feat: enhances deploy/validation scripts, adds manual removal script [#73]

[robert-7]

Summary

This PR turns the manual-removal work into a real cutover toolkit for moving the AWS deployment from the old hand-built environment to the new CDK-owned stack.

The final scope ended up being broader than just adding a teardown script. During the live cutover we found and fixed several real issues in the rebuild path: hidden console-created CloudFormation stacks that still owned ECS resource names, CDK bootstrap missing in the target account, reuse of the existing GitHub OIDC provider, slow/recursive Docker asset packaging, and ECS service creation racing ahead of ALB listener attachment.

What changed

• Added a dedicated infra/manual-removal/ toolkit:
  • remove_manual_stack.sh
  • validate_stack_is_up.sh
  • validate_manual_stack_teardown_completed.sh
  • validate_cdk_predeploy.sh
  • common_manual_stack_vars.sh
  • common_manual_stack_helpers.sh
  • manual-stack-teardown.md
• Updated the teardown flow to also remove leftover console-created CloudFormation stacks for the ECS cluster/service, not just the raw AWS resources.
• Improved the predeploy validator so it now checks:
  • AWS account alignment with infra/.env
  • Docker availability
  • teardown completion
  • ACM certificate status
  • CDK bootstrap presence
  • GitHub OIDC provider reuse configuration
  • cdk synth
• Updated the CDK stack so it can reuse an existing GitHub OIDC provider via CDK_GITHUB_OIDC_PROVIDER_ARN.
• Updated the CDK bootstrap image path so the first rebuild deploy is practical from a local Docker build.
• Reduced the bootstrap Docker asset context so CDK does not drag large local artifacts like infra/.venv, node_modules, and report output into the asset build.
• Fixed stack ordering so the ECS service waits for the ALB/listener/target group attachment before creation.
• Updated the AWS architecture and teardown docs to match the real cutover process.

Live cutover work completed

The rebuilt stack was actually brought up during this work:

• removed the old manual stack resources
• removed leftover console-created CloudFormation stacks that still owned ECS names
• bootstrapped the AWS account/region for CDK
• deployed CourseEnrollmentAppStack successfully
• validated the rebuilt stack in AWS

Validation performed

• bash infra/manual-removal/remove_manual_stack.sh --yes
• bash infra/manual-removal/validate_manual_stack_teardown_completed.sh
• bash infra/manual-removal/validate_cdk_predeploy.sh
• cd infra && cdk bootstrap aws://334348504030/us-east-1 --require-approval never
• cd infra && cdk deploy CourseEnrollmentAppStack --output /tmp/course-enrollment-app-cdk-out-deploy --require-approval never
• bash infra/manual-removal/validate_stack_is_up.sh
• curl -I http://course-enrollment-app-alb-1547360316.us-east-1.elb.amazonaws.com/index
• curl -k -I https://course-enrollment-app-alb-1547360316.us-east-1.elb.amazonaws.com/index

Follow-up

• Namecheap still needs to point course-enrollment-app.robertlech.com at the new ALB DNS name course-enrollment-app-alb-1547360316.us-east-1.elb.amazonaws.com.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!