My company recently formed a Platform Engineering team to build an Internal Developer Platform (IDP) on top of Kubernetes. The goal was to simplify things, but now devs are complaining about "YAML fatigue" from the abstraction layers like Backstage or Crossplane. Are these portals actually helping your teams ship faster, or is it just another layer of complexity that devs have to learn?
3 answers
The "YAML fatigue" is real if you just wrap Kubernetes in another set of complex templates. We moved away from giving devs raw Helm charts and instead built "Golden Paths" in our IDP. Now, a new dev can provision a microservice with a database and a TLS cert via a simple UI form in Backstage. They don't even see the YAML unless they need to do something custom. It’s reduced our onboarding time from two weeks to two days. The secret is to treat the platform as a product—listen to your devs and only abstract what they actually find painful.
Melissa, how do you handle the "Day 2" operations? If something goes wrong with the pod, do the devs have enough visibility in the IDP to debug it, or do they still have to run kubectl?
We use Crossplane to manage our cloud resources via Kubernetes. It’s great because devs use the same GitOps workflow for both their app code and their AWS infrastructure.
Crossplane is the gold standard for unified infrastructure. It really helps in bridging the gap between DevOps and pure software engineering in a way that feels native.
Daniel, we integrated a "Service Health" plugin directly into Backstage that pulls logs and events from Prometheus/Loki. Devs can see 90% of what they need without leaving the portal. If it’s a deep networking issue, they still escalate to the SRE team, but for standard deployment failures, they are now self-sufficient.