ci.yaml

name: CI

on:
  push:
    branches:
      # only on merges to master branch
      - master
      # and version branches, which only include minor versions (eg: v3.4)
      - v[0-9]+.[0-9]+
    tags:
      # also version tags, which include bugfix releases (eg: v3.4.0)
      - v[0-9]+.[0-9]+.[0-9]+
  pull_request:
    types: [opened, reopened, synchronize]
    branches:
      # Only for PRs targeting those branches
      - master
      - v[0-9]+.[0-9]+
  schedule:
    # run every night at midnight
    - cron:  '0 0 * * *'

jobs:
  # TODO: Fix the required checks!
  #       When the pre_job triggers and skips builds, it prevents merging the PR because
  #       the required checks are reported as skipped instead of passed.
  # Special job which automatically cancels old runs for the same branch, prevents runs for the
  # same file set which has already passed, etc.
  pre_job:
    name: Skip Duplicate Jobs Pre Job
    runs-on: ubuntu-20.04
    outputs:
      should_skip: ${{ steps.skip_check.outputs.should_skip }}
    steps:
      - id: skip_check
        uses: fkirc/skip-duplicate-actions@4c656bbdb6906310fa6213604828008bc28fe55d # v3.3.0
        with:
          cancel_others: 'true'
          github_token: ${{ github.token }}

  # Lint checks which don't depend on any service containes, etc. to be running.
  lint-checks:
    needs: pre_job
    # NOTE: We always want to run job on master since we run some additional checks there (code
    # coverage, etc)
    # if: ${{ needs.pre_job.outputs.should_skip != 'true' || github.ref == 'refs/heads/master' }}
    name: '${{ matrix.make.name }} - Python ${{ matrix.python.version-short }}'
    runs-on: ubuntu-20.04
    strategy:
      fail-fast: false
      matrix:
        # NOTE: To speed the CI run, we split unit and integration tests into multiple jobs where
        # each job runs subset of tests.
        # NOTE: We need to use full Python version as part of Python deps cache key otherwise
        # setup virtualenv step will fail.
        python:
          - {version-short: '3.8', version: '3.8.10'}
          - {version-short: '3.9', version: '3.9.14'}
          - {version-short: '3.10', version: '3.10.15'}
          - {version-short: '3.11', version: '3.11.10'}
        make:
          - name: 'Lint Checks (black, flake8, etc.)'
            task: 'ci-checks'
          - name: 'Compile (pip deps, pylint, etc.)'
            task: 'ci-compile'
    env:
      TASK: '${{ matrix.make.task }}'
      COLUMNS: '120'
      PYLINT_CONCURRENCY: '6'
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
      - name: Custom Environment Setup
        run: |
          ./scripts/github/setup-environment.sh
      - name: 'Set up Python (${{ matrix.python.version }}) and Cache Deps'
        uses: ./.github/actions/setup-python
        with:
          python-version: '${{ matrix.python.version }}'
      - name: Cache and Install APT Dependencies
        uses: ./.github/actions/apt-packages
      - name: Install virtualenv
        run: |
          ./scripts/github/install-virtualenv.sh
      - name: Install requirements
        run: |
          ./scripts/ci/install-requirements.sh
      - name: Print versions
        run: |
          ./scripts/ci/print-versions.sh
      - name: make
        # use: script -e -c to print colors
        timeout-minutes: 7
        run: |
          script -e -c "make ${TASK}"
      - name: Nightly
        # Run any additional nightly checks only as part of a nightly (cron) build
        if: "${{ env.IS_NIGHTLY_BUILD == 'yes' }}"
        run: |
          ./scripts/ci/run-nightly-make-task-if-exists.sh "${TASK}"

  self-check:
    needs: pre_job
    # FIXME: dropping the repetition in this name requires changing required checks on GitHub
    name: 'Self-check on Python ${{ matrix.python.version-short }} - Python ${{ matrix.python.version-short }}'
    runs-on: ubuntu-20.04
    strategy:
      fail-fast: false
      matrix:
        python:
          - {version-short: '3.8', version: '3.8.10'}
    services:
      mongo:
        image: mongo:7.0
        ports:
          - 27017:27017

