Requisites

• node (install using nvm)
• aws cli

AWS CloudFormation intro

AWS CDK Code -> synths into -> AWS CF Template -> which deploys to -> AWS CF

AWS CDK

• ! tip: Don't install aws-cdk globally.
  • Instead, for every project install the package locally and set its version in package.json file
  • If you need to use the aws-cdk package commands, such as the init command, you can use npx for that. Like so: npx cdk <command>
  • Afterwards, everytime you want to use cdk use npm run cdk to use the local installation of the package
  • Reference
  • TODO: Agregar un alias a $ZSH_CUSTOM/aliases.zsh para cdk_new='npx cdk init --language=typescript' Inside the init folder, important files are:
• bin/cdk-starter.ts: entry point for the project, as set by the 'app' property on the cdk.json file
• cdk.json: Configuration for the CDK
  • app property sets the entry point of the application
• lib: Stack definitions are stored here

Useful References

• AWS CDK Reference Documentation
• CDK Constructs Library
• AWS CDK GitHub Page
  • Better readability in Constructs Hub
  • Issues GitHub page

CDK CLI Commands

• cdk init app --language=typescript
• cdk bootstrap: prepares the aws environment for usage with cdk. before a cdk app is deployed, the env must first be bootstrapped,
  • it provisions a s3 bucket (used to store cdk project files), a ecr repository (used to store docker images) and iam roles (configured to grant permissions needed by cdk to perform deployments)
• cdk deploy: synthetizes and deploys the app to aws
  • creates/updates a cdk.out folder, which contains the cloudformation template that is synthetized by cdk and deployed to aws
  • to deploy a specific stack, use cdk deploy <stack-id>
• cdk synth: synthetizes the stack(s); generates a cloudformation template to cdk.out folder
  • generates a cf template for each stack specified in the bin file
• cdk list: lists all stacks defined in the app locally
• cdk diff: lists all differences between the local resources and remote/deployed ones
• cdk doctor: checks if there are any problems in the app config or libraries
• cdk destroy <stack-id>: deletes the specified stack in aws

Constructs

• basic building blocks of a cdk app
• levels/types of constructs:
  • L1 - low level constructs: represents raw aws resources (no default options -- must configure all properties)
  • L2 - higher level constructs: represents aws resources with a higher level of abstraction (default options)
  • L3 - patterns: combines multiple resources, to represent a commonly used pattern of resources used in aws

Outputs

• we can define outputs in cdk just like we can in cf.
• to do it, we must use a L1 construct called CfnOutput

Deployment Parameters

• we can define parameters in cdk just like we can in cf
• to do so, we can use a L1 construct called CfnParameter
• then, to pass the parameters to a cdk command, one can use:
  • cdk <command> --parameters <parameter-name>=<parameter-value>
• TODO: Check external parameters

CDK IDs

• any resource declared in a cdk app requires an ID. this is the second parameter given to a construct's constructor method.
• this cdk id is used in the logical id (and physical id, if none is provided in the resource's options) of the resource in cloudformation.
  • the logical id in cf consists of the cdk id + a random string (this is generated by cdk)
    • to override this behavior we can use the following:

const bucket = new Bucket(this, 'PhotoBucket')

(bucket.node.defaultChild as CfnBucket).overrideLogicalId('PhotoBucket_OverridenLogicalId')

• the physical id of a resource can be generated by cdk or can be established in the resource's construct options
• be careful of manually naming resources' logical or physical id, since it could result in some weird behavior such as resources being unable to be replaced.

Cloudformation Intrinsic Functions

• build-in functions to help manage our stacks
• complete reference of cf intrinsic functions here
• these functions can be used in the app's code, importing Fn from aws-cdk-lib
  • Example: Fn.split(), Fn.select()

Multiple stacks

• why?
  • bc of stacks with sensitive info
  • some stacks may take a lot of deploy/deletion time
  • organization of resources
• to deploy multiple stacks, we must use cdk deploy --all

Cross stack references

• Can be done using: new CfnOutput in one stack and referencing using Fn.importValue() in the other

Sharing resources with CDK

• instead of using CfnOutputs and Fn.importValue(), we can make the stack have public readonly properties and pass them to another stack class by extending its constructor props object.
  • this way, cdk is actually handling the outputs + imports behind the scenes for us

CDK Aspects

• feature of cdk that lets us check or modify resources after they were created
  • implements the visitor pattern
  • simple use case: add tag to a resource
  • popular usecase: enforce security or best practices (like a code linter)
• it's useful to apply an operation to all constructs in a given scope (reference)
• cdk-nag: popular library to enforce rules via cdk aspects

Deleting stacks

• if a stack has resources that are referenced in other stacks, this stack cannot be deleted
• to delete them, the stack the reference has to be first removed or the whole stack should be deleted
• ! note: this is only recommended in test/learning environments. never do this in prod.

AWS Lambda CDK Constructs

• CDK offers a construct that provides many utilities to work with AWS Lambda.
• This construct is named NodejsFunction, and provides:
  • Code bundling with tree shaking
  • Compiles TS to JS
  • Leaves out AWS-SDK dependencies
  • Completely editable
  • Library used for bunding is esbuild