Project Lightning Talk: Multi-Cluster Configuration Management With KubeStellar - Andy Anderson

This talk focuses on Kubestellar, specifically the Kubestellar console, which is a cutting-edge Kubernetes management and orchestration solution. The speaker shares lessons learned from their previous work on Kubestellar legacy, emphasizing the integration of AI and cloud-native technologies. The console is designed for multi-cluster access, allowing users to manage their Kubernetes clusters effortlessly while ensuring security through configuration permissions. It employs AI-powered workflows supported by tools like GitHub, Claude, and Codex, minimizing the need for traditional kubectl commands. With support for over 185 CNCF projects, the console enables users to execute complex tasks through simple abstracted requests, allowing for improved deployment and compliance solutions. The platform also boasts a community-driven marketplace and a knowledge base, making it suitable for modern teams with diverse operational needs.

Hi everybody. Welcome to Amsterdam. I love it here. This is second or third time I'm here. Love this town. All right, I'm here to talk to you today about Kubestellar, but specifically about Kubestellar console. So, we uh we did a lot of work in the past on something we call Kubestellar legacy, which just really focused on deployment and orchestration. But this time around I'm coming to visit

you on lessons learned from that and where we've headed next. So, this is all about AI and cloud native, the marriage of it. And this is the world's first fully integrated and automated Kubernetes management and orchestration solution. Yeah, some of you might think it's AI slop. We don't allow human PRs in this repo. So, you can all gasp now. Ooh, how's that happening? Anyway, this is based

on my experience of almost 10 years of working on Kubernetes clusters. I'm tired. My hands have got arthritis. I don't know about any of you, by show of hands, are you tired of running kubectl commands? Yeah, there's a hell of a lot more than you than that. Come on. Admit it. Okay. So, what we designed this console to do, right, is it's multi-cluster first. All right. So,

that means you're going to have all the the access to all of your clusters that you have that you have organized in your kube config. And yes, it dedupes them so you don't get those multiple versions of the same cluster. Okay? This only has access to what you have access to. So, if your kube config and your configuration and your API only allows you to have read-only

access, you only have read-only access. So, we're not exposing anything more than what you naturally have exposure to. This is fully AI-powered as I mentioned before, no human PRs are allowed. We use GitHub agentic workflows, which is fairly new. Has anybody heard of GitHub agentic workflows? A scattering. Okay, not many of you. Okay, and that uses, of course, you know, GitHub agent underneath the GitHub agent workflows

we use things like Claude and Gemini and Codex. And also we use Copilot underneath the covers, which is also using Claude Opus and Gemini and Codex. Okay? So, you can use natural language to design missions that use our MCP servers, which are deployed in ops, which will go out and do them some of the most common exercises that you would normally do with Kubecuddle, but does it

be some behind the scenes. Now, you're saying to me right now, you're probably looking at me going, "Why don't I just use Cloud CLI? Why don't I just use Gemini? Why don't I just use VS Code? Why don't I just use any of these number of things, right?" Well, the answer is really simple. You can work to design prompts over and over again and maybe get lucky

and be able to save them away in an MD file, and maybe you'll remember where that MD file is someday. Or, you can use our console, which automatically abstracts it or supplies it to you, and then you get to save it and share it either publicly or privately. So, that's the real gist. We have support for over 185 of the CNCF's projects. That's There's 210 of them

right now. So, this really unlocks You get away from running like Kubecuddle get pods, right? Or, uh Kubecuddle apply. You get away from that. Now, you're more abstract. You can say, "I want a compliance solution. I want to install Trestle, for instance. Or, I want to have OPA. I want to have Kyverno, etc. I want to have these All of these work together in concert. Or, maybe

I just want to deploy things. Maybe I need a combination of Argo and some other things, right?" So, the the idea is to unlock the ability for you in this audience to use aggregate functions and aggregate projects to do aggregate things, because we don't operate clusters just to do one thing or the other. We many times have to do a lot. So, this is entirely community-driven. We

have a full marketplace and a full knowledge base. I invite you to go check it out. Go to console.kubestellar.io and click on the demo mode, and instantly you'll be given a nice little blurb about how to run that curl command that'll install this in 60 seconds or less. This is what the main dashboard looks like. One of the key things here is that you see on the

left-hand side, you have these dashboards. There's about 30 of these at this time, okay? And that number's I try to keep that around that same number just so there's not too much of a way of sprawl. Don't try to overwhelm you at first, you can discover these by adding more in that bottom left-hand corner. There's stat blocks, these things at the top. These give you quick awareness

of what's going on. They change dynamically based on the number of pods or deployments or whatever else you're monitoring. And then there's these things called cards, and those cards I'll show you in just a second what they do. So again, just to reinforce this, we have this notion of AI missions, right? So you're deploying Kyverno to all my prod clusters. You're which GP which clusters have the

GPU capacity my my my customers are looking for. How many of you have been asked to run an ISO audit compliance recently? How long about how long did that take you? A long time. Well, what if you could see the answer to that in just seconds? It's always on, right? So build audit enforcement policies requiring owner metadata on all deployments. These are the type of meta descriptions

that you could run through the console and monitor simultaneously. So you save and share your missions as I mentioned to you with others. There's the AI missions on the right. So this is a modern platform for modern teams. You know, don't approach this if you really want to just run independent coop cuddle kinds of commands. I invite you to do this only if you're looking to satisfy

your customers with aggregate functions and aggregate projects. I'll leave you with this. We have a leaderboard. It's at It's at coopstella.io/leaderboard. You can join our leaderboard. I promise you this, every one of you There's probably about 200 of you in this room Each and every one of you, I challenge you to open a legitimate bug. It will be implemented by the end of the week. And by

the way, we offer white labeling. So, if any of one of you wanted to steer this project to help your enterprise solve big problems, I promise you, come see us, we'll make it happen for you. Thank you so much.