TheAlanWang

Notes

AWS CDK Deployment

Layer 3 — AWS Resources (Assignment 3)
Layer 1 — Local Workflow
Layer 2 — AWS Deployment Layer (CloudFormation)

Assignment Goal

Reference: Assignment Requirements

Keep the Assignment 2 behavior, but deploy the infrastructure with AWS CDK instead of manually creating resources in the AWS Console.

Business flow stays the same:

  • S3 object changes are tracked
  • total size history is stored in DynamoDB
  • plotting Lambda generates a chart
  • driver Lambda runs the demo sequence

CDK Project Setup

Typical bootstrap commands:

cdk init app --language python
python3 -m venv .venv
source .venv/bin/activate
./.venv/bin/python -m pip install -r requirements.txt

Local synth-style check:

./.venv/bin/python app.py

Project Layout

assignment3/
├── app.py
├── assignment3_app/
│   ├── storage_stack.py
│   └── lambda_stack.py
├── lambdas/
│   ├── size_tracking_lambda.py
│   ├── plotting_lambda.py
│   └── driver_lambda.py
└── tests/

StorageStack

Creates the stateful resources first.

  • data S3 bucket
  • plot S3 bucket
  • DynamoDB table
  • GSI: GSI_SizeByBucket

Purpose:

  • persist bucket size history
  • store generated plot image

LambdaStack

Consumes the storage resources.

Creates:

  • size-tracking Lambda (Python 3.12)
  • plotting Lambda (Python 3.9 + matplotlib layer)
  • driver Lambda (Python 3.12)
  • API Gateway REST API
  • EventBridge rule
  • IAM permissions
  • CloudWatch Log Groups

plotting_lambda.py

Reads history and generates the plot.

Does:

  • get latest DynamoDB record
  • query items within 10 seconds before that latest record
  • query historical maximum from GSI
  • generate matplotlib chart
  • upload image to plot bucket

Only triggered via API Gateway (no automatic trigger)

S3: DataBucket

DynamoDB Data Model

Main table:

  • partition key: bucket_name
  • sort key: time

GSI:

  • partition key: bucket_name
  • sort key: total_size

Why:

  • query recent history from main table
  • query historical max without using scan

S3: PlotBucket

driver_lambda.py

Runs the demo sequence:
0. clean up leftover objects

  1. create assignment1.txt (18 B)
  2. update it (27 B)
  3. delete it (0 B)
  4. create assignment2.txt (2 B)
  5. call plotting API

Expected visible history:
0 -> 18 -> 27 -> 0 -> 2

size_tracking_lambda.py

Triggered by object changes.

Does:

  • list objects in tracked bucket
  • compute object_cnt
  • compute total_size
  • write one history item to DynamoDB

Input:

  • S3 / EventBridge event
    Output:
  • DynamoDB history record

End-to-End Flow

Automated tracking:
S3 object change
→ EventBridge (Object Created / Deleted)
→ size_tracking_lambda
→ DynamoDB history record

Plot generation (manual only):
driver_lambda
→ API Gateway
→ plotting_lambda
→ plot bucket

plotting_lambda has NO automatic trigger.
It is only invoked via API Gateway.

Deploy Workflow

Typical order:

  1. verify AWS account / region
  2. cdk bootstrap if needed
  3. cdk synth
  4. cdk deploy
  5. invoke driver Lambda
  6. validate DynamoDB + plot output
  7. cdk destroy when done

CDK Toolkit (CDKToolkit)

Created by cdk bootstrap.

This is the CDK bootstrap stack, not your business stack.

Typically provides:

  • bootstrap S3 bucket for CDK assets
  • IAM roles used during deployment
  • support resources for synth / deploy flow

Why it appears:

  • CDK needs a place to stage assets before CloudFormation deploys them

CloudFormation

CDK does not create resources directly.

Actual path:
Python CDK code
→ cdk synth
→ CloudFormation template
→ CloudFormation creates / updates stacks

Responsible for:

  • stack lifecycle
  • updates / rollback
  • resource state tracking

Graph View