This repository contains the CloudFormation template to create:

1. a VPC
2. two private subnets
3. a HA RDS cluster running Aurora within the private subnets
4. a role for a lambda function
5. a lambda function that can connect to the Aurora DB

The purpose of this is to illustrate how complex batch jobs with access to dark corners of your team's networking can be replaced by lambda functions for the purposes for cost savings.

Read more about this in our blog post.

This code designed to construct an example of a lambda function connecting to resources in a VPC. To reuse the code with your existing VPC, just take the part of the template that defines the lambda execution role, the target assumed role and the lambda function itself.

Run npm install to install all dependencies for this repository.

In liew of support for ecmascript2015 in the lambda execution environment, we transpile our es6/ecmascript2015 code to es5. We have made a grunt task to kick off this transpiler in watch mode, so during development, as you save your files, they are transpiled automatically.

# if you have grunt installed globally already:
$ grunt run:babel

# otherwise, use the local grunt module:
$ node_modules/grunt-cli/bin/grunt run:babel

NB: your code must have been transpiled already in order to successfully run in s3.

Use the zip_lambda_code.sh script provided in the repository to zip all relevant code and modules for lambda.

$ bash zip_lambda_code.sh

When creating a lambda function from s3 code, we need the s3 bucket to be in the same region as we are running our CloudFormation stack. Using the AWS CLI, take a look at what buckets you have available to store the lambda code:

$ aws s3 ls

2016-02-19 11:21:06 cf-templates-11111-ap-northeast-1
2015-10-08 11:59:04 cf-templates-11111-us-west-1
2015-08-16 12:51:34 random-logs
2015-10-13 22:38:14 some-bucket
...

If there is no bucket appropriate to store your code, create a new one.

$ aws s3 mb s3://my-lambdas --region <region_name>

make_bucket: s3://my-lambdas/

$ aws s3 cp lambda.zip s3://my-lambdas

upload: ./lambda.zip to s3://my-lambdas/lambda.zip

Unfortunately, CloudFormation does not yet support the creation of events and rules. For now, we have to create this via the AWS CLI.

$ aws events put-rule --name fire-every-5-minutes --region <region_name> --schedule-expression "rate(5 minutes)"

{
    "RuleArn": "arn:aws:events:<region_name>:<account_id>:rule/fire-every-5-minutes"
}

The example CloudFormation template in the file 'template.json' builds a VPC with two private subnets, containing a two node HA-Aurora cluster and our lambda function, with the CloudWatch alarm periodic trigger.

Feel free to use the template as it comes, but you will probably just want to lift out the appropriate sections and use in your own existing CloudFormation templates.

Most important one to substitute is the RuleArn from the step above.

This new VPC configuration option for lambda functions is not yet exposed via CloudFormation. Real pain! While we wait for its implementation in CloudFormation, add this configuration via the AWS CLI as follows:

$ aws lambda update-function-configuration --function-name <function_name> --vpc-config SubnetIds=string,string,SecurityGroupIds=string,string

In our CloudFormation template, we create a CloudWatch event rule to periodically create an event. Because there is no support for these event sources anywhere other than in the AWS console, this is one manual step we must make via the console.

1. Under your Lambda function, go to 'Event sources'.
2. Click 'Add event source'
3. Select your event from those listed