Tried Flux's free tier to evaluate it for a team project. It's essentially a demo, not a functional testing environment.
* Max 2 active clusters? That's just a production and staging setup. No room for a disposable test instance.
* The 5-source limit includes Git repos and image repositories. A basic microservice setup blows past that immediately.
* You hit these limits in an afternoon. Then you're forced into a paid plan just to see if it works for your actual use case.
It feels designed to block evaluation, not enable it. You can't properly test integration, scaling, or failure scenarios. Ended up using the open-source version directly in a test cluster, which defeats the purpose of their hosted platform.
> "It's essentially a demo, not a functional testing environment."
Well, yes. That's the point of a free tier on a hosted platform. It's not a sandbox for full-scale chaos engineering. It's a taste, not a feast.
I've watched teams burn months trying to evaluate SaaS tools under free tiers that are deliberately limited to showcase the UI and basic workflow. The 2-cluster cap and 5-source limit you hit? That's not a bug. It's a signal that you're trying to run a production bake-off on a trial budget. The open-source version exists exactly for what you ended up doing - spinning up your own test cluster. You just saved yourself the vendor lock-in evaluation step.
The real question is: what were you hoping to test that required more than 2 clusters and 5 sources in a quick POC? If your microservice setup blows past that immediately, either you're bundling too many repos per service or your evaluation criteria are too broad. Flux's hosted free tier is a toy to see if the UI makes sense and the API feels right. For actual integration, scaling, and failure scenarios, you build a lab. That's always been the case.
What specific failure scenario were you trying to simulate that couldn't fit in a 2-cluster environment?
Test the migration.
> "That's the point of a free tier..."
I disagree on principle. A "taste" that can't swallow a basic integration test is a spit sample.
My rule of thumb: if the free tier can't handle the branching Git workflow (dev/staging/production) and image sources for a *single* microservice app, it's a sales demo, not an evaluation tier. You can't test promotion flows, rollbacks, or even a simple multi-repo setup.
You're right about the open-source version, but the whole pitch for the hosted platform is to avoid managing the controller yourself. If the free tier can't validate that value proposition, what's the point of it? It just forces you to set up the OSS version anyway, which is the very work you're paying them to avoid.
metrics not myths
Yeah, I've hit that same wall trying to test for my small team. You're right that two clusters just locks you into a rigid prod/staging setup. Where's the space for a quick, messy test run you can throw away?
It totally forces you down the OSS path just to see if the hosted version would even solve your problems. Kind of defeats the purpose, doesn't it? How big was the team project you were trying to evaluate for?
Right? The disposable test instance is the whole point. You need to be able to torch a cluster on a Tuesday to see if your rollback actually works, not just shuttle manifests between two pristine environments.
It's the failure scenarios you can't test. Can't simulate a source repo going down, or juggle image updates across three microservices. The limits make your test suite a puppet show.
So you end up running the OSS controller anyway, which honestly, once you've got that running you've done 90% of the work they're supposedly saving you from. Kinda funny.