ci.yml

name: CI

on:
  pull_request:
  workflow_call:
    inputs:
      # comes from cargo-dist workflow call
      plan:
        required: true
        type: string
  workflow_dispatch:
  push:
    branches:
      - main
      - release-*

jobs:
  workspace-hack-check:
    name: Check workspace-hack
    runs-on: warp-ubuntu-latest-x64-4x
    env:
      RUSTFLAGS: -D warnings
    steps:
      - uses: actions/checkout@v4
      - uses: dtolnay/rust-toolchain@stable
      - name: Install cargo-hakari
        uses: taiki-e/install-action@v2
        with:
          tool: cargo-hakari
      - name: Check workspace-hack Cargo.toml is up-to-date
        run: cargo hakari generate --diff
      - name: Check all crates depend on workspace-hack
        run: cargo hakari manage-deps --dry-run
  build-and-test:
    name: Build and test (${{ matrix.os }})
    runs-on: ${{ matrix.os }}
    permissions:
      contents: read
      packages: read
    timeout-minutes: 45
    strategy:
      fail-fast: false
      matrix:
        os: [warp-ubuntu-latest-x64-16x]
    env:
      RUST_BACKTRACE: full
    steps:
      # Disabled as uring is not used in production yet.
      # - name: Install liburing
      #   run: sudo apt-get update && sudo apt-get install -y liburing-dev

      - uses: actions/checkout@v4

      - name: Install Rust toolchain
        uses: actions-rust-lang/setup-rust-toolchain@v1
        with:
          components: clippy
          rustflags: ""
          cache: false

      - name: Setup Rust Caching
        uses: WarpBuilds/rust-cache@v2 # a fork of Swatinem/rust-cache@v2 that uses warpbuild cache
        with:
          prefix-key: "v1-rust"
          cache-on-failure: "true"

      - name: Disk usage after cache restore
        run: df -h / && du -sh target/ 2>/dev/null || true

      - name: Install cargo-hakari
        uses: taiki-e/install-action@v2
        with:
          tool: cargo-hakari

      - name: Disable hakari
        run: cargo hakari disable

      - name: Install protoc
        uses: ./.github/actions/install-protoc

      - name: Install nextest
        uses: taiki-e/install-action@v2
        with:
          tool: nextest@0.9.98

      - name: Setup just
        uses: extractions/setup-just@v3
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - name: Check Cargo.lock file is updated
        run: cargo update -w --locked

      - name: Run lint
        run: just lint

      - name: Disk usage after lint
        run: df -h / && du -sh target/ 2>/dev/null || true

      - name: Run test
        run: just test
        env:
          LOCAL_CLUSTER_RUNNER_FORWARD_LOGS: "true"
          LOCAL_CLUSTER_RUNNER_RETAIN_TEMPDIR: "true"
          LOCAL_CLUSTER_RUNNER_TEMPDIR: ${{ runner.temp }}/lcr

      - name: Disk usage after test
        if: always()
        run: df -h / && du -sh target/ 2>/dev/null || true

      - name: Run doctest
        run: just doctest

      - uses: actions/upload-artifact@v4
        if: always()
        with:
          name: local-cluster-runner
          path: ${{ runner.temp }}/lcr

      - name: Disk usage before cache save
        if: always()
        run: df -h / && du -sh target/ 2>/dev/null || true

  docker:
    name: Create docker image
    uses: ./.github/workflows/docker.yml
    with:
      uploadImageAsTarball: "restate.tar"
      platforms: linux/amd64
      # additional features added for CI validation builds only
      features: metadata-api

  sdk-java:
    name: Run SDK-Java integration tests
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    secrets: inherit
    needs: docker
    uses: restatedev/sdk-java/.github/workflows/integration.yaml@main
    with:
      restateCommit: ${{ github.event.pull_request.head.sha || github.sha }}
      serviceImage: "ghcr.io/restatedev/test-services-java:main"

  sdk-java-journal-table-v2:
    name: Run SDK-Java integration tests with Journal Table v2
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    secrets: inherit
    needs: docker
    uses: restatedev/sdk-java/.github/workflows/integration.yaml@main
    with:
      restateCommit: ${{ github.event.pull_request.head.sha || github.sha }}
      serviceImage: "ghcr.io/restatedev/test-services-java:main"
      testArtifactOutput: sdk-java-integration-test-journal-table-v2
      envVars: |
        RESTATE_INTERNAL_FORCE_MIN_RESTATE_VERSION=1.6.0-dev

