Seamless AWS Cloud Development and Testing Locally

anca.ghenade@localstack.cloud @tinyg210 tinyg210 Anca Ghenade Developer Advocate @ LocalStack

How does one go around developing applications using AWS services? A short story… with cats.

1. So you’re tasked with developing a new web application on AWS Cloud

2. Developing on a local machine, you soon find out that there are lots of dependencies with resources in the cloud

3. You realize that the dev & test loop is extremely slow and tedious. Every local change needs to be packaged and uploaded to the cloud for testing.

4. You have a red build on your feature branch, but can’t efficiently test and debug the code in the CI/CD pipeline.

5. The whole team is using Git flow for development - one CI build per feature branch. There is an explosion of different environments required for development (branches x developers).

6. Your manager approaches the team and complains that AWS dev/test resources are not being cleaned up properly causing cost spikes.

Additional pain points: You’re working in environments that have concerns about: ● Security ● Regulatory compliance ● Data Sovereignty ● Risk Management * We’re talking about banking and healthcare especially here.

So the developer experience is… complex. 󰤆

What is LocalStack? A fully functional local cloud stack that was born out of these exact concerns.

● Enables a highly efficient dev & test loop for cloud apps ● Ships as a Docker image, easy to install and start up ● Support for a growing number of services: ○ compute (Lambda, ECS, EKS) ○ various databases (DynamoDB, RDS) ○ messaging (SQS, Kinesis, MSK) ○ some sophisticated/exotic APIs (Athena, Glue) ● CI integrations & advanced collaboration features ○ redefining the way cloud apps are developed across the lifecycle. ● Support for practices like chaos engineering, ephemeral environments for PRs.

“Your application won’t even know the difference.”

Source: https://github.com/tinyg210/san-francisco-ship

No content

Business Logic: ● Shipment entities ● The shipment has weight defined (for ship/plane calculations) ● We need to add the size as a picture ● For size we’re using the IIS (Internet International System) “Banana for scale”

Technical recap ● Bean configuration - using Java @Configuration classes ● Java AWS SDK - helps interaction with resources ● Spring Profiles - environment specific configuration ● IaC via Terraform - consistency across environments

Scenario 1 Scenario 2

No content

You have reasons: ● Most applications are complex in used resources ● Creating resources is time consuming ● Development cycles can be slow - costs developer time ● Resources are tedious to clean up ● If not disposed of properly, resources can be costly ● Interaction between services can be slow on AWS

No content

IaC Tools Developer Tools Language SDKs App Frame- works CI/CD Systems TESTCONTAINERS Copilot

Some benchmarking: AWS LocalStack

What about testing your AWS-powered applications?

YES

No content

How much of your system you can test Cloud emulation Full cloud emulation with service integration 3 Staging environments Highest fidelity because it actually runs in the cloud 4 Mocking Mock out cloud APIs 1 Service emulation Replace individual services with local versions 2 Testability of Cloud App Deployments

No content

Testcontainers

Test successful POST Test successful GET Test fails validation on incorrect values Test 404 on wrong param

No content

AWS Parity ● AWS Server Framework ○ weekly runs to check parity ○ makes heavy use of botocore ● Snapshot testing ● Metrics collection to track test coverage

Thank You!