awsletter

Complete newsletter application that has no "old-style" backend components. See how does it look like. All features are implemented using Amazon's cloud stack.

To be more specific

• Login with Amazon is used as an authentication provider
• S3 bucket is used to store images
• SES is used to send e-mails
• DynamoDB is used to store e-mail addresses (not implemented yet, using mock)
• Amazon Lambda is used to resize images and communicate with APIs (SES, Twilio)
• AWS SDK for JavaScript is used to call Amazon services both in client applications and Lambda functions
• additionally, Twilio can be used to send SMS notifications

Configuring AWS

It's the trickiest part of the whole setup. Steps depend on each other, so do it in presented order.

Preconditions

• remember to create all services in one region (that will be yours aws.region in config file)
• create "Login with Amazon" application (amazonLogin.clientId)
• create S3 bucket for images (aws.imagesBucketName) and add CORS config (aws/images-bucket-cors-config.xml)
• verify domain or e-mail sender address for SES (sender property in config)

Create IAM role for Lambda

• this role will be used by Lambda when executing functions
• create awsletter_lambda_executor role (of course you can name it as you wish)
• add policy (aws/lambda-executor-policy.json) to this role; provide template with a real bucket name
• add trust policy (aws/lambda-executor-trust-policy.json) to this role; it states that Lambda can assume this role

Create Lambda functions

There's a list of functions to create:

Function name	Code	Memory	Timeout	Notes
awsletter-send-mail	lambdas/send-mail.js	512	5
awsletter-send-sms	lambdas/sens-sms.js	256	10	Provide Twilio credentials
awsletter-resize-image	lambdas/resize-image.js	512	15	Provide image width

Please note:

• previously created role (I call it awsletter_lambda_executor) should be used for all functions
• names of functions can differ in your setup, but remember that these will be used as aws.mailFunctionName and aws.smsFunctionName in config file
• memory and timeout are values that I use, no guarantee that these will work well for you

Create invocation role for resize function

• this role allows S3 bucket to call resize function
• create awsletter_lambda_resize_caller role
• add policy (aws/lambda-resize-caller-policy.json) to this role; provide template with function name
• add trust policy (aws/lambda-resize-caller-trust-policy.json) to this role; provide template with bucket name; this policy states that S3 bucket can assume this role (so it can invoke resize function)

Create web role

• this role is assumed when user logs into client application
• it allows to manipulate bucket and invoke functions
• create awsletter_web_user role
• add policy (aws/web-role-policy.json) to this role; provide template with bucket name and function names
• add trust policy (aws/web-role-trust-policy.json) to this role; provide template with "Login with Amazon" Client ID

That's a lot to do

Yeah, I know. I'm working on automating this process. In case of any (project-related) problems, you can contact me, I'll try to help.

Setting up client locally

Just clone repository, copy reference config file and adjust it:

$ git clone git@github.com:jelz/awsletter.git
$ cd awsletter
$ cp config/config.js.dist public/app/config.js
$ vi public/app/config.js

Then you need to serve public/ directory. Use python -m SimpleHTTPServer or grunt (default task runs server).

How does it look like?

List of recipients

Image management

Composing an e-mail

Dispatch summary

Author

@jelzbieciak

License

MIT