active-job

Rails Active Job Expert

Create robust, well-tested background jobs using Active Job with Solid Queue (the Rails 8 default).

Philosophy

Core Principles:

1. Solid Queue is the default — Rails 8+ ships with Solid Queue. Use it. No Redis dependency needed.
2. Jobs must be idempotent — Jobs can run more than once. Design for it.
3. Arguments must be serializable — ActiveRecord objects via GlobalID, primitives, or custom serializers. Never pass arbitrary Ruby objects.
4. Always handle failure — Every job needs retry_on or discard_on. No exceptions.
5. Test jobs in isolation — Use perform_enqueued_jobs and assert_enqueued_with to verify behavior and enqueuing separately.

When To Use This Skill

• Creating new background jobs
• Moving slow work out of the request cycle (emails, API calls, data processing)
• Configuring Solid Queue (queues, workers, priorities, recurring tasks)
• Handling job retries, errors, and dead letters
• Testing job behavior and enqueuing
• Integrating ActionMailer with deliver_later
• Setting up bulk enqueuing, concurrency controls, or job continuations

Instructions

Step 1: Generate the Job

bin/rails generate job process_payment
# Creates:
#   app/jobs/process_payment_job.rb
#   test/jobs/process_payment_job_test.rb

With a specific queue:

bin/rails generate job process_payment --queue critical

Step 2: Define the Job

class ProcessPaymentJob < ApplicationJob
  queue_as :default

  retry_on Net::OpenTimeout, wait: :polynomially_longer, attempts: 5
  discard_on ActiveJob::DeserializationError

  def perform(order)
    # `order` is an ActiveRecord object — serialized via GlobalID automatically
    PaymentGateway.charge(order.amount, order.payment_method)
    order.update!(paid_at: Time.current)
  end
end

Every job MUST have:

• queue_as — explicit queue assignment
• retry_on and/or discard_on — error handling strategy
• Idempotent perform — safe to run multiple times

Step 3: Enqueue the Job

# Run as soon as a worker is free
ProcessPaymentJob.perform_later(order)

# Run at a specific time
ProcessPaymentJob.set(wait_until: Date.tomorrow.noon).perform_later(order)

# Run after a delay
ProcessPaymentJob.set(wait: 5.minutes).perform_later(order)

# Run with a specific priority (lower = higher priority)
ProcessPaymentJob.set(priority: 0).perform_later(order)

# Run synchronously (useful in console, tests, rake tasks)
ProcessPaymentJob.perform_now(order)

perform_later vs perform_now:

• perform_later — enqueues to background queue. Use in controllers, models, services.
• perform_now — runs inline, blocking. Use in console, rake tasks, tests, and other jobs that need sequential execution.

Step 4: Handle Errors

ALWAYS define error handling. This is non-negotiable.

class ExternalApiJob < ApplicationJob
  # Retry transient failures with exponential backoff
  retry_on Net::OpenTimeout, wait: :polynomially_longer, attempts: 5
  retry_on Faraday::ConnectionError, wait: 30.seconds, attempts: 3

  # Discard permanently broken jobs
  discard_on ActiveJob::DeserializationError  # Record was deleted
  discard_on ArgumentError                     # Bad arguments, retrying won't help

  # Optional: callback when a job is discarded
  after_discard do |job, exception|
    Rails.logger.warn("Discarded #{job.class.name}: #{exception.message}")
    ErrorTracker.notify(exception, job_id: job.job_id)
  end

  def perform(record)
    # ...
  end
end

retry_on options:

• wait: — seconds or :polynomially_longer (default: 3s)
• attempts: — max retries (default: 5)
• queue: — move to a different queue on retry
• priority: — change priority on retry
• jitter: — add randomness to wait time (default: 0.15 = 15%)

discard_on — silently drops the job. Use for unrecoverable errors.

Step 5: Configure Queues and Priority

Queue assignment:

class ImportJob < ApplicationJob
  queue_as :low_priority
end

Dynamic queue based on arguments:

class ProcessVideoJob < ApplicationJob
  queue_as do
    video = self.arguments.first
    video.owner.premium? ? :premium : :default
  end
end

Priority within a queue (lower number = higher priority):

class UrgentNotificationJob < ApplicationJob
  queue_as :notifications
  queue_with_priority 0  # Processed first within :notifications queue
end

class DigestNotificationJob < ApplicationJob
  queue_as :notifications
  queue_with_priority 10  # Processed after urgent ones
end

Solid Queue queue ordering in config/queue.yml:

production:
  workers:
    - queues: [critical, default, low_priority]
      threads: 5
      polling_interval: 0.1

⚠️ Queue order takes precedence over priority. All jobs in critical process before ANY job in default, regardless of individual priority values. Priority only matters within a single queue.

