Mysterious 500 error with AWS Lambda; unable to debug

Question

I have an API that I host using Lambda (nodejs), with API-gateway. I'm using serverless to deploy.

Generally things have been fine, but while I was working on a specific function today, I started to receive HTTP 500 errors when hitting the endpoint. However, while there were still API-Gateway access logs for the end point, there were no Cloudwatch logs for the lambda functions getting hit. I was able to verify that the Authorizer was getting hit successfully, and not returning any issue (if it was, it would have been a 401). After using CLI tools to invoke the function from the command line, the 500 error went away and I was able to successfully hit the endpoints again.

Has anyone ever ran into this before? If I'm missing a debug step, I would really like to know. It was really concerning that my API could be generating 500 errors with no paper trail to help me understand what was happening.

Answer 1

I had the exact same problem i.e. Authorizer running successfully but request returning 500 error without actually hitting the target endpoint lambda. After some trial and error I found that its occurring due to inclusion of array and object based properties in 'context' part of the IAM policy which authoriser was generating. As soon as I replaced array and object properties in context with string values it worked like a charm.