Hi hi — Medium commenting is a little weird.
We were on the API Gateway/Lambda combination very early and we had all kinds of bizarre issues with it. Among those were getting the request and response objects properly. On our long list of grievances, corruption was one of them. I’m sure Amazon has since fixed those issues, but not before we ended up jettisoning API Gateway entirely for various reasons. I’m still not sure I can comfortably recommend the pattern even though it’s now an AWS architectural standard. But that taught us that the bleeding edge on AWS services is not always the place we want to be.
And, of course, we still grab beta products and muck with them today…. Athena being our favorite toy. But year-ago EMR, year ago Cognito, nothing but pain.
Logs were a different beast altogether. While they absolutely went into CloudWatch, and in they went, we found the console interface totally impenetrable and largely useless for our needs. Part of it was the nature of the logging format (not great), part of it was CloudWatch’s nature to be bursty, part of it was the sloggy search interface, and part of it was the terrible UI on CloudWatch itself. We solved it by moving all the logs on SumoLogic, but that took a while because we were on an experimental forwarder running on… Lambda. Now we’re forwarding logs via Lambdas to Splunk and that works perfectly fine but not before going through a couple of iterations to get it right.