Step 6: Serialize Arguments Correctly

Safe argument types (work out of the box):

• Primitives: String, Integer, Float, Boolean, nil, Symbol
• Time types: Date, Time, DateTime, ActiveSupport::TimeWithZone, Duration
• Collections: Array, Hash (string/symbol keys only), Range
• Classes: Module, Class
• ActiveRecord objects — serialized via GlobalID automatically

DANGEROUS — do NOT pass these directly:

# ❌ WRONG — arbitrary Ruby objects are NOT serializable
SomeJob.perform_later(Struct.new(:a).new(1))
SomeJob.perform_later(OpenStruct.new(foo: "bar"))
SomeJob.perform_later(URI.parse("https://example.com"))
SomeJob.perform_later(Money.new(100, "USD"))  # Without custom serializer

# ✅ CORRECT — pass primitives, let the job reconstruct
SomeJob.perform_later("https://example.com")
SomeJob.perform_later(100, "USD")

# ✅ CORRECT — ActiveRecord objects use GlobalID
SomeJob.perform_later(user)  # Serialized as "gid://app/User/123"

Custom serializer for value objects:

# app/serializers/money_serializer.rb
class MoneySerializer < ActiveJob::Serializers::ObjectSerializer
  def serialize(money)
    super("amount" => money.amount, "currency" => money.currency)
  end

  def deserialize(hash)
    Money.new(hash["amount"], hash["currency"])
  end

  def klass
    Money
  end
end

# config/initializers/custom_serializers.rb
Rails.application.config.active_job.custom_serializers << MoneySerializer

Step 7: Use Callbacks

class AuditedJob < ApplicationJob
  before_perform :log_start
  after_perform :log_completion
  around_perform :measure_duration

  # before_enqueue / around_enqueue / after_enqueue also available
  # after_discard — fires when discard_on matches

  private

  def log_start
    Rails.logger.info("Starting #{self.class.name} #{job_id}")
  end

  def log_completion
    Rails.logger.info("Completed #{self.class.name} #{job_id}")
  end

  def measure_duration
    started = Process.clock_gettime(Process::CLOCK_MONOTONIC)
    yield
    elapsed = Process.clock_gettime(Process::CLOCK_MONOTONIC) - started
    StatsD.measure("jobs.#{self.class.name.underscore}.duration", elapsed)
  end
end

Available callbacks: before_enqueue, around_enqueue, after_enqueue, before_perform, around_perform, after_perform, after_discard

Step 8: Test the Job

Test job behavior (does it do the right thing?):

require "test_helper"

class ProcessPaymentJobTest < ActiveSupport::TestCase
  test "marks order as paid" do
    order = orders(:unpaid_order)

    perform_enqueued_jobs do
      ProcessPaymentJob.perform_later(order)
    end

    order.reload
    assert order.paid_at.present?
  end

  test "retries on timeout" do
    order = orders(:unpaid_order)

    assert_enqueued_with(job: ProcessPaymentJob) do
      ProcessPaymentJob.perform_later(order)
    end
  end
end

Test that jobs get enqueued (from calling code):

class OrdersControllerTest < ActionDispatch::IntegrationTest
  test "enqueues payment job on create" do
    assert_enqueued_with(job: ProcessPaymentJob) do
      post orders_path, params: { order: valid_order_params }
    end
  end

  test "enqueues to correct queue" do
    assert_enqueued_with(job: ProcessPaymentJob, queue: "critical") do
      post orders_path, params: { order: valid_order_params }
    end
  end
end

Test helpers:

# Execute all enqueued jobs inline
perform_enqueued_jobs do
  SomeJob.perform_later(args)
end

# Execute only specific job types
perform_enqueued_jobs(only: ProcessPaymentJob) do
  # ...
end

# Assert a job was enqueued
assert_enqueued_with(job: MyJob, args: [record], queue: "default")

# Assert number of jobs enqueued
assert_enqueued_jobs 2 do
  # ...
end

# Assert no jobs enqueued
assert_no_enqueued_jobs do
  # ...
end

Step 9: ActionMailer Integration

# Send email asynchronously — enqueues via Active Job
UserMailer.welcome(@user).deliver_later

# Send with delay
UserMailer.welcome(@user).deliver_later(wait: 1.hour)

# Send immediately (bypasses queue)
UserMailer.welcome(@user).deliver_now

⚠️ In Rake tasks, always use deliver_now — the process may exit before deliver_later jobs are processed.

