Continuous Integration

Playwright tests can be executed in CI environments. We have created sample configurations for common CI providers.

Introduction

3 steps to get your tests running on CI:

  1. Ensure CI agent can run browsers: Use our Docker image in Linux agents. Windows and macOS agents do not require any additional dependencies.
  2. Install Playwright: In most projects, this would be done with npm ci (or npm install). Playwright would install the relevant browsers automatically.
  3. Run your tests: Use npm test or equivalent to execute your tests.

CI configurations

GitHub Actions

The Playwright GitHub Action can be used to run Playwright tests on GitHub Actions.

  1. steps:
  2.   - uses: microsoft/playwright-github-action@v1
  3.   - name: Run your tests
  4.     run: npm test

We run our tests on GitHub Actions, across a matrix of 3 platforms (Windows, Linux, macOS) and 3 browsers (Chromium, Firefox, WebKit).

Docker

We have a pre-built Docker image which can either be used directly, or as a reference to update your existing Docker definitions.

Suggested configuration

  1. By default, Docker runs a container with a /dev/shm shared memory space 64MB. This is typically too small for Chromium and will cause Chromium to crash when rendering large pages. To fix, run the container with docker run --shm-size=1gb to increase the size of /dev/shm. Since Chromium 65, this is no longer necessary. Instead, launch the browser with the --disable-dev-shm-usage flag:

    1. const browser = await playwright.chromium.launch({
    2.   args: ['--disable-dev-shm-usage']
    3. });

    This will write shared memory files into /tmp instead of /dev/shm. See crbug.com/736452 for more details.

  2. Using --ipc=host is also recommended when using Chromium—without it Chromium can run out of memory and crash. Learn more about this option in Docker docs.

  3. Seeing other weird errors when launching Chromium? Try running your container with docker run --cap-add=SYS_ADMIN when developing locally. Since the Dockerfile adds a pwuser user as a non-privileged user, it may not have all the necessary privileges.

  4. dumb-init is worth checking out if you’re experiencing a lot of zombies Chromium processes sticking around. There’s special treatment for processes with PID=1, which makes it hard to terminate Chromium properly in some cases (e.g. in Docker).

Azure Pipelines

For Windows or macOS agents, no additional configuration required, just install Playwright and run your tests.

For Linux agents, you can use our Docker container with Azure Pipelines support for running containerized jobs. Alternatively, you can refer to the Dockerfile to see additional dependencies that need to be installed on a Ubuntu agent.

  1. pool:
  2.   vmImage: 'ubuntu-18.04'
  4. container: mcr.microsoft.com/playwright:bionic
  6. steps:
  7. - script: npm install
  8. - script: npm run test

Travis CI

We run our tests on Travis CI over a Linux agent (Ubuntu 18.04). Use our Travis configuration to see list of additional dependencies to be installed.

Suggested configuration

  1. User namespace cloning should be enabled to support proper sandboxing
  2. xvfb should be launched in order to run Chromium in non-headless mode (e.g. to test Chrome Extensions)
  3. If your project does not have package-lock.json, Travis would be auto-caching node_modules directory. If you run npm install (instead of npm ci), it is possible that the browser binaries are not downloaded. Fix this with these steps outlined below.

To sum up, your .travis.yml might look like this:

  1. language: node_js
  2. dist: bionic
  3. addons:
  4.   apt:
  5.     packages:
  6.     # This is required to run chromium
  7.     - libgbm1
  8.     # These are required to run webkit
  9.     - libwoff1
  10.     - libopus0
  11.     - libwebp6
  12.     - libwebpdemux2
  13.     - libenchant1c2a
  14.     - libgudev-1.0-0
  15.     - libsecret-1-0
  16.     - libhyphen0
  17.     - libgdk-pixbuf2.0-0
  18.     - libegl1
  19.     - libgles2
  20.     - libevent-2.1-6
  21.     - libnotify4
  22.     - libxslt1.1
  23.     - libvpx5
  24.     # For headful execution
  25.     - xvfb
  27. # allow headful tests
  28. before_install:
  29.   # Enable user namespace cloning
  30.   - "sysctl kernel.unprivileged_userns_clone=1"
  31.   # Launch XVFB
  32.   - "export DISPLAY=:99.0"
  33.   - "sh -e /etc/init.d/xvfb start"