      rabbitmq:
        image: rabbitmq:3.8-management
        options: >-
          --name rabbitmq
        ports:
            - 5671:5671/tcp   # AMQP SSL port
            - 5672:5672/tcp   # AMQP standard port
            - 15672:15672/tcp # Management: HTTP, CLI

      redis:
        # Docker Hub image
        image: redis
        # Set health checks to wait until redis has started
        options: >-
          --name "redis"
          --health-cmd "redis-cli ping"
          --health-interval 10s
          --health-timeout 5s
          --health-retries 5
        ports:
          - 6379:6379/tcp
    env:
      # CI st2.conf (with ST2_CI_USER user instead of stanley)
      ST2_CONF: 'conf/st2.ci.conf'

      # Name of the user who is running the CI (on GitHub Actions this is 'runner')
      ST2_CI_USER: 'runner'

      # GitHub is juggling how to set vars for multiple shells. Protect our PATH assumptions.
      PATH: /home/runner/.local/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin

      # Space separated list of tests to be skipped if the self-check is running in GitHub Actions
      TESTS_TO_SKIP: "tests.test_quickstart_rules tests.test_run_pack_tests_tool"
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
      - name: Custom Environment Setup
        run: |
          ./scripts/github/setup-environment.sh
      - name: 'Set up Python (${{ matrix.python.version }}) and Cache Deps'
        uses: ./.github/actions/setup-python
        with:
          python-version: '${{ matrix.python.version }}'
      - name: Cache and Install APT Dependencies
        uses: ./.github/actions/apt-packages
      - name: Install virtualenv
        run: |
          ./scripts/github/install-virtualenv.sh
      - name: Install requirements
        run: |
          ./scripts/ci/install-requirements.sh
      - name: Setup Tests
        run: |
          # prep a ci-specific dev conf file that uses runner instead of stanley
          # this user is the username of the user in GitHub actions, used for SSH, etc during
          # integration tests (important)
          cp conf/st2.dev.conf "${ST2_CONF}"
          sed -i -e "s,/home/vagrant/.ssh/stanley_rsa,/home/stanley/.ssh/stanley_rsa," "${ST2_CONF}"

          sudo -E ./scripts/ci/add-itest-user-key.sh
      - name: Permissions Workaround
        run: |
          sudo ST2_CI_REPO_PATH="${ST2_CI_REPO_PATH}" scripts/ci/permissions-workaround.sh
      - name: Reconfigure RabbitMQ
        # bitnami image allows (see bitnami/rabbitmq readme):
        # Here we're copying a rabbitmq.config file which won't do anything.
        # We need to switch to custom.conf or advanced.config.
        timeout-minutes: 2  # may die if rabbitmq fails to start
        run: |
          ./scripts/github/configure-rabbitmq.sh
      - name: Print versions
        run: |
          ./scripts/ci/print-versions.sh
      - name: make
        timeout-minutes: 14  # may die if rabbitmq fails to start
        # use: script -e -c to print colors
        run: |
          script -e -c "make .ci-prepare-integration" && exit 0
      - name: Extend the path for upcoming tasks
        # pants uses PEP 660 editable wheels to add our code to the virtualenv.
        # But PEP 660 editable wheels do not include 'scripts'.
        # https://peps.python.org/pep-0660/#limitations
        # So, we need to include each bin dir in PATH instead of virtualenv/bin.
        run: |
          for component_bin in ${GITHUB_WORKSPACE}/st2*/bin; do
          echo ${component_bin} | tee -a $GITHUB_PATH
          done
          echo ${GITHUB_WORKSPACE}/virtualenv/bin | tee -a $GITHUB_PATH
      - name: Create symlinks to find the binaries when running st2 actions
        # st2 is actually a console_script entry point, not just a 'script'
        # so it IS included in the virtualenv. But, st2-run-pack-tests might not be included.
        run: |
          ln -s ${GITHUB_WORKSPACE}/virtualenv/bin/st2 /usr/local/bin/st2
          ln -s ${GITHUB_WORKSPACE}/st2common/bin/st2-run-pack-tests /usr/local/bin/st2-run-pack-tests
      - name: Install st2client
        timeout-minutes: 5
        run: |
          cd ./st2client
          pip3 install --upgrade pip
          python3 setup.py develop
      - name: Run self-verification script
        env:
          ST2_CONF: /home/runner/work/st2/st2/conf/st2.ci.conf
        run: |
          sudo -E ST2_AUTH_TOKEN=$(st2 auth testu -p 'testp' -t) PATH=${PATH} st2common/bin/st2-self-check
      - name: Compress Service Logs Before upload
        if: ${{ failure() }}
        run: |
          ./tools/launchdev.sh stop # stop st2 before collecting logs
          tar cvzpf logs.tar.gz logs/*
      - name: Upload StackStorm services Logs
        if: ${{ failure() }}
        uses: actions/upload-artifact@v4
        with:
          name: logs-py${{ matrix.python.version }}
          path: logs.tar.gz
          retention-days: 7

