aws-spa

Deploy a single page app on AWS in one command.

Install & use

npm install --dev aws-spa

npx aws-spa deploy --help

Why?

Configuring the deployment of a single page app is harder than it should be. Most SPA configuration are very similar. aws-spa embodies this idea. It is meant to handle all the quirks associated with SPA configuration.

Features

• Create AWS Bucket & CloudFront distribution & Route 53 record & ACM certificate and configure it
• Serve gzipped file
• Invalidate CloudFront after deployment
• Basic Auth (recommended to avoid search engine indexation)
• idempotent script

Get Started

With create-react-app

npx create-react-app hello-world && cd hello-world
yarn add aws-spa
yarn build

# read about [create-react-app static file caching](https://facebook.github.io/create-react-app/docs/production-build#static-file-cachin)
npx aws-spa deploy hello.example.com --cacheInvalidation "index.html" --cacheBustedPrefix "static/"

API

aws-spa deploy

Deploy a single page app on AWS

Positionals

• domainName:

The domain name on which the SPA will be accessible. For example app.example.com.

You can also specify a path: app.example.com/something. This can be useful to deploy multiple versions of an app in the same s3 bucket. For example one could deploy a feature branch of the SPA like this:

aws-spa deploy app.example.com/$(git branch | grep * | cut -d ' ' -f2)

Options

• --wait: Wait for CloudFront distribution to be deployed & cache invalidation to be completed. If you choose not to wait (default), you won't see site changes as soon as the command ends.
• --directory: The directory where the static files have been generated. It must contain an index.html. Default is build.
• --credentials: This option enables basic auth for the full s3 bucket (even if the domainName specifies a path). Credentials must be of the form "username:password". Basic auth is the recommended way to avoid search engine indexation of non-production apps (such as staging). You can also set credentials through the environment variable "AWS_SPA_CREDENTIALS" (the argument will override the environment variable if both are defined)
• --cacheInvalidation: cache invalidation to be done in CloudFront. Default is *: all files are invalidated. For a create-react-app app you only need to invalidate /index.html
• --cacheBustedPrefix: a folder where files are suffixed with a hash (cash busting). Their cache-control value is set to max-age=31536000. For a create-react-app app you can specify static/.
• --noPrompt: Disable confirm message that prompts on non CI environments (env CI=true).
• --shouldBlockBucketPublicAccess: This option will deploy the SPA with a bucket not being publicly accessible. Access to the bucket will be done through an Origin Access Control (OAC). Default value is false.

Migrate an existing SPA on aws-spa

aws-spa is aware of the resources it is managing thanks to tags.

If a S3 bucket named with the domain name already exists, a prompt will ask you if you want to deleguate the management of this bucket to aws-s3 (this will basically checks that s3 bucket is well configured to serve a static website).

If a CloudFront distribution with this S3 bucket already exists, the script will fail because CloudFront distribution update is quite complicated.

• If you don't care about downtime, you can delete the CloudFront distribution first.
• If you care about downtime, you can configure the CloudFront distribution by yourself (don't forget to gzip the files) and then add the tag key: managed-by-aws-spa, value: v1.

IAM

• cloudfront:CreateDistribution

• cloudfront:ListDistributions

• cloudfront:ListTagsForResource

• cloudfront:TagResource

• cloudfront:GetDistributionConfig

• cloudfront:CreateInvalidation

• cloudfront:UpdateDistribution

• cloudfront:ListOriginAccessControls

• cloudfront:GetOriginAccessControl

• cloudfront:CreateOriginAccessControl

• cloudfront:DeleteOriginAccessControl

• s3:PutBucketPolicy

• s3:GetBucketPolicy

• s3:PutBucketWebsite

• s3:GetBucketWebsite

• s3:DeleteBucketWebsite

• s3:PutBucketTagging

• s3:GetBucketTagging

• s3:DeleteBucketTagging

• s3:ListBucket

• s3:PutBucketPublicAccessBlock

• s3:CreateBucket

• s3:PutObject

• route53:ListHostedZones

• route53:CreateHostedZone

• route53:ListResourceRecordSets

• route53:ChangeResourceRecordSets

• acm:ListCertificates

• acm:DescribeCertificate

• acm:RequestCertificate

• lambda:CreateFunction

• lambda:GetFunctionConfiguration

• lambda:UpdateFunctionCode

• lambda:UpdateFunctionConfiguration

• iam:AttachRolePolicy

• iam:CreateRole

• iam:GetRole

If using simple auth

• lambda:GetFunction
• lambda:EnableReplication*
• iam:CreateServiceLinkedRole
• iam:CreateRole on resource: arn:aws:iam:::role/aws-spa-basic-auth-*

FAQ

Why not using Ansible, Saltstack, Terraform, Cloudformation, Troposphere, etc?

If it better suits your use case, these tools are probably a very good choice because there are done for this. Meanwhile there are some reasons why it is written in javascript:

• in my CI/CD installing Ansible, awscli or Terraform takes more than 1 minute. Since my SPA needs nodejs to be built, having a the same dependency to deploy is convenient & fast.
• Developers would have to learn these tools while they have already tons of things to learn. Using a script in the same language that they develop is nice.
• These tools are quite heavy while deploying a SPA requires only a couple of AWS API calls.