CircleCI

We run our tests on CircleCI, with our pre-built Docker image. Use our CircleCI configuration to create your own. Running Playwright smoothly on CircleCI requires the following steps:

  1. Use the pre-built Docker image in your config like so:

    1. docker:
    2.   - image: mcr.microsoft.com/playwright:bionic
    3.     environment:
    4.       NODE_ENV: development # Needed if playwright is in `devDependencies`

  2. If you’re using Playwright through Jest, then you may encounter an error spawning child processes:

    1. [00:00.0]  jest args: --e2e --spec --max-workers=36
    2. Error: spawn ENOMEM
    3.    at ChildProcess.spawn (internal/child_process.js:394:11)

    This is likely caused by Jest autodetecting the number of processes on the entire machine (36) rather than the number allowed to your container (2). To fix this, set jest --maxWorkers=2 in your test command.

AppVeyor

We run our tests on Windows agents in AppVeyor. Use our AppVeyor configuration to create your own.

Bitbucket Pipelines

Bitbucket Pipelines can use public Docker images as build environments. To run Playwright tests on Bitbucket, use our public Docker image (see Dockerfile).

  1. image: mcr.microsoft.com/playwright:bionic

While the Docker image supports sandboxing for Chromium, it does not work in the Bitbucket Pipelines environment. To launch Chromium on Bitbucket Pipelines, use the --no-sandbox launch argument.

  1. const { chromium } = require('playwright');
  2. const browser = await chromium.launch({ args: ['--no-sandbox'] });

GitLab CI

To run Playwright tests on GitLab, use our public Docker image (see Dockerfile).

  1. stages:
  2.   - test
  4. tests:
  5.   stage: test
  6.   image: mcr.microsoft.com/playwright:bionic
  7.   script:
  8.     - npm install # This should install playwright
  9.     - npm run test

Caching browsers

By default, Playwright downloads browser binaries when the Playwright NPM package is installed. The NPM packages have a postinstall hook that downloads the browser binaries. This behavior can be customized with environment variables.

Caching browsers on CI is strictly optional: The postinstall hooks should execute and download the browser binaries on every run.

Exception: node_modules are cached

Most CI providers cache the npm-cache directory (located at $HOME/.npm). If your CI pipelines caches the node_modules directory and you run npm install (instead of npm ci), the default configuration will not work. This is because the npm install step will find the Playwright NPM package on disk and not execute the postinstall step.

Travis CI automatically caches node_modules if your repo does not have a package-lock.json file.

This behavior can be fixed with one of the following approaches:

  1. Move to caching $HOME/.npm or the npm-cache directory. (This is the default behavior in most CI providers.)
  2. Set PLAYWRIGHT_BROWSERS_PATH=0 as the environment variable before running npm install. This will download the browser binaries in the node_modules directory and cache them with the package code. See installation docs.
  3. Use npm ci (instead of npm install) which forces a clean install: by removing the existing node_modules directory. See npm docs.
  4. Cache the browser binaries, with the steps below.

Directories to cache

With the default behavior, Playwright downloads the browser binaries in the following directories:

  • %USERPROFILE%\AppData\Local\ms-playwright on Windows
  • ~/Library/Caches/ms-playwright on MacOS
  • ~/.cache/ms-playwright on Linux

To cache the browser downloads between CI runs, cache this location in your CI configuration, against a hash of the Playwright version.

Debugging browser launches

Playwright supports the DEBUG environment variable to output debug logs during execution. Setting it to pw:browser* is helpful while debugging Error: Failed to launch browser errors.

  1. DEBUG=pw:browser* npm run test

Running headful

By default, Playwright launches browsers in headless mode. This can be changed by passing a flag when the browser is launched.

  1. // Works across chromium, firefox and webkit
  2. const { chromium } = require('playwright');
  3. const browser = await chromium.launch({ headless: false });

On Linux agents, headful execution requires Xvfb to be installed. Our Docker image and GitHub Action have Xvfb pre-installed. To run browsers in headful mode with Xvfb, add xvfb-run before the Node.js command.

  1. xvfb-run node index.js