  unit-tests:
    needs: pre_job
    # NOTE: We always want to run job on master since we run some additional checks there (code
    # coverage, etc)
    # NB: disabled. See TODO above pre_job
    # if: ${{ needs.pre_job.outputs.should_skip != 'true' || github.ref == 'refs/heads/master' }}
    name: '${{ matrix.make.name }} - Python ${{ matrix.python.version-short }}'
    runs-on: ubuntu-20.04
    strategy:
      fail-fast: false
      matrix:
        # NOTE: To speed the CI run, we split unit and integration tests into multiple jobs where
        # each job runs subset of tests.
        python:
          - {version-short: '3.8', version: '3.8.10'}
          - {version-short: '3.9', version: '3.9.14'}
          - {version-short: '3.10', version: '3.10.15'}
          - {version-short: '3.11', version: '3.11.10'}
        make:
          - name: 'Unit Tests (chunk 1)'
            task: 'ci-unit'
            shard: {k: 0, n: 2}
          - name: 'Unit Tests (chunk 2)'
            task: 'ci-unit'
            shard: {k: 1, n: 2}

          # This job is slow so we only run in on a daily basis
          # - name: 'Micro Benchmarks'
          #   task: 'micro-benchmarks'
          #   shard: {k: 0, n: 1}
    services:
      mongo:
        image: mongo:7.0
        ports:
          - 27017:27017
      redis:
        # Docker Hub image
        image: redis
        # Set health checks to wait until redis has started
        options: >-
          --name "redis"
          --health-cmd "redis-cli ping"
          --health-interval 10s
          --health-timeout 5s
          --health-retries 5
        ports:
          - 6379:6379/tcp


      rabbitmq:
        image: rabbitmq:3.8-management
        options: >-
          --name rabbitmq
        ports:
          - 5671:5671/tcp   # AMQP SSL port
          - 5672:5672/tcp   # AMQP standard port
          - 15672:15672/tcp # Management: HTTP, CLI

    env:
      TASK: '${{ matrix.make.task }}'
      PYTHON_VERSION_SHORT: '${{ matrix.python.version-short }}'
      NODE_TOTAL: '${{ matrix.make.shard.n }}'
      NODE_INDEX: '${{ matrix.make.shard.k }}'

      # We need to explicitly specify terminal width otherwise some CLI tests fail on container
      # environments where small terminal size is used.
      COLUMNS: '120'

      # CI st2.conf (with ST2_CI_USER user instead of stanley)
      ST2_CONF: 'conf/st2.ci.conf'

      # Tell StackStorm that we are indeed in CI mode, previously we hard coded a Travis specific
      # environment variable in our test code, making it a PITA when we switch CI providers.
      # Now, we simply set this environment varible here in the CI portion of our testing and
      # it avoids any CI provider type lock-in.
      ST2_CI: 'true'

