Puppeteer Cluster

Create a cluster of puppeteer workers. This library spawns a pool of Chromium instances via Puppeteer and helps to keep track of jobs and errors. This is helpful if you want to crawl multiple pages or run tests in parallel. Puppeteer Cluster takes care of reusing Chromium and restarting the browser in case of errors.

• Installation
• Usage
• Examples
• Concurreny implementations
• Debugging
• API

What does this library do?

• Handling of crawling errors
• Auto restarts the browser in case of a crash
• Can automatically retry if a job fails
• Different concurrency models to choose from (pages, contexts, browsers)
• Simple to use, small boilerplate
• Progress view and monitoring statistics (see below)

Installation

Install puppeteer (if you don't already have it installed):

npm install --save puppeteer

Install puppeteer-cluster:

npm install --save puppeteer-cluster

Usage

The following is a typical example of using puppeteer-cluster. A cluster is created with 2 concurrent workers. Then a task is defined which includes going to the URL and taking a screenshot. We then queue two jobs and wait for the cluster to finish.

const { Cluster } = require('puppeteer-cluster');

(async () => {
  const cluster = await Cluster.launch({
    concurrency: Cluster.CONCURRENCY_CONTEXT,
    maxConcurrency: 2,
  });

  await cluster.task(async ({ page, data: url }) => {
    await page.goto(url);
    const screen = await page.screenshot();
    // Store screenshot, do something else
  });

  await cluster.queue('http://www.google.com/');
  await cluster.queue('http://www.wikipedia.org/');
  // many more pages

  await cluster.idle();
  await cluster.close();
})();

Examples

• Simple example
• Wait for a task to be executed
• Minimal screenshot server with express
• Deep crawling the Google search results
• Crawling the Alexa Top 1 Million
• Queuing functions (simple)
• Queuing functions (complex)
• Error handling
• Using a different puppeteer library (like puppeteer-core)

Concurreny implementations

There are different concurrency models, which define how isolated each job is run. You can set it in the options when calling Cluster.launch. The default option is Cluster.CONCURRENCY_CONTEXT, but it is recommended to always specify which one you want to use.

Concurrency	Description	Shared data
CONCURRENCY_PAGE	One Page for each URL	Shares everything (cookies, localStorage, etc.) between jobs.
CONCURRENCY_CONTEXT	Incognito page (see IncognitoBrowserContext) for each URL	No shared data.
CONCURRENCY_BROWSER	One browser (using an incognito page) per URL. If one browser instance crashes for any reason, this will not affect other jobs.	No shared data.
Custom concurrency (experimental)	You can create your own concurrency implementation. Copy one of the files of the concurrency/built-in directory and implement ConcurrencyImplementation. Then provide the class to the option concurrency. This part of the library is currently experimental and might break in the future, even in a minor version upgrade while the version has not reached 1.0.	Depends on your implementation

Debugging

Try to checkout the puppeteer debugging tips first. Your problem might not be related to puppeteer-cluster, but puppteer itself. Additionally, you can enable verbose logging to see which data is consumed by which worker and some other cluster information. Set the DEBUG environment variable to puppeteer-cluster:*. See an example below or checkout the debug docs for more information.

# Linux
DEBUG='puppeteer-cluster:*' node examples/minimal
# Windows Powershell
$env:DEBUG='puppeteer-cluster:*';node examples/minimal

API

• class: Cluster
  • Cluster.launch(options)
  • cluster.task(taskFunction)
  • cluster.queue([data] [, taskFunction])
  • cluster.execute([data] [, taskFunction])
  • cluster.idle()
  • cluster.close()

class: Cluster

Cluster module provides a method to launch a cluster of Chromium instances.

event: 'taskerror'

• <Error>
• <string|Object>

Emitted when a queued task ends in an error for some reason. Reasons might be a network error, your code throwing an error, timeout hit, etc. The first argument will the error itself. The second argument is the URL or data of the job (as given to Cluster.queue). If retryLimit is set to a value greater than 0, the cluster will automatically requeue the job and retry it again later. In case the task was queued via Cluster.execute there will be no event fired.

  cluster.on('taskerror', (err, data) => {
      console.log(`Error crawling ${data}: ${err.message}`);
  });

Cluster.launch(options)

