Weather

Simple webapp displaying a grid of weather forecasts. A deployment of this using Bay Area locales is running at https://alecholmes.com/weather.

The frontend is implemented completely in index.html without using any third party dependencies.

Weather data is stored in a public JSON blob which is periodically updated by the backend.

The backend is a Golang CLI app that will run as an AWS lambda if no arguments are provided. See main.tf for all the necessary boilerplate to deploy the backend.

Building and deploying the lambda

Prerequisites:

• OpenWeather API key
• AWS account
• Golang toolchain
• Terraform

make build-aws-lambda

cd terraform
terraform apply

Terraform will need to run in the context of an AWS user with the following permission policies (these can be scoped down):

• IAMFullAccess
• AmazonS3FullAccess
• CloudWatchEventsFullAccess
• AWSLambda_FullAccess

To run Terraform locally, AWS credentials can be set in environment variables:

export AWS_ACCESS_KEY_ID=your_access_key
export AWS_SECRET_ACCESS_KEY=your_secret_key
export AWS_REGION=your_region # e.g. us-east-1

Creating weather snapshot JSON locally

# If -snapshot-to-s3 is omitted, the snapshot is written to stdout instead.
go run ./cmd/syncweather -local -config config/config.json -snapshot-to-s3

Notes

The OpenWeather API free tier only allows 1000 API calls per day. Since there is an API call for each location, the number of locations and lambda frequency must be tweaked to remain within this limit. If data is updated hourly, there can be no more than 41 locations.