Amazon Web Services DOP-C02 - AWS Certified DevOps Engineer - Professional

ï‚· Turn off cross-zone load balancing on the ALB ' s target group:

Cross-zone load balancing distributes traffic evenly across all registered targets in all enabled Availability Zones. Turning this off will ensure that each target group only handles requests from its respective Availability Zone.

To disable cross-zone load balancing:

Go to the Amazon EC2 console.

Navigate to Load Balancers and select the ALB.

Choose the Target Groups tab, select the target group, and then select the Group details tab.

Click on Edit and turn off Cross-zone load balancing.

ï‚· Use Amazon Route 53 Application Recovery Controller to start a zonal shift away from the Availability Zone:

Amazon Route 53 Application Recovery Controller provides the ability to control traffic flow to ensure high availability and disaster recovery.

By using Route 53 Application Recovery Controller, you can perform a zonal shift to redirect traffic away from the unhealthy Availability Zone.

To start a zonal shift:

Configure Route 53 Application Recovery Controller by creating a cluster and control panel.

Create routing controls to manage traffic shifts between Availability Zones.

Use the routing control to shift traffic away from the affected Availability Zone.

https://docs.aws.amazon.com/IAM/latest/UserGuide/tutorial_cross-account-with-roles.html

The requirement is to prevent any product files containing unannounced product IDs (prefixed with a specific UUID) from being stored in the production S3 bucket that users can access.

To achieve this, a best practice is to use a staging bucket as a control point before files go to production, combined with Amazon Macie’s data classification capabilities.

Creating a custom data identifier in Amazon Macie allows precise detection of product IDs starting with the specific UUID, which default managed identifiers will not detect.

By running a Macie sensitive data discovery job on the staging bucket, you can identify files containing these sensitive product IDs.

Only files without findings (i.e., files that do not contain unannounced product IDs) are copied to the production bucket, ensuring no sensitive information is exposed. This approach aligns with AWS best practices for data classification and staged deployment workflows, maximizing control and reducing risk. Using Macie on the production bucket directly (options B and D) risks exposing sensitive data before detection and deletion. Option C uses managed data identifiers, which will likely not detect the custom UUID prefix pattern.

Reference from AWS Official Documentation and Study Guide:

Amazon Macie Custom Data Identifiers: " You can create custom data identifiers in Amazon Macie to find sensitive data that is unique to your organization. " (Amazon Macie User Guide)

Data Security Best Practices: " Use staging environments to inspect and sanitize data before moving it to production to reduce exposure risks. " (AWS Security Best Practices)

Option A is incorrect because creating an Amazon EventBridge rule that invokes an AWS Lambda function to query the applications on a 5-minute interval is not a valid solution. EventBridge rules can only trigger Lambda functions based on events, not on time intervals. Moreover, querying the applications on a 5-minute interval might incur unnecessary costs and network overhead, and might not detect performance issues in real time.

Option B is correct because creating an Amazon CloudWatch Synthetics canary that runs a custom script to query the applications on a 5-minute interval is a valid solution. CloudWatch Synthetics canaries are configurable scripts that monitor endpoints and APIs by simulating customer behavior. Canaries can run as often as once per minute, and can measure the latency and availability of the applications. Canaries can also send notifications to an Amazon SNS topic when they detect errors or performance issues1.

Option C is incorrect because creating an Amazon CloudWatch alarm that uses the AWS/ApplicationELB namespace RequestCountPerTarget metric is not a valid solution. The RequestCountPerTarget metric measures the number of requests completed or connections made per target in a target group2. This metric does not reflect the application response time, which is the requirement. Moreover, configuring the CloudWatch alarm to send a notification when the number of connections becomes greater than the configured number of threads that the application supports is not a valid way to measure the application performance, as it depends on the application design and implementation.

Option D is incorrect because creating an Amazon CloudWatch alarm that uses the AWS/ApplicationELB namespace RequestCountPerTarget metric is not a valid solution, for the same reason as option C. The RequestCountPerTarget metric does not reflect the application response time, which is the requirement. Moreover, configuring the CloudWatch alarm to send a notification when the average response time becomes greater than the longest response time that the application supports is not a valid way to measure the application performance, as it does not account for variability or outliers in the response time distribution.

This is a classic AWS Organizations governance design:

Restrict Regions and allowed services across all accounts

The correct Organizations mechanism for account-wide guardrails is a Service Control Policy (SCP) .

SCPs set the maximum permissions that accounts/OUs can use, making them the right tool to enforce “only these Regions” and “only these services” consistently across the org.

Authentication from Active Directory + consistent job-function roles in every account

With AD as the identity source, the standard approach is federation to AWS using an IAM identity provider (SAML) and role assumption based on job function.

AWS CloudFormation StackSets is the operationally efficient way to deploy identical IAM roles/policies into multiple accounts/OUs so job-function permissions are consistent everywhere.

The roles’ trust policies can be configured to allow federated identities (from the AD-backed IdP) to assume the correct job role in each account.

