Step 1: Reacting to RDS Storage Autoscaling Events Using Amazon EventBridgeAmazon RDS emits events when storage autoscaling occurs. To visualize these events in a CloudWatch dashboard, you can create an EventBridge rule that listens for these specific autoscaling events.
Action: Create an EventBridge rule that reacts to RDS storage autoscaling events from the RDS event stream.
Why: EventBridge allows you to listen to RDS events and route them to specific AWS services for processing.
Step 2: Creating a Custom CloudWatch Metric via LambdaOnce the EventBridge rule detects a storage autoscaling event, you can use a Lambda function to publish a custom metric to CloudWatch. This metric can then be visualized in a CloudWatch dashboard.
Action: Use a Lambda function to publish custom metrics to CloudWatch based on the RDS storage autoscaling events.
Why: Custom metrics allow you to track specific events like autoscaling and visualize them easily on a CloudWatch dashboard.
Reference: AWS documentation on Publishing Custom Metrics to CloudWatch.
This corresponds to Option A: Create an Amazon EventBridge rule that reacts to RDS storage autoscaling events from RDS events. Create an AWS Lambda function that publishes a CloudWatch custom metric.
Configure the EventBridge rule to invoke the Lambda function. Visualize the custom metric by using the CloudWatch dashboard.

Option A is incorrect because setting the deployment configuration to LambdaAllAtOnce means that the new version of the application will be deployed to all Lambda functions at once, affecting all customers. This does not meet the requirement of affecting the fewest customers possible. Moreover, configuring automatic rollbacks on the deployment group is not operationally efficient, as it requires manual intervention to fix the errors and redeploy the application.
Option B is correct because setting the deployment configuration to LambdaCanary10Percent10Minutes means that the new version of the application will be deployed to 10 percent of the Lambda functions first, and then to the remaining 90 percent after 10 minutes. This minimizes the impact of errors on customers, as only 10 percent of them will be affected by a faulty deployment. Configuring automatic rollbacks on the deployment group also meets the requirement of reverting to the most recent stable version of the application when an error is detected. Creating a CloudWatch alarm that detects HTTP Bad Gateway errors on API Gateway is a valid way to monitor the health of the application and trigger a rollback if needed.
Option C is incorrect because setting the deployment configuration to LambdaAllAtOnce means that the new version of the application will be deployed to all Lambda functions at once, affecting all customers. This does not meet the requirement of affecting the fewest customers possible. Moreover, configuring manual rollbacks on the deployment group is not operationally efficient, as it requires human intervention to stop the current deployment and start a new one. Creating an SNS topic to send notifications every time a deployment fails is not sufficient to detect errors in the application, as it does not monitor the API Gateway responses.
Option D is incorrect because configuring manual rollbacks on the deployment group is not operationally efficient, as it requires human intervention to stop the current deployment and start a new one. Creating a metric filter on a CloudWatch log group for API Gateway to monitor HTTP Bad Gateway errors is a valid way to monitor the health of the application, but invoking a new Lambda function to perform a rollback is unnecessary and complex, as CodeDeploy already provides automatic rollback functionality.
References:
AWS CodeDeploy Deployment Configurations
[AWS CodeDeploy Rollbacks]
Amazon CloudWatch Alarms

* Havingseparate CodeArtifact repositoriesin dev and prod accounts provides clear isolation and control.
* Repository policies can restrict dev repo access to developers.
* EventBridge triggers pipelines to test and promote packages only if tests pass, ensuring safe deployment to production.
* Using S3 (C and D) is not ideal for package management.
* A single repo (A) complicates access and version control across accounts.
References:
CodeArtifact Repository Policies
Cross-Account Package Promotion

To meet the requirements of failover and disaster recovery, the company should use the following deployment strategies:
Create an Amazon Aurora global database in two AWS Regions as the data store. In the event of a failure, promote the secondary Region to the primary for the application. Update the application to use the Aurora cluster endpoint in the secondary Region. This strategy can provide a low RPO and RTO for the data, as Aurora global database replicates data with minimal latency across Regions and allows fast and easy failover12. The company can use the Amazon Aurora cluster endpoint to connect to the current primary DB cluster without needing to change any application code1.
Set up the application in two AWS Regions. Configure AWS Global Accelerator to point to Application Load Balancers (ALBs) in both Regions. Add both ALBs to a single endpoint group. Use health checks and Auto Scaling groups in each Region. This strategy can provide high availability and performance for the application, as AWS Global Accelerator uses the AWS global network to route traffic to the closest healthy endpoint3. The company can also use static IP addresses that are assigned by Global Accelerator as a fixed entry point for their application1. By using health checks and Auto Scaling groups, the company can ensure that their application can scale up or down based on demand and handle any instance failures4.
The other options are incorrect because:
Creating an Amazon Aurora Single-AZ cluster in multiple AWS Regions as the data store would not provide a fast failover or disaster recovery solution, as the company would need to manually restore data from backups or snapshots in another Region in case of a failure.
Creating an Amazon Aurora cluster in multiple AWS Regions as the data store and using a Network Load Balancer to balance the database traffic in different Regions would not work, as Network Load Balancers do not support cross-Region routing. Moreover, this strategy would not provide a consistent view of the data across Regions, as Aurora clusters do not replicate data automatically between Regions unless they are part of a global database.
Setting up the application in two AWS Regions and using Amazon Route 53 failover routing that points to Application Load Balancers in both Regions would not provide a low RTO, as Route 53 failover routing relies on DNS resolution, which can take time to propagate changes across different DNS servers and clients.
Moreover, this strategy would not provide deterministic routing, as Route 53 failover routing depends on DNS caching behavior, which can vary depending on different factors.

https://aws.amazon.com/caching/session-management/