Hmm, lots of things have happened in this space since I last looked, found so far: • https://github.com/babashka/pod-babashka-aws • https://github.com/FieryCod/holy-lambda • https://fierycod.github.io/holy-lambda/#/

If I was doing this today, I would copy the ions approach.

Dumb lambda talking to an autoscaling clojure application running on EC2 or containers or whatever is cool.

Is there a but there? 😃

But I don't have to suffer lambdas 😅

Ah ok 😃

Sure that's fair, at the moment it seems like lambdas will be enough, at some point we'll transition, perhaps to just using ions. Not sure on that front. I think the real question is how hard will it be to debug, lambdas can be really finicky to debug I've found.

Though maybe just a dumb server that's just sitting around waiting for work...

Yeah, that's why I like the option of having a dumb lambda that just forwards everything.

You make it really good with fantastic logging and then use it everywhere.

God, logging is still super painful as well, I don't know how people are productive with cloudlogs, all of the ui for them I've ever seen has terrible defaults... That's not just AWS, all cloud logging I've ever seen is just >_<...

Ok, one completely golden thing so far is the REPL seems to both working and a valid representation (so far at least) of what actual lambda code looks like...

If the orchestration part is significant to you, you may wish to look into https://aws.amazon.com/step-functions/