Why the other options don’t fit:

A “OU with group policies” isn’t an AWS Organizations control mechanism for restricting Regions/services. The AWS-native control for this is SCPs , not “group policies.”

B Permissions boundaries are useful to limit what an IAM principal/role can do within an account , but they are not the org-wide enforcement mechanism for “all accounts must stay in these Regions / use only these services.” That’s what SCPs do.

C AWS RAM is for sharing resources (like subnets, Transit Gateway, etc.), not for “sharing roles” across accounts as a governance baseline. Also, the question emphasizes identical permissions in each account, which is best achieved by provisioning roles in each account (StackSets), not attempting to “share” roles.

AWS Systems Manager Patch Manager automates patch scanning and installation. Integrating with AWS Config enables compliance reporting and auto-remediation. This provides centralized patch compliance management per AWS Ops best practices.

The best solution to implement failover for the application is to use CloudFront origin groups. Origin groups allow CloudFront to automatically switch to a secondary origin when the primary origin is unavailable or returns specific HTTP status codes that indicate a failure1. This way, CloudFront can serve the requests from the secondary ALB in the secondary Region without any delay or redirection. To set up origin groups, the DevOps engineer needs to create a new origin on the distribution for the secondary ALB, create a new origin group with the original ALB as the primary origin and the secondary ALB as the secondary origin, and configure the origin group to fail over for HTTP 5xx status codes. Then, the DevOps engineer needs to update the default behavior to use the origin group instead of the single origin2.

The other options are not as effective or efficient as the solution in option B. Option A is not suitable because creating a second CloudFront distribution will increase the complexity and cost of the application. Moreover, using Route 53 alias records with a failover policy will introduce some delay in detecting and switching to the secondary CloudFront distribution, which may not meet the zero-second RTO requirement. Option C is not feasible because CloudFront does not support using Route 53 alias records as origins3. Option D is not advisable because using a CloudFront function to redirect the requests to the secondary ALB will add an extra round-trip and latency to the failover process, which may also not meet the zero-second RTO requirement.

1: Optimizing high availability with CloudFront origin failover - Amazon CloudFront

2: Creating an origin group - Amazon CloudFront

3: Values That You Specify When You Create or Update a Web Distribution - Amazon CloudFront

The correct answer is D. The scenario requires a solution that can migrate an application from on premises to AWS, run on specific versions of Apache Tomcat, HAProxy, and Varnish Cache, tune the operating system-level parameters, automate the deployment of new application versions, and scale and replace faulty servers automatically. Option D meets all these requirements by using AWS CodeCommit, AWS CodeDeploy, AWS CodePipeline, and Amazon EC2 Auto Scaling. AWS CodeCommit is a fully managed source control service that hosts Git repositories and works with Git-based tools. AWS CodeDeploy is a fully managed deployment service that automates software deployments to a variety of compute services, including Amazon EC2, AWS Fargate, AWS Lambda, and on-premises servers. AWS CodePipeline is a fully managed continuous delivery service that helps automate the release pipelines for fast and reliable application updates. Amazon EC2 Auto Scaling helps maintain application availability and allows scaling of Amazon EC2 capacity up or down automatically according to the defined conditions. By using these services together, the DevOps engineer can migrate the application code to AWS, configure and install the necessary software using the appspec.yml file, automate the deployment process using the pipeline, and scale and replace the servers using the Auto Scaling group.

Option A is incorrect because AWS Fargate is a serverless compute engine for containers that works with Amazon ECS and Amazon EKS. Fargate removes the need to provision and manage servers, but it also limits the ability to tune the operating system-level parameters, which is a requirement in the scenario. Moreover, Fargate does not support HAProxy and Varnish Cache as sidecar containers, which are needed to run the application properly.

Option B is incorrect because AWS Elastic Beanstalk is a fully managed service that automates the deployment and scaling of web applications and services using familiar servers such as Apache, Nginx, Passenger, and IIS. However, Elastic Beanstalk does not support HAProxy and Varnish Cache as part of the Tomcat solution stack, which are needed to run the application properly. Moreover, Elastic Beanstalk web server tier environments are designed to serve HTTP requests, not to process background tasks, which is the purpose of worker tier environments.

Option C is incorrect because AWS Elastic Beanstalk worker tier environments are designed to process background tasks using a daemon process that runs on each Amazon EC2 instance in the environment. Worker tier environments are not suitable for running web applications that serve HTTP requests, which is the case in the scenario. Moreover, Elastic Beanstalk does not support HAProxy and Varnish Cache as part of the Tomcat solution stack, which are needed to run the application properly.

AWS CodeCommit

AWS CodeDeploy

AWS CodePipeline

Amazon EC2 Auto Scaling

AWS Fargate

AWS Elastic Beanstalk

Amazon ECR enhanced scanning uses Amazon Inspector for vulnerability detection.

EventBridge can capture Inspector scan findings.

Lambda can process scan findings and reject manual approval if critical vulnerabilities exist.

Options A and C use incorrect or less integrated services (basic scanning or Detective).

Option B adds unnecessary complexity with SBOM and Athena.