run-performance.yaml

name: Run Performance test

on:
  workflow_dispatch: # manually triggered
    inputs:
      test_name:
        type: string
        description: Folder name under "performance" that contains yaml
        required: true
      js_ceramic_image:
        type: string
        description: Image for js-ceramic
        required: true
        default: ceramicnetwork/js-ceramic:latest
      rust_ceramic_image:
        type: string
        description: Image for rust-ceramic
        required: true
        default: public.ecr.aws/r5b3e0r5/3box/ceramic-one:latest
      scenario_image:
        type: string
        description: Image for scenario
        required: true
        default: public.ecr.aws/r5b3e0r5/3box/keramik-runner:latest

env:
  CARGO_TERM_COLOR: always
  SCCACHE_GHA_ENABLED: "true"
  RUSTC_WRAPPER: "sccache"

jobs:
  run-performance-test:
    name: Run Performance Test ${{ github.run_id }}
    runs-on: ubuntu-latest
    environment: cas-scaling-2024
    steps:
      - name: Checkout
        uses: actions/checkout@v3

      - name: Install yq
        run: |
          sudo apt-get update
          sudo apt-get install -y software-properties-common
          sudo add-apt-repository -y ppa:rmescandon/yq
          sudo apt-get update
          sudo apt-get install -y yq

      - name: Template network
        run: |
          set -exo pipefail
          export RUST_CERAMIC_IMAGE=${{ inputs.rust_ceramic_image }}
          export SCENARIO_IMAGE=${{ inputs.scenario_image }}
          export JS_CERAMIC_IMAGE=${{ inputs.js_ceramic_image }}
          export TEST_NAME=${{ inputs.test_name }}
          ls -l performance/$TEST_NAME
          export THIS_TEST=${{ inputs.test_name }}-${{ github.run_id }}
          mkdir -p performance/$THIS_TEST
          # set the network name yaml key to the test names
          yq e '.metadata.name = env(THIS_TEST)' performance/$TEST_NAME/network.yaml \
            > performance/$THIS_TEST/network.yaml
          cat performance/$THIS_TEST/network.yaml
          echo "THIS_TEST=$THIS_TEST" >> $GITHUB_ENV
          echo "THIS_TEST_NAMESPACE=keramik-${THIS_TEST}" >> $GITHUB_ENV
          echo "TEST_NAME=${TEST_NAME}" >> $GITHUB_ENV

      - name: Setup GKE auth
        uses: 'google-github-actions/auth@v1'
        with:
          credentials_json: ${{ secrets.GKE_SA_KEY }}

      - name: Get GKE credentials
        uses: 'google-github-actions/get-gke-credentials@v1'
        with:
          cluster_name: ${{ vars.GKE_CLUSTER }}
          location: ${{ vars.GKE_ZONE }}

      - name: Deploy network
        run: |
          kubectl apply -f performance/$THIS_TEST/network.yaml

      - name: Wait for bootstrap to complete
        timeout-minutes: 8
        run: |
          set -exo pipefail
          sleep 60
          kubectl wait --for=condition=ready \
            --timeout=240s \
            pod \
            -l app=ceramic \
            -n ${THIS_TEST_NAMESPACE}
          sleep 60
          kubectl wait --for=condition=complete \
            --timeout=120s \
            job/bootstrap \
            -n ${THIS_TEST_NAMESPACE}

      - name: Template simulation
        run: |
          set -exo pipefail
          yq e '.metadata.namespace = env(THIS_TEST_NAMESPACE)' performance/$TEST_NAME/simulation.yaml \
            > performance/$THIS_TEST/simulation.yaml
          cat performance/$THIS_TEST/simulation.yaml

      - name: Run simulation
        run: |
          set -exo pipefail
          kubectl apply -f performance/$THIS_TEST/simulation.yaml
          SIMULATION_RUNTIME=$(yq e '.spec.runTime' performance/$THIS_TEST/simulation.yaml)
          echo "SIMULATION_RUNTIME=${SIMULATION_RUNTIME}" >> $GITHUB_ENV

      - name: Notify Discord
        env:
          SIMULATION_COLOR: 3066993
        run: |
          set -exo pipefail
          export SIMULATION_STATUS_TAG="Simulation $THIS_TEST started"
          envsubst < notifications/notification-template.json  > message.json
          cat message.json
          curl -v -H "Content-Type: application/json" -X POST -d @./message.json "${{ secrets.DISCORD_WEBHOOK_URL_SUCCEEDED }}"

      - name: Wait for simulation to complete
        run: |
          set -exo pipefail
          # runtime is in minutes
          sleep $((${SIMULATION_RUNTIME} * 60))
          kubectl wait --for=condition=complete \
            --timeout=120s \
            job/simulate-manager \
            -n ${THIS_TEST_NAMESPACE}

      - name: Cleanup
        if: always()
        run: |
          kubectl delete -f performance/$THIS_TEST/network.yaml