Skip to main content
BullMQ’s flow system is built on parent-child dependencies. Understanding how these dependencies work is crucial for building complex workflows.

​
How Dependencies Work

In BullMQ flows, a parent job will not be processed until all of its children have completed successfully. This creates a dependency tree where:
  • Children are processed first
  • Parents wait in the “waiting-children” state
  • Parent processing begins only after all children complete

​
Dependency States

Jobs in a flow can have dependencies in different states:

Processed

Children that have completed successfully

Unprocessed

Children that are waiting or in progress

Failed

Children that have failed during processing

Ignored

Failed children that are ignored by parent

​
Managing Dependencies

​
Get Dependencies

Retrieve all dependencies of a parent job: 
const dependencies = await job.getDependencies();
Get specific types of dependencies with pagination: 
const { processed, nextProcessedCursor } = await job.getDependencies({
  processed: {
    count: 5,
    cursor: 0,
  },
});

const { unprocessed, nextUnprocessedCursor } = await job.getDependencies({
  unprocessed: {
    count: 5,
    cursor: 0,
  },
});

const { failed, nextFailedCursor } = await job.getDependencies({
  failed: {
    count: 5,
    cursor: 0,
  },
});

const { ignored, nextIgnoredCursor } = await job.getDependencies({
  ignored: {
    count: 5,
    cursor: 0,
  },
});

​
Get Dependencies Count

Get counts of dependencies by state: 
const { failed, ignored, processed, unprocessed } =
  await job.getDependenciesCount();
Or retrieve specific counts: 
const { failed } = await job.getDependenciesCount({
  failed: true,
});

const { ignored, processed } = await job.getDependenciesCount({
  ignored: true,
  processed: true,
});

​
Get Children Values

Access the return values from all child jobs: 
const childrenValues = await job.getChildrenValues();
// Returns: { 'child-job-id-1': result1, 'child-job-id-2': result2, ... }

​
Parent Options

When creating child jobs, you can specify how they relate to their parent: 
interface ParentOptions {
  id: string;
  queue: string;
}
Children automatically inherit the parent relationship: 
const flow = await flowProducer.add({
  name: 'parent-job',
  queueName: 'parent-queue',
  data: {},
  children: [
    {
      name: 'child-job',
      queueName: 'child-queue',
      data: {},
      // Parent relationship is automatically set
    },
  ],
});

​
Accessing Parent Information

Child jobs have access to their parent information: 
// Get the parent key
const parentKey = job.parentKey;
// Format: "bull:parent-queue:parent-job-id"

// Parse parent information
if (parentKey) {
  const [prefix, queueName, jobId] = parentKey.split(':');
  console.log(`Parent Queue: ${queueName}, Parent ID: ${jobId}`);
}

​
Checking Parent State

Parent jobs waiting for children have a special state: 
const state = await parentJob.getState();
if (state === 'waiting-children') {
  console.log('Parent is waiting for children to complete');
}

​
Complex Dependency Patterns

​
Serial Execution

Create a chain where jobs execute one after another: 
const chain = await flowProducer.add({
  name: 'step-3',
  queueName: 'processing',
  data: { step: 3 },
  children: [
    {
      name: 'step-2',
      queueName: 'processing',
      data: { step: 2 },
      children: [
        {
          name: 'step-1',
          queueName: 'processing',
          data: { step: 1 },
        },
      ],
    },
  ],
});
// Execution order: step-1 → step-2 → step-3

​
Parallel Execution with Aggregation

Process multiple children in parallel, then aggregate results: 
const flow = await flowProducer.add({
  name: 'aggregate-results',
  queueName: 'aggregation',
  data: {},
  children: [
    { name: 'task-1', queueName: 'processing', data: { id: 1 } },
    { name: 'task-2', queueName: 'processing', data: { id: 2 } },
    { name: 'task-3', queueName: 'processing', data: { id: 3 } },
  ],
});

// In the parent worker:
const worker = new Worker('aggregation', async job => {
  const childrenValues = await job.getChildrenValues();
  const results = Object.values(childrenValues);
  return results.reduce((sum, val) => sum + val, 0);
});

​
Mixed Serial and Parallel

Combine serial and parallel execution patterns: 
const flow = await flowProducer.add({
  name: 'final-step',
  queueName: 'final',
  data: {},
  children: [
    {
      name: 'parallel-aggregator',
      queueName: 'aggregation',
      data: {},
      children: [
        { name: 'parallel-1', queueName: 'processing', data: {} },
        { name: 'parallel-2', queueName: 'processing', data: {} },
        { name: 'parallel-3', queueName: 'processing', data: {} },
      ],
    },
  ],
});
// parallel-1, parallel-2, parallel-3 run in parallel
// → parallel-aggregator runs after all complete
// → final-step runs last

​
Dependency Failure Handling

By default, if any child fails, the parent will not be processed. However, you can customize this behavior:

Remove Dependency

Remove failed child from parent dependencies

Ignore Dependency

Ignore failed child but keep dependency

Fail Parent

Make parent fail when child fails

Continue Parent

Process parent immediately on child failure

​
Example: Multi-Stage Processing

Here’s a complete example of a multi-stage data processing workflow: 
import { FlowProducer, Worker, Queue } from 'bullmq';

const flowProducer = new FlowProducer({ connection });

// Create a flow for processing a dataset
const flow = await flowProducer.add({
  name: 'generate-report',
  queueName: 'reporting',
  data: { datasetId: 'dataset-123' },
  children: [
    {
      name: 'process-batch',
      queueName: 'processing',
      data: { batch: 1 },
      children: [
        { name: 'fetch-data', queueName: 'fetching', data: { batch: 1 } },
        { name: 'validate-data', queueName: 'validation', data: { batch: 1 } },
      ],
    },
    {
      name: 'process-batch',
      queueName: 'processing',
      data: { batch: 2 },
      children: [
        { name: 'fetch-data', queueName: 'fetching', data: { batch: 2 } },
        { name: 'validate-data', queueName: 'validation', data: { batch: 2 } },
      ],
    },
  ],
});

// Worker for processing batches
const processingWorker = new Worker('processing', async job => {
  const childrenValues = await job.getChildrenValues();
  const fetchResult = Object.values(childrenValues)[0];
  const validationResult = Object.values(childrenValues)[1];
  
  return {
    batch: job.data.batch,
    recordsProcessed: fetchResult.count,
    validationPassed: validationResult.success,
  };
});

// Worker for generating final report
const reportingWorker = new Worker('reporting', async job => {
  const childrenValues = await job.getChildrenValues();
  const batchResults = Object.values(childrenValues);
  
  const totalRecords = batchResults.reduce(
    (sum, batch) => sum + batch.recordsProcessed,
    0
  );
  
  return {
    datasetId: job.data.datasetId,
    totalRecords,
    batchesProcessed: batchResults.length,
  };
});

​
API Reference

Build docs developers (and LLMs) love

Get started for freeTalk to us