• options <Object> Set of configurable options for the cluster. Can have the following fields:
  • concurrency <Cluster.CONCURRENCY_PAGE|Cluster.CONCURRENCY_CONTEXT|Cluster.CONCURRENCY_BROWSER|ConcurrencyImplementation> The chosen concurrency model. See Concurreny models for more information. Defaults to Cluster.CONCURRENCY_CONTEXT. Alternatively you can provide a class implementing ConcurrencyImplementation.
  • maxConcurrency <number> Maximal number of parallel workers. Defaults to 1.
  • puppeteerOptions <Object> Object passed to puppeteer.launch. See puppeteer documentation for more information. Defaults to {}.
  • retryLimit <number> How often do you want to retry a job before marking it as failed. Ignored by tasks queued via Cluster.execute. Defaults to 0.
  • retryDelay <number> How much time should pass at minimum between the job execution and its retry. Ignored by tasks queued via Cluster.execute. Defaults to 0.
  • sameDomainDelay <number> How much time should pass at minimum between two requests to the same domain. If you use this field, the queued data must be your URL or data must be an object containing a field called url.
  • skipDuplicateUrls <boolean> If set to true, will skip URLs which were already crawled by the cluster. Defaults to false. If you use this field, the queued data must be your URL or data must be an object containing a field called url.
  • timeout <number> Specify a timeout for all tasks. Defaults to 30000 (30 seconds).
  • monitor <boolean> If set to true, will provide a small command line output to provide information about the crawling process. Defaults to false.
  • workerCreationDelay <number> Time between creation of two workers. Set this to a value like 100 (0.1 seconds) in case you want some time to pass before another worker is created. You can use this to prevent a network peak right at the start. Defaults to 0 (no delay).
  • puppeteer <Object> In case you want to use a different puppeteer library (like puppeteer-core or puppeteer-extra), pass the object here. If not set, will default to using puppeteer. When using puppeteer-core, make sure to also provide puppeteerOptions.executablePath.
• returns: <Promise<Cluster>>

The method launches a cluster instance.

cluster.task(taskFunction)

• taskFunction <function(string|Object, Page, Object)> Sets the function, which will be called for each job. The function will be called with an object having the following fields:
  • page <Page> The page given by puppeteer, which provides methods to interact with a single tab in Chromium.
  • data The data of the job you provided to Cluster.queue.
  • worker <Object> An object containing information about the worker executing the current job.
    • id <number> ID of the worker. Worker IDs start at 0.
• returns: <Promise>

Specifies a task for the cluster. A task is called for each job you queue via Cluster.queue. Alternatively you can directly queue the function that you want to be executed. See Cluster.queue for an example.

cluster.queue([data] [, taskFunction])

• data Data to be queued. This might be your URL (a string) or a more complex object containing data. The data given will be provided to your task function(s). See [examples] for a more complex usage of this argument.
• taskFunction <function> Function like the one given to Cluster.task. If a function is provided, this function will be called (only for this job) instead of the function provided to Cluster.task. The function will be called with an object having the following fields:
  • page <Page> The page given by puppeteer, which provides methods to interact with a single tab in Chromium.
  • data The data of the job you provided as first argument to Cluster.queue. This might be undefined in case you only specified a function.
  • worker <Object> An object containing information about the worker executing the current job.
    • id <number> ID of the worker. Worker IDs start at 0.
• returns: <Promise>

Puts a URL or data into the queue. Alternatively (or even additionally) you can queue functions. See the examples about function queuing for more information: (Simple function queuing, complex function queuing)

cluster.execute([data] [, taskFunction])

• data Data to be queued. This might be your URL (a string) or a more complex object containing data. The data given will be provided to your task function(s). See [examples] for a more complex usage of this argument.
• taskFunction <function> Function like the one given to Cluster.task. If a function is provided, this function will be called (only for this job) instead of the function provided to Cluster.task. The function will be called with an object having the following fields:
  • page <Page> The page given by puppeteer, which provides methods to interact with a single tab in Chromium.
  • data The data of the job you provided as first argument to Cluster.queue. This might be undefined in case you only specified a function.
  • worker <Object> An object containing information about the worker executing the current job.
    • id <number> ID of the worker. Worker IDs start at 0.
• returns: <Promise>

Works like Cluster.queue, just that this function returns a Promise which will be resolved after the task is executed. In case an error happens during the execution, this function will reject the Promise with the thrown error. There will be no "taskerror" event fired. In addition, tasks queued via execute will ignore "retryLimit" and "retryDelay". For an example see the Execute example.

cluster.idle()

• returns: <Promise>

Promise is resolved when the queue becomes empty.

cluster.close()

• returns: <Promise>

Closes the cluster and all opened Chromium instances including all open pages (if any were opened). It is recommended to run Cluster.idle before calling this function. The Cluster object itself is considered to be disposed and cannot be used anymore.