What’s the largest Step Function state machine you’ve encountered?

I've come across Step Functions that have a couple hundred states in production. Honestly, when they get that large, you start to sense that it might not be the best approach. The more states you add, the trickier it becomes to debug and manage. It's usually a sign that too much business logic is crammed into one state machine instead of breaking it down into smaller, manageable parts. It's best to use Step Functions for orchestration rather than as a catch-all solution. Keep it simpler by chaining smaller functions together—it’ll save you headaches in the long run!

Yeah, I've definitely seen several Step Functions with hundreds of states, and I wouldn’t be surprised if there are even bigger ones out there. Whenever we start designing with Step Functions, we usually pull back to avoid going too big.

The largest Step Function I personally saw was a POC for handling insurance claims. It involved several steps for human intervention, file uploads to S3, and all kinds of validations. However, it ended up being scrapped due to time constraints.

At one job, there was a Windows image builder Step Function with about 20 steps. It wasn't excessive, but it took around 3 hours to complete due to all the reboots windows goes through!

I once built a Step Function to decommission IoT devices, and I used CDK for that. It worked, but I'm not sure if there are larger implementations out there. I did see teams with longer chains in the same organization!