Procedure from AWS CDK (Python) development to AWS resource construction * For beginners of development

Because I came with almost one infrastructure and had little development experience Here's a summary of what development beginners did and how to build AWS resources using CDK.

• Please note that this may not work depending on the version. I am using the latest version as of December 2019

Coding environment

Refer to the following article for environment preparation Procedure for building a CDK environment on Windows (Python) https://qiita.com/toma_shohei/items/985916e1a95ec4c38121

First get an overview

■ Outline understanding material https://pages.awscloud.com/rs/112-TZM-766/images/B-3.pdf

◾️ Try WorkShop ・ Workshop (English) https://cdkworkshop.com/30-python.html

・ Workshop (Japanese: download version) http://bit.ly/cdkworkshopjp

CDK (Python) coding preparation

-Open Visual Studio Code and open the directory to specify the working directory.

・ Placement of cdk initialization file Execute the following command from the terminal cdk init app --language=python

• This command creates the necessary directories, but the sources are also initialized.

-Edit requirements.txt (install resources to handle) ＝＝＝＝＝＝＝ aws-cdk.core

aws-cdk.aws-autoscaling aws-cdk.aws-ec2 aws-cdk.aws-elasticloadbalancingv2 aws-cdk.aws-rds aws-cdk.aws-ssm aws-cdk.aws-route53

aws-cdk.aws-autoscaling-common =====＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝ pip install -r requirements.txt

If you can do this, start coding.

Folder structure

At the stage of cdk init, the following folder structure will be created without permission. https://cdkworkshop.com/30-python/20-create-project/300-structure.html

I have introduced pyyaml to separate the parameters from the code. The folder structure will look like the image below.

・ Folder structure image

CDK-WORKSHOP □┣config □┃┣prd □┃┃┣ec2.yaml □┃┃┣vpc.yaml □ ┃┃┣ ・ ・ ・ □┃┗stg □┃□┣ec2.yaml □┃□┣vpc.yaml □ ┃ □ ┣ ・ ・ ・ □┣cdk-workshop_stack.py □ ┣ ・ ・ ・

• A folder called config is cut directly under the TOP of the CDK. Under that, cut folders by environment (prd, stg, etc.) and Parameters are described for each resource (ec2, vpc, etc.) under that. (Switch the directory to read with cdk.json) Write the code to build all resources in cdk-workshop_stack.py (read the parameters under config). It was not possible to build in multiple regions in the same folder, so separate the folders. The following method cannot be used when using pyyaml (it seems to be a bug, but the output is always JSON) 　https://dev.classmethod.jp/cloud/aws-cdk-multi-account-region/

Provisional support (planned to be fixed in the future, solution unknown at present)

When writing! Ref in the parameter, use "\! Ref" Batch conversion "\! Ref" ⇒ "! Ref" of the output yaml. *! GetAtt is the same

• I'm using yaml.safe_load in pyyaml, but I don't know how to escape the "!" If you use "! Ref" + characters read from yaml on .py without writing in yaml, "" will be applied to the entire character string, and even! Ref will be treated as a character string.

CloudFormation execution

・ CloudFormation template output cdk synth --version-reporting false --path-metadata false > C:\Users\test\Desktop\test.txt

・ Implementation of the above provisional measures

・ CloudFormation deploy command (until changeset creation) aws cloudformation deploy --stack-name testStack --template-file C:\Users\test\Desktop\test.txt --no-fail-on-empty-changeset --no-execute-changeset

・ Change set confirmation ⇒ execution When you open CloudFormation from the management console Check the change set of the created stack and press Execute if there is no problem

Frequently viewed documents

・ AWS CDK Python Reference https://docs.aws.amazon.com/cdk/api/latest/python/index.html

・ Python examples https://github.com/aws-samples/aws-cdk-examples

Task

・ Use of High-level constructs For operation (?) In which a resource to be newly constructed is constructed with code on an existing resource that has already been constructed manually. Only Low-level constructs (library named CfnXXX) are available

• Originally, I would like to use High-level constructs to reduce the amount of code, Is it only possible to create on resources created with High-level constructs? (⇒ This is under investigation)

・ No test code is applied It is unexamined what to check in the first place

Impressions

・ About construction with IaC The quality will definitely go up. (I always make a mistake if it is a construction project of a certain scale I did not make a mistake) Once the verification environment is created, the production environment can be created almost by copying, so the reproducibility is quite high.

・ Characteristics of CDK I have experience building AWS with Terraform, but I have the impression that CDK fits in and has fewer points. Since it is a latecomer tool, the interval between version upgrades is short (committed almost every day), There is a risk because the latest version may not be backward compatible, but at the moment there are only a few things that have plagued me.