Amazon CloudWatch -> Rollbar

This project uses the Serverless Framework to deploy an AWS Lambda function which subscribes to Amazon CloudWatch log groups, consumes their log events, and continuously pushes them into a Rollbar project.

Get Started

You will need to add two configuration files to run this project: a dotenv, and a YAML file.

Environment config

First, create a dotenv environment file such as .env.{stage} (e.g. .env.live) in the project directory. Use the following configuration as a template to get you started, adding your own values where appropriate:

STAGE=live
LOG_LEVEL=DEBUG

ROLLBAR_ACCESS_TOKEN={POST_SERVER_ITEM_ACCESS_TOKEN}
ROLLBAR_ENVIRONMENT=production
ROLLBAR_LANGUAGE=Java
ROLLBAR_FILTER_THRESHOLD=WARNING

The ROLLBAR_FILTER_THRESHOLD variable will set a threshold to limit the number of log messages sent to Rollbar. The threshold will send higher priority messages and is inclusive of the value. e.g. WARNING will send WARNING and ERROR logs. INFO will send INFO, WARNING, and ERROR logs.

Serverless config

You will then need to add a .env-config.yml file, to specify the CloudWatch log subscriptions to subscribe our AWS Lambda to, in the project's serverless.yml file.

cloudwatchLogs:
  - cloudwatchLog: /aws/lambda/my-project-${self:provider.stage}-execute

Deploy to AWS

$ serverless deploy -v --env=live

How it works

Determine Log Levels

Amazon CloudWatch logs are provided to AWS Lambda as strings, and so we need to parse the message to determine the logging level, to better inform Rollbar. We analyse the first 200 characters of the log message, and look for standalone keywords, such as CRITICAL, SEVERE, FATAL, and WARN etc. It's typical your log messages would have these log level keywords. If we are unable to determine the log level, we default to INFO.