Skip to main content

Cleaning up a persistent build environment

Abstract

You can configure the Bitrise CLI to run in agent mode: this allows cleaning up persistent build environments when running builds on self-hosted agents.

On Bitrise, a new virtual machine is created every time a build starts and it is destroyed when the build is finished. You can, however run Bitrise builds in a persistent build environment: for example, you can use our on-premise runner or run builds on an AWS EC2 instance. In such an environment, the products of one build might affect subsequent builds.

On self-hosted agents, one agent executes multiple builds. This allows sharing data between builds on the local filesystem, but it also requires care in order to avoid one build affecting another.

To avoid the problem, you can configure the Bitrise CLI to run in agent mode. Agent mode requires placing an agent-config.yml in the host machine's ~/.bitrise/ directory. In the file, you can specify which directories to clean up when starting a new build or at the end of a build. It also allows you to run your own custom scripts if you have a more advanced use case than a simple cleanup.

Configuring agent mode

  1. Add the agent-config.yml file to your ~/.bitrise/ directory.

  2. If you wish to configure the exact folders in which you want to do cleanup operations, define a bitrise_dirs property. The property overrides the default settings: for example, you can set a different path for the BITRISE_DEPLOY_DIR environment variable than the default.

  3. Under bitrise_dirs, define the directories that you wish to perform some action on at the start or at the end of builds in a KEY: path format. For example, you can define separate directories for all source code checkout, deployable artifacts, and deployable test result artifacts.

    # Customize the common Bitrise directories
    bitrise_dirs:
      # Root directory for all Bitrise data produced at runtime
      BITRISE_DATA_HOME_DIR: /opt/bitrise
      
      # Directory for source code checkout.
      BITRISE_SOURCE_DIR: /opt/bitrise/workspace/$BITRISE_APP_SLUG
      
      # Directory for deployable artifacts.
      BITRISE_DEPLOY_DIR: /opt/bitrise/$BITRISE_APP_SLUG/$BITRISE_BUILD_SLUG/artifacts
      
      # Directory for deployable test result artifacts.
      BITRISE_TEST_DEPLOY_DIR: /opt/bitrise/$BITRISE_APP_SLUG/$BITRISE_BUILD_SLUG/test_results
    
    

    Multiple projects of the same Workspace

    Don’t forget that multiple projects of the same Workspace could run on the same agent. Make sure to always include $BITRISE_APP_SLUG in the directory hierarchy to separate projects and their source code checkouts.

    # wrong:
    BITRISE_SOURCE_DIR: /opt/bitrise/workspace
    # correct:
    BITRISE_SOURCE_DIR: /opt/bitrise/$BITRISE_APP_SLUG/workspace
  4. Add a hooks property to the agent-config.yml file. This property will define the actions to perform at the start or end of builds. It has four different parameters:

    • cleanup_on_build_start: Defines a directory that is cleaned up when a new build is started.

    • cleanup_on_build_end: Defines a directory that is cleaned up whenever a build is finished. Not guaranteed to run in the case of build failure.

    • do_on_build_start: Defines a custom script that runs when a new build is started.

    • do_on_build_end: Defines a custom script that runs when a build is finished. Not guaranteed to run in the case of build failure.

    Nested Workflows

    A Script Step in a Workflow can execute bitrise run nested_workflow and trigger a nested workflow. This nested Workflow inherits all the envs and parameters of the parent Workflow, and the parent Workflow waits for the completion of the nested Workflow. When the nested Workflow is launched this way, hooks and directory cleanups are not executed in this process to avoid unexpected behavior.

  5. To test the agent mode, start a build. The log should show the following message at the top:

    Running in agent mode
    Config file: .bitrise/agent-config.yml

Common configuration examples

Example 1. Fully isolated builds

To minimize state and maximize reliability, each build is assigned unique directories. Source code is checked out from scratch in each build.

bitrise_dirs:
  BITRISE_DATA_HOME_DIR: /opt/bitrise
  BITRISE_SOURCE_DIR: /opt/bitrise/$BITRISE_APP_SLUG/$BITRISE_BUILD_SLUG/workspace
  BITRISE_DEPLOY_DIR: /opt/bitrise/$BITRISE_APP_SLUG/$BITRISE_BUILD_SLUG/artifacts
  BITRISE_TEST_DEPLOY_DIR: /opt/bitrise/$BITRISE_APP_SLUG/$BITRISE_BUILD_SLUG/test_results

hooks:
  # Since dirs are unique to each build, there is nothing to clean up here:
  cleanup_on_build_start: []
  
  # Clean up everything after the build ends:
  cleanup_on_build_end:
    - $BITRISE_SOURCE_DIR
    - $BITRISE_DEPLOY_DIR
    - $BITRISE_TEST_DEPLOY_DIR

Example 2. Shared source code directory for warm checkouts

For faster source code checkouts, it’s possible to reuse a previous build’s source code directory.

bitrise_dirs:
  BITRISE_DATA_HOME_DIR: /opt/bitrise
  # Use a warm clone of the repo for all builds:
  BITRISE_SOURCE_DIR: /opt/bitrise/$BITRISE_APP_SLUG/workspace
  # For artifacts and test results, it's still a good idea to place them into unique dirs
  BITRISE_DEPLOY_DIR: /opt/bitrise/$BITRISE_APP_SLUG/$BITRISE_BUILD_SLUG/artifacts
  BITRISE_TEST_DEPLOY_DIR: /opt/bitrise/$BITRISE_APP_SLUG/$BITRISE_BUILD_SLUG/test_results

hooks:
  # Since dirs are unique to each build, there is nothing to clean up here:
  cleanup_on_build_start: []
  
  # Optional: clean up artifacts and test results.
  # Note that $BITRISE_SOURCE_DIR is NOT clean up here!
  cleanup_on_build_end:
    - $BITRISE_DEPLOY_DIR
    - $BITRISE_TEST_DEPLOY_DIR