Pros: * AWS config done using yaml files that can then be version controlled. * Avoid the horrible AWS GUIs. * Easier migration to other cloud providers (not sure how true this is).
Cons: * Potential lag time between AWS features and serverless? * Additional dependency on startup that might go bust. * Diverging from "standard" way of doing things? * Others?
What have you guys chosen to do?
The problem with the console is that some options might be missing. For instance, when creating a cognition userpool, any auth0 standard attribute may be set to immutable; however, this option isn't available in the console. Serverless might be missi g options, but it allows that to be supplemented with writing raw cloudformation in the serverless document.
I enjoy the CDK because it's one less layer of abstraction to deal with compared to serverless, and what's actually being deployed isn't hidden behind the yaml. It's programmatic, maintained by aws, and the documentation is pretty stellar. The only downside is that it's not complete, so sometimes raw cloudformation will have to be written.
Another thing to keep in mind that serverless and cdk allow for you to recycle time spent while just using the console does not in lost cases.
Say what ?
"What have you guys chosen to do?"
If you are making heavy use of the AWS GUI for any aspect of AWS then You're Not Doing it Right (TM).
Use the AWS SDKs, or if you don't like the SDKs, the APIs.
Hell, if you can't be bothered to program against the SDKs or APIs, then just make naïve shell script calls against the AWS CLI for your platform.
That's what all the native tools Amazon gives you are there for. No need to add un-necessary layers of third-party abstraction, which, above anything else, is a vast potential security risk.
RTFM as the cool kids say. ;-)