AWS infrastructure to ship CloudTrail logs from S3 to Kinesis

This repository contains a Terraform module to ship CloudTrail logs stored in a S3 bucket into a Kinesis stream for further processing and real-time analysis.

Use-cases include shipping CloudTrail logs of an organization trail to a SIEM (such as Splunk) or third-party service provider.

Architecture

In summary:

• A S3 bucket notification is configured to trigger a SNS event every time AWS delivery a new CloudTrail log file to the S3 bucket.

• A subscription is set up so that a Lambda function is invoked for every SNS event, i.e. every time AWS delivers a new CloudTrail log file to the S3 bucket.

• The Lambda function picks up the information that a new file was uploaded, reads this file from S3 and writes it to Kinesis.

Pre-requisites

• A CloudTrail organization trail must already be configured to ship organization-wide CloudTrail logs to a S3 bucket

• This S3 bucket ("CloudTrail S3 bucket") must already exist

• The CloudTrail S3 bucket must be in the same AWS account and region as the one the module is instantiated

Usage

Inputs

Outputs

Example usage

Sample usage:

provider "aws" {
  region = "eu-west-1"
}

// Ship CloudTrail logs that AWS ships in "my-organization-trail-s3-bucket" into a Kinesis stream "cloudtrail-logs"
module "cloudtrail-to-kinesis" {
  source                      = "github.com/nexthink/terraform-aws-cloudtrail-s3-to-kinesis?ref=v1.0.0"
  cloudtrail-bucket-name      = "my-organization-trail-s3-bucket"
  kinesis-stream-name         = "cloudtrail-logs"
  kinesis-retention-time-days = 7
}

Monitoring

Logs

Logs of the Lambda function are sent to a CloudWatch log group. Invocation rate and success/error rate can be seen directly from the AWS console, under the "Monitoring" tab of the Lamdba function

Observability

The module creates a custom CloudWatch metric NumberOfCloudTrailRecordsShippedToKinesis, allowing to have visibility on the total number of CloudTrail events shipped by the Lambda function. When used conjointly to the default AWS Lambda CloudWatch metrics, it allows to easily build a CloudWatch dashboard to monitor the volume of logs processed by the Lambda function, as well as any potential errors.

Unit tests

Run the Lambda function unit tests using:

$ cd lambda
$ make test
AWS_SECRET_ACCESS_KEY= AWS_ACCESS_KEY_ID= python -m unittest discover tests -v
test_extract (test_extract.TestExtract) ... ok
test_entrypoint (test_lambda_entrypoint.TestLambdaEntryPoint) ... 2021-04-09 16:38:33,062 Reading CloudTrail log file s3://my-cloudtrail-bucket/dir/to/cloudtrail.json.gz
2021-04-09 16:38:33,066 Writing CloudTrail log payload to Kinesis (27596 bytes)
ok
test_writes_to_kinesis (test_load.TestLoad) ... 2021-04-09 16:38:33,124 Found credentials in environment variables.
2021-04-09 16:38:33,155 Writing CloudTrail log payload to Kinesis (24 bytes)
ok
test_returns_correct_data_from_s3 (test_transform.TestExtract) ... 2021-04-09 16:38:33,236 Found credentials in environment variables.
2021-04-09 16:38:33,300 Reading CloudTrail log file s3://my-bucket//path/to/cloudtrail-file.json.gz
ok

----------------------------------------------------------------------
Ran 4 tests in 0.532s

OK

End-to-end tests

End-to-end tests use Terratest to spin up real infrastructure, create a S3 bucket, ship a sample CloudTrail log file to it and check that the expected record is produced in Kinesis.

# Pre-requisite: ensure you are authenticated against a test/CI AWS account
# $ aws-vault-exec ci-account

$ cd test/
$ go test

End-to-end tests take around 2-3 minutes to run.

Examples

See the examples/ folder for complete usage examples.