  sdk-java-vqueues:
    name: Vqueues Run SDK-Java integration tests
    # for the time being we don't expect all tests to pass
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    secrets: inherit
    needs: docker
    uses: restatedev/sdk-java/.github/workflows/integration.yaml@main
    with:
      restateCommit: ${{ github.event.pull_request.head.sha || github.sha }}
      serviceImage: "ghcr.io/restatedev/test-services-java:main"
      testArtifactOutput: sdk-java-integration-test-vqueues
      envVars: |
        RESTATE_EXPERIMENTAL_ENABLE_VQUEUES=true
      # we don't expect all tests to pass, that's why we continue on error
      continueOnError: true

  sdk-python:
    name: Run SDK-Python integration tests
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    secrets: inherit
    needs: docker
    uses: restatedev/sdk-python/.github/workflows/integration.yaml@main
    with:
      restateCommit: ${{ github.event.pull_request.head.sha || github.sha }}
      serviceImage: "ghcr.io/restatedev/test-services-python:main"

  sdk-go:
    name: Run SDK-Go integration tests
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    secrets: inherit
    needs: docker
    uses: restatedev/sdk-go/.github/workflows/integration.yaml@main
    with:
      restateCommit: ${{ github.event.pull_request.head.sha || github.sha }}

  sdk-typescript:
    name: Run SDK-Typescript integration tests
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    secrets: inherit
    needs: docker
    uses: restatedev/sdk-typescript/.github/workflows/integration.yaml@main
    with:
      restateCommit: ${{ github.event.pull_request.head.sha || github.sha }}
      serviceImage: "ghcr.io/restatedev/test-services-node:main"

  sdk-rust:
    name: Run SDK-Rust integration tests
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    secrets: inherit
    needs: docker
    uses: restatedev/sdk-rust/.github/workflows/integration.yaml@main
    with:
      restateCommit: ${{ github.event.pull_request.head.sha || github.sha }}

  e2e:
    name: Run E2E tests
    runs-on: warp-ubuntu-latest-x64-4x
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    needs: docker
    steps:
      - name: Set up Docker containerd snapshotter
        uses: docker/setup-docker-action@v4
        with:
          version: "v28.5.2"
          set-host: true
          daemon-config: |
            {
              "features": {
                "containerd-snapshotter": true
              }
            }

      ### Download the Restate container image, if needed
      - name: Download restate snapshot from in-progress workflow
        uses: actions/download-artifact@v4
        with:
          name: restate.tar
      - name: Install restate snapshot
        run: |
          output=$(docker load --input restate.tar | head -n 1)
          docker tag "${output#*: }" "localhost/restatedev/restate-commit-download:latest"
          docker image ls -a

      ### Run e2e tests
      - name: Run E2E tests
        uses: restatedev/e2e@main
        with:
          testArtifactOutput: e2e-test-report
          restateContainerImage: localhost/restatedev/restate-commit-download:latest
          # RESTATE_WORKER__INVOKER__experimental_features_allow_protocol_v6=true is needed for forward compatibility
          # tests 1.6 -> 1.5. The e2e tests create a v1.5 Restate server which gets the same env variables passed in,
          # and we need to enable the protocol v6 which is being used by 1.6.
          # todo remove once we use v1.6 as the LAST_COMPATIBLE_RESTATE_SERVER_VERSION in the e2e tests
          envVars: |
            RESTATE_WORKER__INVOKER__experimental_features_allow_protocol_v6=true

  e2e-enable-journal-table-v2:
    name: Run E2E tests with Journal Table V2 feature
    runs-on: warp-ubuntu-latest-x64-4x
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    needs: docker
    steps:
      - name: Set up Docker containerd snapshotter
        uses: docker/setup-docker-action@v4
        with:
          version: "v28.5.2"
          set-host: true
          daemon-config: |
            {
              "features": {
                "containerd-snapshotter": true
              }
            }

