Install serverless

You need to have serverless installed and your AWS Credentials in place.

Serverless requires an IAM user that is used to build the stack. You can use their recommended policy or this one.

Deploy stack

The stack consists of 1 sqs queue, 1 dynamodb, 7 lambdas, 1 s3 bucket & an API Gateway with 6 routes

serverless deploy

or with options (default values shown)

serverless deploy --stage dev --bucket cdeno --region eu-west-2

AWS Bucket names are unique so choose something that's unlikely to clash.

The stage is suffixed to the bucket name i.e. cdeno-dev

Test stack

If the deployment worked serverless should report the API gateway endpoints. To test the stack works, try POSTing to the /register endpoint using the following body:

{
  "url": "https://github.com/denoland/deno_std/releases/tag/v0.8.0",
  "description": "deno standard modules"
}

Wait a few seconds and then check your resources. The s3 bucket should have a copy of the source code and the dynamo tables a registry entry for the module & version.

Remove the stack

serverless remove

Should I use this?

Probably not, no. It was a learning exercise to understand more about how serverless works with AWS Services. It needs lots more rigor to be useful. I was interested in building a registry of sorts for deno and things went from there.

If you are interested in using s3 as a backup store for your code files or CDN then I'd recommend this AWS Quick start guide. It fairly easy to set up and supports other SCP like bitbucket and tfs. It doesn't have a notion of a registry though, it's just a secure and robust way to get files from a git repo to s3.

If you do go ahead and install the cdeno stack, please know the following:

• There's no security around the endpoints, significantly the two POST endpoints. Basically, anyone with a link can put files into your bucket.
• There's no security around the /webhook endpoint, nor is there any validation that it's really github making the call.
• No credentials are used when pulling the git repo
• The bucket is public read

Known issues

• The AWS CLI aws s3 sync command is used to sync the extract release.zip to s3. It has mime-type guessing built in but it's not great. At the moment, it's disabled and content-type 'text/plain' is used.

Fun

• AWS Lambdas don't have the aws cli pre-installed. Getting this to work for someone like me who doesn't know much python was a war of attrition but lots of fun. It involves building a virtualenv and following quite a few steps. It took me ages to get it right - if you have a need to run the aws cli on lambda, it's now easily packaged as a zip in this project. You can use it as a lambda layer in serverless like this

Thanks to

• https://bezdelev.com/hacking/aws-cli-inside-lambda-layer-aws-s3-sync/
• https://alestic.com/2016/11/aws-lambda-awscli/
• https://stackoverflow.com/questions/33513604/call-aws-cli-from-aws-lambda

MIT Licence