카나리 및 블루/그린 배포

Deploying new features or bug fixes is exciting, but also carries risk. What if the new code has an issue?

• Downtime: Users can't access your service.
• Bugs: New errors impact user experience.
• Rollback Challenges: Reverting to a previous version can be slow or complex.

Advanced deployment strategies help minimize these risks.

A Canary Deployment is a strategy where you release a new version of your application to a small subset of users first.

Think of a canary in a coal mine – it's an early warning system. If the new version (the "canary") fails, only a few users are affected, and you can quickly roll back.

For AWS Lambda, Canary deployments are typically managed using Lambda Aliases and weighted routing.

• An Alias is like a pointer to a specific Lambda function version.
• Weighted routing allows you to direct a percentage of traffic to one version (e.g., the new canary) and the rest to another (the stable old version).

This allows for a gradual and controlled rollout.

Here's a simple Python Lambda function. We'll imagine deploying different versions of this function using Canary. Notice how it returns a version string.

In a real-world scenario, you would define your Lambda function and its aliases using infrastructure-as-code tools like AWS SAM or CloudFormation.

You'd specify a DeploymentPreference for your alias, telling AWS Lambda to shift traffic gradually (e.g., 10% every 5 minutes) while monitoring CloudWatch alarms for errors.

A Blue/Green Deployment involves running two identical, but separate, production environments:

• The Blue environment hosts the current, stable version.
• The Green environment hosts the new version.

Once the Green environment is fully tested, all user traffic is switched from Blue to Green simultaneously. The Blue environment is kept as a fallback.

For serverless applications, Blue/Green deployments are often managed at the API Gateway level.

• You can have two separate API Gateway stages (e.g., prod-blue and prod-green) pointing to different Lambda function versions.
• A DNS record (e.g., using Route 53) can then be updated to switch traffic instantly from the Blue stage to the Green stage.

Both strategies reduce risk, but have differences:

• Canary: Gradual traffic shift, immediate feedback from small user group, complex to manage multiple small shifts.
• Blue/Green: Instant traffic switch, full environment testing, simpler rollback (switch back to Blue), higher resource cost (two full environments).

Choose based on your risk tolerance and operational complexity.

Regardless of the strategy, robust monitoring is crucial. Use Amazon CloudWatch to track:

• Error rates: Any increase in errors?
• Latency: Is the new version slower?
• Application-specific metrics: Are key business metrics impacted?

Automated rollbacks, triggered by CloudWatch alarms, are essential for quickly reverting to a stable state if issues arise.

Which of the following are key benefits of using advanced deployment strategies like Canary or Blue/Green for serverless applications?

You've learned about two powerful advanced deployment strategies for serverless applications:

• Canary Deployments: Gradually shift traffic to a new version, often using Lambda aliases and weighted routing, to test with a small user subset.
• Blue/Green Deployments: Deploy a new version to a separate environment (Green) and then switch all traffic from the old (Blue) environment at once.

Both approaches, combined with strong monitoring, significantly reduce deployment risk and improve application reliability.