Question

A DevOps team uses AWS CodePipeline to deploy a serverless application. Between the build and deploy stages, they add a CodePipeline Lambda invoke action that triggers an external load-testing service via its REST API. The load test typically takes 15-20 minutes to complete, well beyond a single Lambda invocation's practical duration. The team notices the pipeline stage is marked as 'Failed' almost immediately, even when the load test itself later passes. How should the Lambda function be designed so the pipeline correctly waits for and reflects the external load test result?

Accepted Answer

The Lambda function should start the load test and, while it is still running, return without calling PutJobSuccessResult or PutJobFailureResult, and instead re-invoke itself using the continuationToken until the test finishes, then report the final result.

Answer

The Lambda function should start the load test, then immediately call PutJobSuccessResult so CodePipeline continues; a separate process monitors the test asynchronously.

Answer

The Lambda function should synchronously poll the external service in a loop and increase the Lambda timeout to 30 minutes so it blocks until the load test completes.

Answer

The Lambda function should start the load test and throw an exception so CodePipeline automatically retries the action every few minutes until the test passes.