      ### Download the Restate container image, if needed
      - name: Download restate snapshot from in-progress workflow
        uses: actions/download-artifact@v4
        with:
          name: restate.tar
      - name: Install restate snapshot
        run: |
          output=$(docker load --input restate.tar | head -n 1)
          docker tag "${output#*: }" "localhost/restatedev/restate-commit-download:latest"
          docker image ls -a

      ### Run e2e tests
      - name: Run E2E tests
        uses: restatedev/e2e@main
        with:
          testArtifactOutput: e2e-journal-table-v2-test-report
          restateContainerImage: localhost/restatedev/restate-commit-download:latest
          envVars: |
            RESTATE_INTERNAL_FORCE_MIN_RESTATE_VERSION=1.6.0-dev
          # Ignore forward compatibility tests as long as the previous version is not at least v1.6.0-dev. Otherwise,
          # the previous version will hang at applying the version barrier with version 1.6.0-dev.
          exclusions: |
            exclusions:
              "versionCompat":
                - "dev.restate.sdktesting.tests.ForwardCompatibilityTest.*"

  e2e-vqueues:
    name: Run E2E tests with vqueues
    runs-on: warp-ubuntu-latest-x64-4x
    # we don't expect all tests to pass, that's why we continue on error
    continue-on-error: true
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    needs: docker
    steps:
      - name: Set up Docker containerd snapshotter
        uses: docker/setup-docker-action@v4
        with:
          version: "v28.5.2"
          set-host: true
          daemon-config: |
            {
              "features": {
                "containerd-snapshotter": true
              }
            }

      ### Download the Restate container image, if needed
      - name: Download restate snapshot from in-progress workflow
        uses: actions/download-artifact@v4
        with:
          name: restate.tar
      - name: Install restate snapshot
        run: |
          output=$(docker load --input restate.tar | head -n 1)
          docker tag "${output#*: }" "localhost/restatedev/restate-commit-download:latest"
          docker image ls -a

      ### Run e2e tests with vqueues enabled.
      - name: Run E2E tests
        uses: restatedev/e2e@main
        with:
          testArtifactOutput: e2e-test-vqueues-report
          restateContainerImage: localhost/restatedev/restate-commit-download:latest
          envVars: |
            RESTATE_EXPERIMENTAL_ENABLE_VQUEUES=true
          # We will never support the forward compatibility tests for v1.5 when using vqueues.
          # The backward compatibility tests are temporarily disabled until we support migration from v1.5 -> v1.6
          exclusions: |
            exclusions:
              "versionCompat":
                - "dev.restate.sdktesting.tests.ForwardCompatibilityTest.*"
                - "dev.restate.sdktesting.tests.BackwardCompatibilityTest.*"

  e2e-experimental-kafka-ingestion:
    name: Run E2E tests with experimental Kafka ingestion
    runs-on: warp-ubuntu-latest-x64-4x
    # we don't expect all tests to pass, that's why we continue on error
    continue-on-error: true
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    needs: docker
    steps:
      - name: Set up Docker containerd snapshotter
        uses: docker/setup-docker-action@v4
        with:
          version: "v28.5.2"
          set-host: true
          daemon-config: |
            {
              "features": {
                "containerd-snapshotter": true
              }
            }

      ### Download the Restate container image, if needed
      - name: Download restate snapshot from in-progress workflow
        uses: actions/download-artifact@v4
        with:
          name: restate.tar
      - name: Install restate snapshot
        run: |
          output=$(docker load --input restate.tar | head -n 1)
          docker tag "${output#*: }" "localhost/restatedev/restate-commit-download:latest"
          docker image ls -a

      ### Run e2e tests with experimental kafka ingestion enabled.
      - name: Run E2E tests
        uses: restatedev/e2e@main
        with:
          testArtifactOutput: e2e-test-experimental-kafka-ingestion-report
          restateContainerImage: localhost/restatedev/restate-commit-download:latest
          envVars: |
            RESTATE_experimental_kafka_batch_ingestion=true

  e2e-experimental-shuffler:
    name: Run E2E tests with experimental Shuffler
    runs-on: warp-ubuntu-latest-x64-4x
    # we don't expect all tests to pass, that's why we continue on error
    continue-on-error: true
    permissions:
      contents: read
      issues: read
      checks: write
      pull-requests: write
      actions: read
    needs: docker
    steps:
      - name: Set up Docker containerd snapshotter
        uses: docker/setup-docker-action@v4
        with:
          version: "v28.5.2"
          set-host: true
          daemon-config: |
            {
              "features": {
                "containerd-snapshotter": true
              }
            }