Error reporting for mailer jobs:

class ApplicationMailer < ActionMailer::Base
  ActionMailer::MailDeliveryJob.rescue_from(Exception) do |exception|
    Rails.error.report(exception)
    raise exception
  end
end

Step 10: Bulk Enqueuing

# Create job instances (don't call perform_later individually)
jobs = users.map { |user| WelcomeEmailJob.new(user) }

# Enqueue all at once — single round-trip to the queue backend
ActiveJob.perform_all_later(jobs)

# With options
jobs = users.map { |user| WelcomeEmailJob.new(user).set(wait: 1.day) }
ActiveJob.perform_all_later(jobs)

# Mix different job classes
jobs = [
  CleanupJob.new(record),
  NotifyJob.new(user),
  ExportJob.new(data)
]
ActiveJob.perform_all_later(jobs)

⚠️ around_enqueue callbacks are NOT triggered during bulk enqueuing. Use ActiveSupport::Notifications with the enqueue_all.active_job event if you need to hook into bulk operations.

Solid Queue Configuration

See references/queues-and-priority.md for full multi-worker queue.yml examples.

Starting the Queue Worker

bin/jobs start  # Development — run in separate terminal

Concurrency Controls

class ImportJob < ApplicationJob
  # Only 2 imports per account at a time
  limits_concurrency to: 2, key: ->(account) { account }, duration: 5.minutes

  def perform(account)
    account.run_import
  end
end

Recurring Tasks (config/recurring.yml)

production:
  cleanup_old_records:
    class: CleanupJob
    schedule: every day at 3am
  send_daily_digest:
    class: DailyDigestJob
    args: [summary]
    schedule: every day at 9am
  health_check:
    command: "HealthCheck.run"
    schedule: every 5 minutes

Job Continuations (Rails 8.1+)

For long-running jobs that may be interrupted:

class LargeImportJob < ApplicationJob
  include ActiveJob::Continuable

  def perform(import_id)
    @import = Import.find(import_id)

    step :validate do
      @import.validate_data!
    end

    step :process do |step|
      @import.rows.find_each(start: step.cursor) do |row|
        row.process!
        step.advance! from: row.id
      end
    end

    step :finalize do
      @import.mark_complete!
    end
  end
end

Common Agent Mistakes

1. Passing non-serializable objects — Only primitives, ActiveRecord (via GlobalID), and registered serializer types. Everything else will raise SerializationError.
2. No error handling — Every job needs retry_on or discard_on. A job without error handling is a ticking time bomb.
3. Not handling DeserializationError — If a record is deleted between enqueue and perform, you get this error. Always discard_on ActiveJob::DeserializationError.
4. Blocking jobs without timeouts — HTTP calls, external APIs must have timeouts. Add retry_on for timeout errors.
5. Non-idempotent jobs — If the job charges money, send a guard: return if order.paid?. Jobs CAN run twice.
6. Using perform_now in production controllers — This blocks the request. Use perform_later.
7. Forgetting queue workers in dev — bin/jobs start or configure config.active_job.queue_adapter = :async for development.
8. Testing with perform_now when you mean to test enqueuing — Use assert_enqueued_with to verify the job was enqueued, perform_enqueued_jobs to verify behavior.

ApplicationJob Base Class

Set up sensible defaults in app/jobs/application_job.rb:

class ApplicationJob < ActiveJob::Base
  # Automatically retry transient failures
  retry_on ActiveRecord::Deadlocked, wait: 5.seconds, attempts: 3

  # Discard if the record no longer exists
  discard_on ActiveJob::DeserializationError

  # Report errors to your error tracker
  rescue_from(Exception) do |exception|
    Rails.error.report(exception)
    raise exception
  end

  # Wait for transaction commit before enqueuing (prevents race conditions)
  self.enqueue_after_transaction_commit = true
end

Running and Debugging

# Start queue worker
bin/jobs start

# Run a job from console
ProcessPaymentJob.perform_now(Order.last)

# Check queue status (with mission_control-jobs gem)
# Visit /jobs in browser

# Enable verbose enqueue logging
# config/environments/development.rb
config.active_job.verbose_enqueue_logs = true

Reference Files

See the references/ directory for detailed patterns and edge cases:

• references/serialization.md — Job lifecycle, GlobalID, argument types, custom serializers
• references/error-handling.md — retry_on, discard_on, idempotency guards
• references/queues-and-priority.md — Multi-worker setup, naming, priority, ActionMailer integration
• references/testing.md — Enqueue assertions, job behavior, error handling, scheduled job tests
• references/solid-queue.md — Database setup, concurrency controls, recurring tasks, monitoring
• references/advanced-patterns.md — Continuations, performance, anti-patterns, migration guides