Thread by @rossalexwilson

I stood up a Kubernetes cluster from scratch this evening, to get a better understanding of the internals. You see, I’m building a new web service, and I’m thinking about deployment strategies. K8S was top of mind. Here’s some thoughts.

Firstly, it’s a pretty involved process, with a bunch of decision points at every turn. I was console and CLI prodding, but I’d want to script all this. I value infrastructure as code too. That’s another rabbit hole: CloudFormation vs Terraform vs CDK vs DIY.

All the big cloud providers have managed K8S solutions, but I can’t help shake the idea that they’re not *quite* there yet. For example, IAM Role integration just doesn’t seem fully baked. K8S feels like a leaky abstraction (on AWS, at least).

@QuinnyPig has discussed how K8S provides foundational constructs that were envisioned to be built upon, perhaps by cloud service providers. Perhaps one day K8S will be a seldomly discussed concept - an abstraction later that has becomes dull.

Ultimately I’ve spent a few days thinking about how I’m going to run my relatively simple @elixirphoenix service. The fanfare of serverless is ringing, and right now I wish there was an obvious go-to solution for non-serverless workloads.

I thought I wanted infrastructure as code, but maybe I don’t actually care about my infrastructure. I just have some code, that serves some business value, and I just want it to run somewhere. Along with being scalable and resilient to underlying failure.

For now, I think I’ve settled on a set of CloudFormation stacks, managing an ASG with the AWS-provided ECS AMI. I’ll run my service inside containers, and use ECS to orchestrate zero-downtime deployments.

Thanks for sticking with me, as I dump my thoughts down here. I’m still thinking about what modern deployment/infrastructure strategies look like for a small scale outfit. There’s piles of large-scale examples, but fewer targeting solopreneur side gigs.