      ### Download the Restate container image, if needed
      - name: Download restate snapshot from in-progress workflow
        uses: actions/download-artifact@v4
        with:
          name: restate.tar
      - name: Install restate snapshot
        run: |
          output=$(docker load --input restate.tar | head -n 1)
          docker tag "${output#*: }" "localhost/restatedev/restate-commit-download:latest"
          docker image ls -a

      ### Run e2e tests with experimental shuffler enabled.
      - name: Run E2E tests
        uses: restatedev/e2e@main
        with:
          testArtifactOutput: e2e-test-experimental-shuffler-report
          restateContainerImage: localhost/restatedev/restate-commit-download:latest
          envVars: |
            RESTATE_experimental_shuffler_batch_ingestion=true

  jepsen:
    if: github.event.repository.fork == false && github.event.pull_request.head.repo.full_name == 'restatedev/restate' && github.ref == 'refs/heads/main'
    runs-on: warp-ubuntu-latest-arm64-4x
    name: Run Jepsen tests
    env:
      CLUSTER_NAME: restatedev-restate-${{ github.event.pull_request.number != null && format('pr{0}', github.event.pull_request.number) || format('run{0}', github.run_id) }}
    permissions:
      actions: read
      checks: write
      contents: read
      id-token: write # NB: can obtain OIDC tokens on behalf of this repository!
      issues: read
      pull-requests: write

    steps:
      - uses: aws-actions/configure-aws-credentials@v4
        with:
          aws-region: us-east-1 # co-locate with WarpBuild worker
          role-to-assume: "arn:aws:iam::339713187748:role/github-restatedev-restate-actions-role"
      - name: Setup Jepsen cluster ${{ env.CLUSTER_NAME }}
        uses: restatedev/jepsen/.github/actions/setup@main
        with:
          ref: ${{ inputs.ref }}
          clusterName: ${{ env.CLUSTER_NAME }}
          bucketName: "restate-jepsen-test-clusters-us-east-1"
      - name: Drop AWS credentials
        run: |
          echo "AWS_ACCESS_KEY_ID=" >> $GITHUB_ENV
          echo "AWS_SECRET_ACCESS_KEY=" >> $GITHUB_ENV
          echo "AWS_SESSION_TOKEN=" >> $GITHUB_ENV

      - name: Wait for Docker image artifact
        id: wait-for-docker
        uses: fountainhead/action-wait-for-check@v1.2.0
        with:
          token: ${{ secrets.GITHUB_TOKEN }}
          checkName: "Create docker image / build-and-push-image"
          ref: ${{ github.event.pull_request.head.sha || github.sha }}
          timeoutSeconds: 1800
          intervalSeconds: 10

      - name: Run Jepsen tests
        uses: restatedev/jepsen/.github/actions/run-tests@main
        if: steps.wait-for-docker.outputs.conclusion == 'success'
        with:
          restateCommit: ${{ github.event.pull_request.head.sha || github.sha }}
          testConfig: '{ "workloads": "set-vo set-mds set-mds-s3", "nemeses": "partition-random-node", "duration": "60", "rate": "10", "concurrency": "5n", "testCount": "1" }'

      - name: Fail if no Docker artifact
        if: steps.wait-for-docker.outputs.conclusion != 'success'
        run: |
          echo "::error::Docker step did not complete?"
          exit 1

      - uses: aws-actions/configure-aws-credentials@v4
        with:
          aws-region: us-east-1
          role-to-assume: "arn:aws:iam::339713187748:role/github-restatedev-restate-actions-role"
        if: always()
      - name: Tear down Jepsen cluster ${{ env.CLUSTER_NAME }}
        uses: restatedev/jepsen/.github/actions/teardown@main
        if: always()
        with:
          clusterName: ${{ env.CLUSTER_NAME }}
          bucketName: "restate-jepsen-test-clusters-us-east-1"