Add DrainJobsMode (aka UpdateStrategy feature) (#2569)

.github/actions/execute-assert-arc-e2e/action.yaml

@@ -23,6 +23,14 @@ inputs:
   arc-controller-namespace:
     description: 'The namespace of the configured gha-runner-scale-set-controller'
     required: true
+  wait-to-finish:
+    description: 'Wait for the workflow run to finish'
+    required: true
+    default: "true"
+  wait-to-running:
+    description: 'Wait for the workflow run to start running'
+    required: true
+    default: "false"
 
 runs:
   using: "composite"
@@ -118,7 +126,36 @@ runs:
         | ${{steps.query_workflow.outputs.workflow_run_url}} |
         EOF
 
+    - name: Wait for workflow to start running
+      if: inputs.wait-to-running == 'true' && inputs.wait-to-finish == 'false'
+      uses: actions/github-script@v6
+      with:
+        script: |
+          function sleep(ms) {
+            return new Promise(resolve => setTimeout(resolve, ms))
+          }
+          const owner = '${{inputs.repo-owner}}'
+          const repo = '${{inputs.repo-name}}'
+          const workflow_run_id = ${{steps.query_workflow.outputs.workflow_run}}
+          const workflow_job_id = ${{steps.query_workflow.outputs.workflow_job}}
+          let count = 0
+          while (count++<10) {
+            await sleep(30 * 1000);
+            let getRunResponse = await github.rest.actions.getWorkflowRun({
+              owner: owner,
+              repo: repo,
+              run_id: workflow_run_id
+            })
+            console.log(`${getRunResponse.data.html_url}: ${getRunResponse.data.status} (${getRunResponse.data.conclusion})`);
+            if (getRunResponse.data.status == 'in_progress') {
+              console.log(`Workflow run is in progress.`)
+              return
+            }
+          }
+          core.setFailed(`The triggered workflow run didn't start properly using ${{inputs.arc-name}}`)
+
     - name: Wait for workflow to finish successfully
+      if: inputs.wait-to-finish == 'true'
       uses: actions/github-script@v6
       with:
         script: |
@@ -151,10 +188,15 @@ runs:
           }
           core.setFailed(`The triggered workflow run didn't finish properly using ${{inputs.arc-name}}`)
 
+    - name: cleanup
+      if: inputs.wait-to-finish == 'true'
+      shell: bash
+      run: |
+        helm uninstall ${{ inputs.arc-name }} --namespace ${{inputs.arc-namespace}} --debug
+        kubectl wait --timeout=10s --for=delete AutoScalingRunnerSet -n ${{inputs.arc-name}} -l app.kubernetes.io/instance=${{ inputs.arc-name }}
+
     - name: Gather logs and cleanup
       shell: bash
       if: always()
       run: |
-        helm uninstall ${{ inputs.arc-name }} --namespace ${{inputs.arc-namespace}} --debug
-        kubectl wait --timeout=10s --for=delete AutoScalingRunnerSet -n ${{inputs.arc-name}} -l app.kubernetes.io/instance=${{ inputs.arc-name }}
         kubectl logs deployment/arc-gha-runner-scale-set-controller -n ${{inputs.arc-controller-namespace}}

