Optimizations

This document provides an overview of several methods for optimizing your CircleCI configuration. Each optimization method will be described briefly, will present possible use cases, and will provide an example optimization for speeding up your jobs.

Docker image choice
Caching dependencies
Workflows
Workspaces
Parallelism
Resource class
Docker layer caching

Note: For Cloud customers, some of the features discussed in this document may require a specific pricing plan. Visit our pricing page to get an overview of the plans CircleCI offers. Or, if you are a logged in to the CircleCI web application, go to Plan from the sidebar to view and make adjustments to your plan.

Docker image choice

Choosing the right docker image for your project can have huge impact on build time. For example, choosing a basic language image means dependencies and tools need to be downloaded each time your pipeline is run, whereas, if you choose or build an image that has these dependencies and tools already installed, this time will be saved for each build run. When configuring your projects and specifying images, consider the following options:

CircleCI provides a range of convenience images, typically based on official Docker images, but with a range of useful language tools pre-installed.
You can create your own images, maximizing specificity for your projects. To help with this we provide both a Docker image build wizard, and guidance for building images manually.

Caching dependencies

Caching should be one of the first things you consider when trying to optimize your jobs. If a job fetches data at any point, it is likely that you can make use of caching. A common example is the use of a package/dependency manager. If your project uses Yarn, Bundler, or Pip, for example, the dependencies downloaded during a job can be cached for later use rather than being re-downloaded on every build.

version: 2
jobs:
  build:
    steps: # a collection of executable commands making up the 'build' job
      - checkout # pulls source code to the working directory
      - restore_cache: # **restores saved dependency cache if the Branch key template or requirements.txt files have not changed since the previous run**
          key: deps1-{{ .Branch }}-{{ checksum "requirements.txt" }}
      - run: # install and activate virtual environment with pip
          command: |
            python3 -m venv venv
            . venv/bin/activate
            pip install -r requirements.txt
      - save_cache: # ** special step to save dependency cache **
          key: deps1-{{ .Branch }}-{{ checksum "requirements.txt" }}
          paths:
            - "venv"

Make note of the use of a checksum in the cache key; this is used to calculate when a specific dependency-management file (such as a package.json or requirements.txt in this case) changes and so the cache will be updated accordingly. In the above example, the restore_cache example uses interpolation to put dynamic values into the cache-key, allowing more control in what exactly constitutes the need to update a cache.

We recommend that you verify that the dependencies installation step succeeds before adding caching steps. Caching a failed dependency step will require you to change the cache key in order to avoid failed builds due to a bad cache.

Consult the caching document to learn more.

Workflows

Workflows provide a means to define a collection of jobs and their run order. If at any point in your build you see a step where two jobs could happily run independent of one another, workflows may be helpful. Workflows also provide several other features to augment and improve your build configuration. Read more about workflows in the workflow documentation.

Note: Workflows are available to all plans, but running jobs concurrently assumes that your plan provides multiple machines to execute on.

version: 2.1
jobs: # here we define two jobs: "build" and "test"
  build:
    docker: # the docker executor is used
      - image: cimg/<language>:<version TAG> # An example docker image
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
    steps:
      - checkout # Pulls code down from your VCS
      - run: <command> # An example command
  test:
    docker: # same as previous docker key.
      - image: cimg/<language>:<version TAG>
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
    steps:
      - checkout
      - run: <command>
workflows: # Here we can orchestrate our jobs into a workflow
  version: 2
  build_and_test: # A single workflow named "build_and_test"
    jobs: # we run our `build` job and `test` job concurrently.
      - build
      - test

You can view more examples of workflows in the CircleCI demo workflows repo.

Workspaces

Note: Using workspaces presumes that you are also using workflows.

Workspaces are used to pass along data that is unique to a run and is needed for downstream jobs. So, if you are using workflows, a job run earlier in your build might fetch data and then make it available later for jobs that run later in a build.

To persist data from a job and make it available to downstream jobs via the attach_workspace key, configure the job to use the persist_to_workspace key. Files and directories named in the paths: property of persist_to_workspace will be uploaded to the workflow’s temporary workspace relative to the directory specified with the root key. The files and directories are then uploaded and made available for subsequent jobs (and re-runs of the workflow) to use.

Read more about how to use workspaces in the workflows document.

Parallelism

Note: Your CircleCI plan determines what level of parallelism you can use in your builds (1x, 2x, 4x, etc)

If your project has a large test suite, you can configure your build to use parallelism together with either CircleCI’s test splitting functionality or a third party application or library to split your tests across multiple machines. CircleCI supports automatic test allocation across machines on a file-basis, however, you can also manually customize how tests are allocated.

# ~/.circleci/config.yml
version: 2
jobs:
  docker:
    - image: cimg/<language>:<version TAG>
      auth:
        username: mydockerhub-user
        password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
  test:
    parallelism: 4

Read more in-depth about splitting tests in our document on parallelism.

Resource class

Note: An eligible plan is required to use the resource_class feature on Cloud. If you are on a container-based plan you will need to open a support ticket to enable this feature on your account. Resource class options for self hosted installations are set by system administrators.

Using resource_class, it is possible to configure CPU and RAM resources for each job. For Cloud, see this table for a list of available classes, and for self hosted installations contact your system administrator for a list.

Below is an example use case of the resource_class feature.

jobs:
  build:
    docker:
      - image: buildpack-deps:trusty
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
    environment:
      FOO: bar
    parallelism: 3
    resource_class: large # implements a machine with 4 vCPUS and 8gb of ram.
    steps:
      - run: make test
      - run: make

Docker layer caching

Note: The Performance plan is required to use Docker Layer Caching. If you are on the container-based plan you will need to upgrade to the Performance plan to enable DLC for your organization.

DLC is a feature that can help to reduce the build time of a Docker image in your build. Docker Layer Caching is useful if you find yourself frequently building Docker images as a regular part of your CI/CD process.

DLC is similar to caching dependencies mentioned above in that it saves the image layers that you build within your job, making them available on subsequent builds.

version: 2
jobs:
 build:
    docker:
      - image: circleci/node:14.17.3-buster-browsers # DLC does nothing here, its caching depends on commonality of the image layers.
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
    steps:
      - checkout
      - setup_remote_docker:
          docker_layer_caching: true # DLC will explicitly cache layers here and try to avoid rebuilding.
      - run: docker build .

Learn more about Docker Layer Caching