      # Name of the user who is running the CI (on GitHub Actions this is 'runner')
      ST2_CI_USER: 'runner'

      # GitHub is juggling how to set vars for multiple shells. Protect our PATH assumptions.
      PATH: /home/runner/.local/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
      - name: Custom Environment Setup
        run: |
          ./scripts/github/setup-environment.sh
      - name: 'Set up Python (${{ matrix.python.version }}) and Cache Deps'
        uses: ./.github/actions/setup-python
        with:
          python-version: '${{ matrix.python.version }}'
      - name: Cache and Install APT Dependencies
        uses: ./.github/actions/apt-packages
      - name: Install virtualenv
        run: |
          ./scripts/github/install-virtualenv.sh
      - name: Install requirements
        run: |
          ./scripts/ci/install-requirements.sh
      - name: Setup Tests
        run: |
          # prep a ci-specific dev conf file that uses runner instead of stanley
          # this user is the username of the user in GitHub actions, used for SSH, etc during
          # integration tests (important)
          cp conf/st2.dev.conf "${ST2_CONF}" ; sed -i -e "s/stanley/${ST2_CI_USER}/" "${ST2_CONF}"

          sudo -E ./scripts/ci/add-itest-user-key.sh
      - name: Permissions Workaround
        run: |
          echo "$ST2_CI_REPO_PATH"
          sudo ST2_CI_REPO_PATH="${ST2_CI_REPO_PATH}" scripts/ci/permissions-workaround.sh
      - name: Reconfigure RabbitMQ
        if: "${{ env.TASK == 'ci-unit' || env.TASK == 'ci-integration' }}"
        # bitnami image allows (see bitnami/rabbitmq readme):
        # Here we're copying a rabbitmq.config file which won't do anything.
        # We need to switch to custom.conf or advanced.config.
        timeout-minutes: 2  # may die if rabbitmq fails to start
        run: |
          ./scripts/github/configure-rabbitmq.sh
      - name: Print versions
        run: |
          ./scripts/ci/print-versions.sh
      - name: make
        # use: script -e -c to print colors
        # TODO: Use dynamic timeout value based on the branch - for master we
        # need to use timeout x2 due to coverage overhead
        # timeout-minutes: 8
        timeout-minutes: 19
        run: |
          script -e -c "make ${TASK}"
      - name: Nightly
        # Run any additional nightly checks only as part of a nightly (cron) build
        if: "${{ env.IS_NIGHTLY_BUILD == 'yes' }}"
        run: |
          ./scripts/ci/run-nightly-make-task-if-exists.sh "${TASK}"
      - name: Codecov
        # NOTE: We only generate and submit coverage report for master and version branches and only when the build succeeds (default on GitHub Actions, this was not the case on Travis so we had to explicitly check success)
        if: "${{ success() && (env.ENABLE_COVERAGE == 'yes') && (env.PYTHON_VERSION_SHORT == '3.8')}}"
        run: |
          ./scripts/ci/submit-codecov-coverage.sh
        env:
          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

  integration-tests:
    needs: pre_job
    # NOTE: We always want to run job on master since we run some additional checks there (code
    # coverage, etc)
    # if: ${{ needs.pre_job.outputs.should_skip != 'true' || github.ref == 'refs/heads/master' }}
    name: '${{ matrix.make.name }} - Python ${{ matrix.python.version-short }}'
    runs-on: ubuntu-20.04
    strategy:
      fail-fast: false
      matrix:
        # NOTE: To speed the CI run, we split unit and integration tests into multiple jobs where
        # each job runs subset of tests.
        python:
          - {version-short: '3.8', version: '3.8.10'}
          - {version-short: '3.9', version: '3.9.14'}
          - {version-short: '3.10', version: '3.10.15'}
          - {version-short: '3.11', version: '3.11.10'}
        make:
          # We run pack tests here since they rely on some integration tests set
          # up (aka stanley user being present, etc.)
          - name: 'Pack Tests'
            task: 'ci-packs-tests'
            shard: {k: 0, n: 1}
          - name: 'Integration Tests (chunk 1)'
            task: 'ci-integration'
            shard: {k: 0, n: 2}
          - name: 'Integration Tests (chunk 2)'
            task: 'ci-integration'
            shard: {k: 1, n: 2}
    services:
      mongo:
        image: mongo:7.0
        ports:
          - 27017:27017