.github/workflows/gha-e2e-tests.yaml

@@ -710,3 +710,173 @@ jobs:
           arc-name: ${{steps.install_arc.outputs.ARC_NAME}}
           arc-namespace: "arc-runners"
           arc-controller-namespace: "arc-systems"
+
+  update-strategy-tests:
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    if: github.event_name != 'pull_request' || github.event.pull_request.head.repo.id == github.repository_id
+    env:
+      WORKFLOW_FILE: "arc-test-sleepy-matrix.yaml"
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          ref: ${{github.head_ref}}
+
+      - uses: ./.github/actions/setup-arc-e2e
+        id: setup
+        with:
+          app-id: ${{secrets.E2E_TESTS_ACCESS_APP_ID}}
+          app-pk: ${{secrets.E2E_TESTS_ACCESS_PK}}
+          image-name: ${{env.IMAGE_NAME}}
+          image-tag: ${{env.IMAGE_VERSION}}
+          target-org: ${{env.TARGET_ORG}}
+
+      - name: Install gha-runner-scale-set-controller
+        id: install_arc_controller
+        run: |
+          helm install arc \
+          --namespace "arc-systems" \
+          --create-namespace \
+          --set image.repository=${{ env.IMAGE_NAME }} \
+          --set image.tag=${{ env.IMAGE_VERSION }} \
+          --set flags.updateStrategy="eventual" \
+          ./charts/gha-runner-scale-set-controller \
+          --debug
+          count=0
+          while true; do
+            POD_NAME=$(kubectl get pods -n arc-systems -l app.kubernetes.io/name=gha-runner-scale-set-controller -o name)
+            if [ -n "$POD_NAME" ]; then
+              echo "Pod found: $POD_NAME"
+              break
+            fi
+            if [ "$count" -ge 60 ]; then
+              echo "Timeout waiting for controller pod with label app.kubernetes.io/name=gha-runner-scale-set-controller"
+              exit 1
+            fi
+            sleep 1
+            count=$((count+1))
+          done
+          kubectl wait --timeout=30s --for=condition=ready pod -n arc-systems -l app.kubernetes.io/name=gha-runner-scale-set-controller
+          kubectl get pod -n arc-systems
+          kubectl describe deployment arc-gha-runner-scale-set-controller -n arc-systems
+
+      - name: Install gha-runner-scale-set
+        id: install_arc
+        run: |
+          ARC_NAME=${{github.job}}-$(date +'%M%S')$((($RANDOM + 100) % 100 + 1))
+          helm install "$ARC_NAME" \
+          --namespace "arc-runners" \
+          --create-namespace \
+          --set githubConfigUrl="https://github.com/${{ env.TARGET_ORG }}/${{env.TARGET_REPO}}" \
+          --set githubConfigSecret.github_token="${{ steps.setup.outputs.token }}" \
+          ./charts/gha-runner-scale-set \
+          --debug
+          echo "ARC_NAME=$ARC_NAME" >> $GITHUB_OUTPUT
+          count=0
+          while true; do
+            POD_NAME=$(kubectl get pods -n arc-systems -l actions.github.com/scale-set-name=$ARC_NAME -o name)
+            if [ -n "$POD_NAME" ]; then
+              echo "Pod found: $POD_NAME"
+              break
+            fi
+            if [ "$count" -ge 60 ]; then
+              echo "Timeout waiting for listener pod with label actions.github.com/scale-set-name=$ARC_NAME"
+              exit 1
+            fi
+            sleep 1
+            count=$((count+1))
+          done
+          kubectl wait --timeout=30s --for=condition=ready pod -n arc-systems -l actions.github.com/scale-set-name=$ARC_NAME
+          kubectl get pod -n arc-systems
+
+      - name: Trigger long running jobs and wait for runners to pick them up
+        uses: ./.github/actions/execute-assert-arc-e2e
+        timeout-minutes: 10
+        with:
+          auth-token: ${{ steps.setup.outputs.token }}
+          repo-owner: ${{ env.TARGET_ORG }}
+          repo-name: ${{env.TARGET_REPO}}
+          workflow-file: ${{env.WORKFLOW_FILE}}
+          arc-name: ${{steps.install_arc.outputs.ARC_NAME}}
+          arc-namespace: "arc-runners"
+          arc-controller-namespace: "arc-systems"
+          wait-to-running: "true"
+          wait-to-finish: "false"
+
+      - name: Upgrade the gha-runner-scale-set
+        shell: bash
+        run: |
+          helm upgrade --install "${{ steps.install_arc.outputs.ARC_NAME }}" \
+            --namespace "arc-runners" \
+            --create-namespace \
+            --set githubConfigUrl="https://github.com/${{ env.TARGET_ORG }}/${{ env.TARGET_REPO }}" \
+            --set githubConfigSecret.github_token="${{ steps.setup.outputs.token }}" \
+            --set template.spec.containers[0].name="runner" \
+            --set template.spec.containers[0].image="ghcr.io/actions/actions-runner:latest" \
+            --set template.spec.containers[0].command={"/home/runner/run.sh"} \
+            --set template.spec.containers[0].env[0].name="TEST" \
+            --set template.spec.containers[0].env[0].value="E2E TESTS" \
+            ./charts/gha-runner-scale-set \
+            --debug
+
+      - name: Assert that the listener is deleted while jobs are running
+        shell: bash
+        run: |
+          count=0
+          while true; do
+            LISTENER_COUNT="$(kubectl get pods -l actions.github.com/scale-set-name=${{ steps.install_arc.outputs.ARC_NAME }} -n arc-systems --field-selector=status.phase=Running -o=jsonpath='{.items}' | jq 'length')"
+            RUNNERS_COUNT="$(kubectl get pods -l app.kubernetes.io/component=runner -n arc-runners --field-selector=status.phase=Running -o=jsonpath='{.items}' | jq 'length')"
+            RESOURCES="$(kubectl get pods -A)"
+
+            if [ "$LISTENER_COUNT" -eq 0 ]; then
+              echo "Listener has been deleted"
+              echo "$RESOURCES"
+              exit 0
+            fi
+            if [ "$count" -ge 60 ]; then
+              echo "Timeout waiting for listener to be deleted"
+              echo "$RESOURCES"
+              exit 1
+            fi
+
+            echo "Waiting for listener to be deleted"
+            echo "Listener count: $LISTENER_COUNT target: 0 | Runners count: $RUNNERS_COUNT target: 3"
+
+            sleep 1
+            count=$((count+1))
+          done
+
+      - name: Assert that the listener goes back up after the jobs are done
+        shell: bash
+        run: |
+          count=0
+          while true; do
+            LISTENER_COUNT="$(kubectl get pods -l actions.github.com/scale-set-name=${{ steps.install_arc.outputs.ARC_NAME }} -n arc-systems --field-selector=status.phase=Running -o=jsonpath='{.items}' | jq 'length')"
+            RUNNERS_COUNT="$(kubectl get pods -l app.kubernetes.io/component=runner -n arc-runners --field-selector=status.phase=Running -o=jsonpath='{.items}' | jq 'length')"
+            RESOURCES="$(kubectl get pods -A)"
+
+            if [ "$LISTENER_COUNT" -eq 1 ]; then
+              echo "Listener is up!"
+              echo "$RESOURCES"
+              exit 0
+            fi
+            if [ "$count" -ge 120 ]; then
+              echo "Timeout waiting for listener to be recreated"
+              echo "$RESOURCES"
+              exit 1
+            fi
+
+            echo "Waiting for listener to be recreated"
+            echo "Listener count: $LISTENER_COUNT target: 1 | Runners count: $RUNNERS_COUNT target: 0"
+
+            sleep 1
+            count=$((count+1))
+          done
+      
+      - name: Gather logs and cleanup
+        shell: bash
+        if: always()
+        run: |
+          helm uninstall "${{ steps.install_arc.outputs.ARC_NAME }}" --namespace "arc-runners" --debug
+          kubectl wait --timeout=10s --for=delete AutoScalingRunnerSet -n "${{ steps.install_arc.outputs.ARC_NAME }}" -l app.kubernetes.io/instance="${{ steps.install_arc.outputs.ARC_NAME }}"
+          kubectl logs deployment/arc-gha-runner-scale-set-controller -n "arc-systems"