to this point, all we've covered is how to queue up jobs to run, but so far, we aren't actually running any jobs.

When you go to run jobs, Payload will query for any jobs that are added to the queue and then run them. By default, all queued jobs are added to the default queue.

But, imagine if you wanted to have some jobs that run nightly, and other jobs which should run every five minutes.

By specifying the queue name when you queue a new job using payload.jobs.queue(), you can queue certain jobs with queue: 'nightly', and other jobs can be left as the default queue.

Then, you could configure two different runner strategies:

1. A cron that runs nightly, querying for jobs added to the nightly queue
2. Another that runs any jobs that were added to the default queue every ~5 minutes or so

As mentioned above, you can queue jobs, but the jobs won't run unless a worker picks up your jobs and runs them. This can be done in four ways:

For dedicated servers, the recommended approach is to use Payload's bin script to run jobs. This creates a completely separate process from your Next.js server, making it easier to deploy, scale, and manage workers independently.

Benefits of using bin scripts:

• Separate process: Runs completely independently from your Next.js server, preventing any impact on API response times
• Easy deployment: Can be deployed as a separate service/container, making it simple to scale workers independently
• Simple monitoring: Each worker process can be monitored, restarted, and scaled independently
• No Next.js overhead: Workers don't load the entire Next.js application, making them lighter and faster

Deployment example:

This makes it easy to run multiple workers for different queues, scale them independently, and deploy them to different servers if needed.

The jobs.autoRun property allows you to configure cron jobs that automatically execute jobs from your queue at specified intervals. This runs within your Next.js process.

Think of it this way:

• Queuing = Adding jobs to the database to be run later
• Running = Actually executing the job handler functions

autoRun only handles the running part.

Example:

In this example:

• processPayment jobs are queued manually (e.g., from API endpoints using payload.jobs.queue())
• generateReport jobs are auto-queued at midnight via the schedule property
• The first autoRun entry processes manually-queued jobs from the 'default' queue every 5 minutes
• The second autoRun entry processes scheduled jobs from the 'nightly' queue every minute

You can execute jobs by making a fetch request to the /api/payload-jobs/run endpoint:

This endpoint is automatically mounted for you and is helpful in conjunction with serverless platforms like Vercel, where you might want to use Vercel Cron to invoke a serverless function that executes your jobs.

• limit: The maximum number of jobs to run in this invocation (default: 10).
• queue: The name of the queue to run jobs from. If not specified, jobs will be run from the default queue.
• allQueues: If set to true, all jobs from all queues will be run. This will ignore the queue parameter.

If you're deploying on Vercel, you can add a vercel.json file in the root of your project that configures Vercel Cron to invoke the run endpoint on a cron schedule.

Here's an example of what this file will look like:

The configuration above schedules the endpoint /api/payload-jobs/run to be invoked every 5 minutes.

The last step will be to secure your run endpoint so that only the proper users can invoke the runner.

To do this, you can set an environment variable on your Vercel project called CRON_SECRET, which should be a random string—ideally 16 characters or longer.

Then, you can modify the access function for running jobs by ensuring that only Vercel can invoke your runner.

This works because Vercel automatically makes the CRON_SECRET environment variable available to the endpoint as the Authorization header when triggered by the Vercel Cron, ensuring that the jobs can be run securely.

After the project is deployed to Vercel, the Vercel Cron job will automatically trigger the /api/payload-jobs/run endpoint in the specified schedule, running the queued jobs in the background.

If you want to process jobs programmatically from your server-side code, you can use the Local API:

Run all jobs:

Run a single job:

By default, jobs are processed first in, first out (FIFO). This means that the first job added to the queue will be the first one processed. However, you can also configure the order in which jobs are processed.

You can configure the order in which jobs are processed in the jobs configuration by passing the processingOrder property. This mimics the Payload sort property that's used for functionality such as payload.find().

You can also set this on a queue-by-queue basis:

If you need even more control over the processing order, you can pass a function that returns the processing order - this function will be called every time a queue starts processing jobs.

You can configure the order in which jobs are processed in the payload.jobs.queue method by passing the processingOrder property.

Here are typical patterns for organizing your queues:

Separate jobs by priority to ensure critical tasks run quickly:

Then queue jobs to appropriate queues:

Only run jobs on specific servers:

Use cases:

• Dedicate specific servers to job processing
• Disable job processing during deployments
• Scale job workers independently from API servers

Group jobs by feature or domain:

This makes it easy to:

• Monitor specific features
• Scale individual features independently
• Pause/resume specific types of work

Here's a quick guide to help you choose:

Recommendations:

• Production (Dedicated Server): Use Bin script with --cron flag
• Production (Serverless): Use Endpoint + Vercel Cron
• Development: Use Bin script or Local API
• Testing: Use Local API with payload.jobs.runByID()

If you configured autoRun but jobs aren't executing, check these common causes in order:

Symptom: autoRun is configured but no jobs ever execute.

Diagnosis: autoRun runs jobs that are already in the queue. If nothing is adding jobs to the queue, there's nothing to run.

Solution: Queue jobs either:

• Manually via payload.jobs.queue() in your code (e.g., in hooks, endpoints), OR
• Automatically via the schedule property on tasks/workflows

Example of the problem:

Solution:

Symptom: Jobs appear in the payload-jobs collection but never execute.

Diagnosis: The queue name in your task's schedule doesn't match the queue name in autoRun.

Example of the problem:

Solution: Make sure queue names match exactly:

Symptom: Jobs work locally but don't run in production on serverless platforms.

Diagnosis: autoRun requires a long-running server and won't work on serverless platforms where instances shut down between requests.

Solution: Use the endpoint method with Vercel Cron instead of autoRun.

Symptom: Jobs exist in the payload-jobs collection with completedAt: null but aren't running.

Diagnosis: Jobs may have waitUntil set to a future date - they won't run until that time passes.

Solution: Check the payload-jobs collection:

Symptom: Jobs work initially but stop running after you make code changes in development.

Diagnosis: Hot Module Reload (HMR) in Next.js disrupts cron schedules. This is expected behavior in development.

Solution: Restart your dev server after making changes to job configurations. This is not an issue in production.

Is shouldAutoRun returning true?

Is autoRun configured correctly?

Are jobs in the correct queue?

Check the jobs collection

Enable the jobs collection in admin:

Look for jobs with:

• processing: true but stuck → Worker may have crashed
• hasError: true → Check the log field for errors
• completedAt: null → Job hasn't run yet

Check the job logs in the payload-jobs collection:

Increase limit

Run more frequently

Add more workers

Scale horizontally by running multiple servers with ENABLE_JOB_WORKERS=true.