rationalise lambda deployment

[danmermel]

There seems to be a lot of repetition in this... we can turn it into a more streamlined script.

• The ./deploy.sh should have an array of clue types (anagram, container etc).
• It should loop through this list, performing the tasks currently in ./<cluetype>/prepare.sh
• then it should deploy the built zip and tidy up

This will save us modifying bash scripts many times.

[danmermel]

have started this... now need: Potentially remove docker bit Have only one package.json and one npm install rationalise the prepare script (see above)

[danmermel]

There is another angle here: can we let terraform do the deploy of lambda?

This TF will zip up a directory

data "archive_file" "triggerOnUploadLambdaZip" {
    type        = "zip"
    source_dir  = "lambda/triggerOnUpload"
    output_path = "lambda/triggerOnUpload.zip"
}

and then you can add the code like this...

resource "aws_lambda_function" "funguyLambda" {
  filename = "lambda/triggerOnUpload.zip"
  source_code_hash = data.archive_file.triggerOnUploadLambdaZip.output_base64sha256
  function_name = "triggerOnUpload"
  role = aws_iam_role.funguyLambdaRole.arn
  description = "triggers meshroom build on file drop in funguy S3"
  handler = "index.handler"
  runtime = "nodejs12.x"
}

I cannot quite remember why we did it separately and/or whether the above will do us... to discuss!

[glynnbird]

You can run terraform on deploy (in Travis) if you like - unfortunately, it doesn't work properly (the taint thing....). But it could upload the code too if it wanted. We currently have a separation between infrastructure which is triggered manually and the code which is automated.