alt Hacker NewsI really don’t understand why they positioned Openshift so far from the open source K8s project. It’s even more opinionated than AKS/EKS, and this says a lot.
The only reason I see is that they wanted more vendor lock-in, but that rarely pays in the long run.
Their goal is to create a system that's easy to run and manage even if you don't really know much about k8s - which I guess is a common thing, since k8s is a nightmare to manage directly until you learn everything about everything.
More importantly, they want to build an ecosystem around their ecosystem around k8s. They want people publishing their software to work on OpenShift (e.g. in the operator catalog), they have certification processes for software compatibility (e.g. CNI certification), and now that they've come as far as they have they're going after VMWare's lunch, madly scooping up every customer they can get their hands on to try to offer them a user-friendly kubevirt implementation to replace VMWare (and NSX).
If OpenShift was close to k8s then there wouldn't be enough differentiation between the two to make it worth the money so they're building out everything they think that execs and overworked infra teams will be willing to pay for to make k8s migration a workable option (rather than an internal migration project that stagnates and drags on for ten years).
OCP is just RHEL but for Kubernetes, and RHEL has done pretty well over the years all in all. RHEL is "just Linux", but it's "just Linux" with a bunch of RHEL-specific stuff in it and on top of it, and RHEL-specific policies around it, such that you can buy something from a software or hardware vendor that works with and is tested on RHEL and you know it'll work with your systems.
The goal is the same; company sells this software/solution, it runs on OpenShift v4.xx and up; you need RH's mesh and virt operators installed and configured, and then you can install our operator for our software and you're done. Now your whatever, database or CMS or whatever it is, just works. You don't have to customize the charts or define your own ingress gateway, it's all handled already if your system is set up properly. Then you get the meta-control panel that lets you see and manage all of your clusters and etc., so you can get a higher-level view, etc. etc.
They want vendor lock-in to an extent, but they've seen that what customers want isn't "Linux", it's a Linux distribution that gives them enterprise reliability (read: never getting new versions of stuff, only security patches) and compatibility guarantees from vendors. They're just trying to do the same thing with kubernetes. Will it work? Seems like it's going okay so far, I guess we'll see.