If you’re interested, I can create a gist with an example.

There’s not a real standard yet. OAuth by itself is not enough. The credentials provided could be exposed rather easily. Security guardrails are needed badly.

www.reco.ai/learn/mcp-se...

It’s quite simple really. The handler function signatures are slightly different from one cloud provider to another but it’s real easy boilerplate to handle the differences. Access to path, query parameters, body and headers are very similar. The responses are even easier.

Absolutely!

Vendor lock-in is such a weak argument in regards to serverless. It’s usually promoted by devs that don’t know how to do serverless in multiple clouds.

Yeah that’s only true if you get a lot of frequent traffic. Functions destroy servers especially when there is a lot of troughs in traffic. Functions can run as containers so probably best to point to the specific type (ex. K8s vs container functions like cloud run and azure functions by container.

There are ways to run containers as lambdas in AWS and as functions in Azure and they are significantly cheaper.

Yeah I hear that quite a bit. It’s a shame because there’s no silver bullet and use cases vary widely.

Cache the connection in memory as a global. That will help tremendously. Also ensure the database connection has the best route in the cloud. Private endpoints are not only secure but performant.

🙂 The path people take that don’t really understand serverless workloads. Cold starts are easy to manage. 🤷🏻‍♂️

I would not use v2 any longer. I have this open-source repository that may come in handy: github.com/kevboutin/az... the conversion is in the git history as well.

You can use ACR (azure container registry) for your azure functions. I do this today with most of my function apps. Just build the container from azure’s base image and send it to ACR. Automatic deployment.

It’s extremely similar to how I build lambdas. In fact, I can write the same function to work in both with very little boilerplate. Not sure what your issues were writing azure functions.

Or deploy the container to lambda

What is stopping you? Azure and AWS have base container images. The secret sauce is in managing the infrastructure, monitoring and deployment to it.

You have to learn the yaml operations. It’s not going to configure itself. If it’s taking you 5 minutes to test a change, you’re doing something wrong.

GitHub actions wins on just about every metric. Throw GitLab into the mix and it still does. It’s not even a close comparison unless you get GitLab into the mix. Azure DevOps is very clunky and Jenkins is for dinosaurs.