Question

A DevOps team runs 40 CodePipeline pipelines across multiple applications. When a pipeline stage fails, engineers currently discover it hours later. Leadership wants an automated system that, within seconds of any pipeline stage entering a FAILED state, posts a message to a Slack channel that includes the pipeline name, the failed stage, and the failed action's execution ID so the on-call engineer can jump straight to troubleshooting. The solution must require no code changes to individual pipelines and must scale as new pipelines are added. What is the MOST operationally efficient approach?

Accepted Answer

Create a single EventBridge rule matching the 'CodePipeline Action Execution State Change' event with state FAILED on the default event bus, targeting a Lambda function that parses the event detail and posts the pipeline name, stage, and action execution ID to Slack via an incoming webhook.

Answer

Add a manual approval action with an SNS topic to the end of each pipeline so that failures trigger an email that engineers forward to Slack.

Answer

Enable CloudTrail data events for CodePipeline and configure a CloudWatch Logs metric filter with an alarm that publishes to an SNS topic subscribed by a Lambda function that posts to Slack.

Answer

Create a separate EventBridge rule for each pipeline that polls the CodePipeline GetPipelineState API every minute using a scheduled rule and invokes Lambda when a stage reports Failed.