      # In GHA, these services are started first before the code is checked out.
      # We use bitnami images to facilitate reconfiguring RabbitMQ during ci-integration tests.
      # We rely on custom config and SSL certs that are in the repo.
      # Many images require config in env vars (which we can't change during the test job)
      # or they require config in entrypoint args (which we can't override for GHA services)
      # bitnami builds ways to get config files from mounted volumes.
      rabbitmq:
        image: bitnami/rabbitmq:3.8
        volumes:
          - /home/runner/rabbitmq_conf:/bitnami/conf  # RABBITMQ_MOUNTED_CONF_DIR
        env:
          # tell bitnami/rabbitmq to enable this by default
          RABBITMQ_PLUGINS: rabbitmq_management
          RABBITMQ_USERNAME: guest
          RABBITMQ_PASSWORD: guest

        # These are strictly docker options, not entrypoint args (GHA restriction)
        options: >-
          --name rabbitmq

        ports:
          # These 6 ports are exposed by bitnami/rabbitmq (see https://www.rabbitmq.com/networking.html#ports)
          # host_port:container_port/protocol
          - 5671:5671/tcp   # AMQP SSL port
          - 5672:5672/tcp   # AMQP standard port
          - 15672:15672/tcp # Management: HTTP, CLI
          #- 15671:15671/tcp # Management: SSL port
          #- 25672:25672/tcp # inter-node or CLI
          #- 4369:4369/tcp   # epmd
          #

      redis:
        # Docker Hub image
        image: redis
        # Set health checks to wait until redis has started
        options: >-
          --name "redis"
          --health-cmd "redis-cli ping"
          --health-interval 10s
          --health-timeout 5s
          --health-retries 5
        ports:
          - 6379:6379/tcp

    env:
      TASK: '${{ matrix.make.task }}'
      PYTHON_VERSION_SHORT: '${{ matrix.python.version-short }}'
      NODE_TOTAL: '${{ matrix.make.shard.n }}'
      NODE_INDEX: '${{ matrix.make.shard.k }}'

      # We need to explicitly specify terminal width otherwise some CLI tests fail on container
      # environments where small terminal size is used.
      COLUMNS: '120'

      # CI st2.conf (with ST2_CI_USER user instead of stanley)
      ST2_CONF: 'conf/st2.ci.conf'

      # Tell StackStorm that we are indeed in CI mode, previously we hard coded a Travis specific
      # environment variable in our test code, making it a PITA when we switch CI providers.
      # Now, we simply set this environment varible here in the CI portion of our testing and
      # it avoids any CI provider type lock-in.
      ST2_CI: 'true'

      # Name of the user who is running the CI (on GitHub Actions this is 'runner')
      ST2_CI_USER: 'runner'

      # GitHub is juggling how to set vars for multiple shells. Protect our PATH assumptions.
      PATH: /home/runner/.local/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin

