如何实现CodeDeploy失败时终止Bitbucket流水线？

Absolutely! You can absolutely implement this by polling the AWS CodeDeploy API in a loop within your Bitbucket Pipeline. Here's a step-by-step breakdown of how to do it, including the specific API you'll need and example configurations for your files.

Polling the AWS CodeDeploy API is a reliable way to wait for your deployment to fully complete (including the ValidateService step) and fail the pipeline if the deployment doesn't succeed. The key API you'll use is get-deployment — this returns detailed status information for a specific CodeDeploy deployment.

First, when you trigger your CodeDeploy deployment from the pipeline, make sure to capture the deploymentId returned by the create-deployment command. You'll need this ID to poll the status later.

Example snippet for your deployment step in bitbucket-pipeline.yml:

- step:
    name: Deploy to CodeDeploy
    script:
      # Assume AWS credentials are configured via secured pipeline variables
      - DEPLOYMENT_ID=$(aws deploy create-deployment --application-name YOUR_APP_NAME --deployment-group-name YOUR_DEPLOYMENT_GROUP --github-location repository=YOUR_REPO,commitId=$BITBUCKET_COMMIT --output text --query deploymentId)
      - echo "Initiated deployment with ID: $DEPLOYMENT_ID"
      # Save the ID to a file to pass to the next step (Bitbucket allows sharing via artifacts)
      - echo $DEPLOYMENT_ID > deployment_id.txt
    artifacts:
      - deployment_id.txt

Next, add a new step that reads the deployment ID, then repeatedly calls aws deploy get-deployment to check the status until it reaches a terminal state (SUCCEEDED, FAILED, or STOPPED). If it fails, we'll exit with a non-zero code to terminate the pipeline.

This step uses jq (a JSON parsing tool) to extract the status from the API response — install it if your pipeline image doesn't include it by default.

Example polling step:

- step:
    name: Monitor CodeDeploy Deployment
    script:
      # Install jq (adjust command if using a non-Ubuntu image, e.g., yum install jq for Amazon Linux)
      - apt-get update && apt-get install -y jq
      # Retrieve the deployment ID from the artifact
      - DEPLOYMENT_ID=$(cat deployment_id.txt)
      # Configure polling parameters (adjust based on your deployment's typical duration)
      - POLL_INTERVAL=30
      - MAX_ATTEMPTS=20
      - ATTEMPT=0

      echo "Starting monitoring for deployment $DEPLOYMENT_ID..."

      while [ $ATTEMPT -lt $MAX_ATTEMPTS ]; do
        # Fetch deployment status
        DEPLOYMENT_STATUS=$(aws deploy get-deployment --deployment-id $DEPLOYMENT_ID --output json | jq -r '.deploymentInfo.status')
        
        echo "Attempt $((ATTEMPT+1)): Current status → $DEPLOYMENT_STATUS"

        case $DEPLOYMENT_STATUS in
          "SUCCEEDED")
            echo "Deployment completed successfully!"
            exit 0
            ;;
          "FAILED" | "STOPPED")
            echo "Deployment failed with status: $DEPLOYMENT_STATUS"
            # Optional: Fetch failure details for debugging
            aws deploy get-deployment --deployment-id $DEPLOYMENT_ID --output json | jq '.deploymentInfo.errorInformation'
            exit 1
            ;;
          *)
            echo "Deployment still in progress. Waiting $POLL_INTERVAL seconds..."
            sleep $POLL_INTERVAL
            ATTEMPT=$((ATTEMPT+1))
            ;;
        esac
      done

      # Fail pipeline if deployment times out
      echo "Deployment timed out after $MAX_ATTEMPTS attempts."
      exit 1

• The API returns a JSON object where deploymentInfo.status can be:
  • PENDING: Deployment is queued to start
  • IN_PROGRESS: Deployment is running (including your ValidateService hook)
  • SUCCEEDED: All steps (including validation) completed successfully
  • FAILED: One or more deployment steps failed
  • STOPPED: Deployment was manually halted
• Ensure your AWS IAM user (used by the Bitbucket Pipeline) has the codedeploy:GetDeployment permission added to its policy.

Since you mentioned Bitbucket wasn't waiting for the ValidateService step to finish, this polling approach solves that problem — it waits for the entire deployment lifecycle to conclude. The SUCCEEDED status is only returned after all hooks defined in your appspec.yml (including ValidateService) have passed.

• Tune POLL_INTERVAL and MAX_ATTEMPTS to match your deployment's average runtime to avoid unnecessary waits or infinite loops.
• Store AWS credentials as secured variables in Bitbucket to prevent exposure in logs.
• If using a custom pipeline image, pre-install jq to speed up the step.

内容的提问来源于stack exchange，提问作者KarmicJohn