    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
      - name: Custom Environment Setup
        run: |
          ./scripts/github/setup-environment.sh
      - name: 'Set up Python (${{ matrix.python.version }}) and Cache Deps'
        uses: ./.github/actions/setup-python
        with:
          python-version: '${{ matrix.python.version }}'
      - name: Cache and Install APT Dependencies
        uses: ./.github/actions/apt-packages
      - name: Install virtualenv
        run: |
          ./scripts/github/install-virtualenv.sh
      - name: Install requirements
        run: |
          ./scripts/ci/install-requirements.sh
      - name: Setup Integration Tests
        run: |
          # prep a ci-specific dev conf file that uses runner instead of stanley
          # this user is the username of the user in GitHub actions, used for SSH, etc during
          # integration tests (important)
          cp conf/st2.dev.conf "${ST2_CONF}" ; sed -i -e "s/stanley/${ST2_CI_USER}/" "${ST2_CONF}"

          sudo -E ./scripts/ci/add-itest-user-key.sh
      - name: Permissions Workaround
        run: |
          echo "$ST2_CI_REPO_PATH"
          sudo ST2_CI_REPO_PATH="${ST2_CI_REPO_PATH}" scripts/ci/permissions-workaround.sh
      - name: Reconfigure RabbitMQ
        # bitnami image allows (see bitnami/rabbitmq readme):
        # Here we're copying a rabbitmq.config file which won't do anything.
        # We need to switch to custom.conf or advanced.config.
        timeout-minutes: 2  # may die if rabbitmq fails to start
        run: |
          ./scripts/github/configure-rabbitmq.sh
      - name: Print versions
        run: |

          ./scripts/ci/print-versions.sh
      - name: make
        #timeout-minutes: 7
        # TODO: Use dynamic timeout value based on the branch - for master we
        # need to use timeout x2 due to coverage overhead
        timeout-minutes: 14  # may die if rabbitmq fails to start
        # use: script -e -c to print colors
        run: |
          script -e -c "make ${TASK}" && exit 0
      - name: Codecov
        # NOTE: We only generate and submit coverage report for master and version branches and only when the build succeeds (default on GitHub Actions, this was not the case on Travis so we had to explicitly check success)
        if: "${{ success() && (env.ENABLE_COVERAGE == 'yes') && (env.TASK == 'ci-integration') && (env.PYTHON_VERSION_SHORT == '3.8')}}"
        run: |
          ./scripts/ci/submit-codecov-coverage.sh
        env:
          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
      - name: Compress Service Logs Before upload
        if: ${{ failure() && env.TASK == 'ci-integration' }}
        run: |
          ./tools/launchdev.sh stop # stop st2 before collecting logs
          tar cvzpf logs.tar.gz logs/*
      - name: Upload StackStorm services Logs
        if: ${{ failure() && env.TASK == 'ci-integration' }}
        uses: actions/upload-artifact@v4
        with:
          name: logs-py${{ matrix.python.version }}-pytest-${{ matrix.make.shard.k }}
          path: logs.tar.gz
          retention-days: 7

  slack-notification:
    name: Slack notification for failed master builds
    if: always()
    needs:
      - lint-checks
      - self-check
      - unit-tests
      - integration-tests
    runs-on: ubuntu-20.04
    steps:
      - name: Workflow conclusion
        # this step creates an environment variable WORKFLOW_CONCLUSION and is the most reliable way to check the status of previous jobs
        uses: technote-space/workflow-conclusion-action@v2
      - name: CI Run Failure Slack Notification
        if: ${{ env.WORKFLOW_CONCLUSION == 'failure' && github.ref == 'refs/heads/master' }}
        env:
          SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}
        uses: voxmedia/github-action-slack-notify-build@v1
        with:
          channel: development
          status: FAILED
          color: danger

      # HELPER FOR FUTURE DEVELOPERS:
      #  If your GitHub Actions job is failing and you need to debug it, by default there is
      #  no way to SSH into the container.
      #  The step below can be uncommeted and will stop here and allow you to SSH in.
      #  When this step is reached, simply refresh the GitHub Actions output for this build
      #  and this SSH command will be printed every 5 seconds to the output.
      #  Once you are done debugging in your SSH session, simply: touch /continue
      #  and this will continue the build.
      #
      # - name: Setup tmate session for debugging failed jobs (allows SSH into the container)
      #   uses: mxschmitt/action-tmate@v3
      #   if: "${{ failure() }}"
      #