Alexis Richardson on GitOps, CNCF Evolution, and the Launch of ConfigHub

Hey everyone. We're back here. We're live at CubeCon.

It's, uh, Wednesday, day two of the, of the main, main, uh, floor being open and the keynotes and so forth. Let me introduce you to my next guest. His name is Alexis Richardson.

And, um, well if you've ever used Flux or Heard of Flux or GI Ops, you have, you know, a little bit. You can thank him. Alexis, thank you.

Welcome to Textron tv. Thank You Shimmy. I'm, It's great to Have you here.

I'm really happy to be Here. It's great. It's great to have you.

Um, before we get into your company and, and everything else, let's, let's talk a little bit about you. Thank you. If you wouldn't mind, I, you and I said some things, but share your story with our audience.

So I guess I started my career at Goldman Sachs. I was a trader. Really?

Derivatives. Yep. High stakes.

High risk. High high risk. Yeah.

High pressure. High pressure. Not a long lifespan though.

Very much not. No, no, no. And I, I actually found it boring.

Did you? Yeah. All you do all day, you sit at your desk, you have these boxes squawking at you, and you make a decision whether or not to buy yourself some derivatives.

So for all you day traders out there, and you know who I'm looking at? Boring. Boring.

Alright. And I wanted the connection with humanity. Uhhuh.

I wanted to build things that other people would get excited about Uhhuh. So I started a software company and I, that led me to create something called RabbitMQ, which is my first big success. It's still going cool.

As a piece of software stand. A piece of software that was acquired by VMware Uhhuh after, for a few years. Then I was part of the pivotal spin out.

I led really one of the app teams there, created the whole new generation of Spring Uhhuh product. I'm not an engineer, I'm a product person. Uhhuh.

And then I left to do a company called Weaveworks. Sure. Which is where we invented GID Ops and Flux.

Very Cool. Did the C-N-C-F-I was the chair of the TOC for three, maybe four years. Wow.

Helped get all this stuff going. It was, I went to the first Cube Con. It was a hundred people In San Francisco.

A hundred people. There's nothing going on. No, nothing.

See, I didn't start going, I think till the third one in. See it was Seattle. Yep.

There you go. So, I mean, that was big by then. Yeah.

Anyhow, so, so exciting to see the change. It's just all incredible. So anyway, um, after doing that, we created GI Ops and every single customer told me, you know, it's a good Alexis, I've got this automation.

I've got all these different tools I can choose. 'cause you did as an open thing. The problem is my ops team really hate me.

'cause you've got it. You're making them all learn this thing called yaml. It has been a scourge that people laugh about, I Don't wanna learn yaml, I don't wanna learn Terraform, I don't wanna learn vol, and if the systems go down, I want to go to the screen and I want to press the on off button until I get it right.

And so, you know, we, we, we call this being in configuration Hal config Hal. And that's what led me to do the next company, which is called Config Hub. So we being in it right up to today.

Up to today. Wait, I wanna go back a little bit though. Please do.

So when you say you, you know GI ups. Yeah. First of all, I'm not sure if everyone out here has the same definition of GI ups.

Yeah, yeah. What does GI UPS mean to you? Right, right.

GID ops is when you run, you operate a system automatically based on a plan which lives in a separate place. We call that plan the desired state that explains everything the system is supposed to do. It's the what, not the how, but the what that you're expecting to happen.

It's like, imagine you're running a Fractory floor and you have a plan of where all the robots are, where all the pipelines are, what everything is supposed to do. And then you have autonomous agents that use the plan to tell all the operating pieces what to do and continually check that they're doing it correctly. And they're always checking back to see if the plan has changed or if the, um, or if the operation needs to be corrected and they'll automatically fix any differences.

So it's always doing what we call continuous drift detection and reconciliation and that continuous loop of operations. That is GI ops. So it's automatic management of systems based on a plan which we put into Git when we wrote down the rules of GI ops in the GI ops working group in the CNCF, which you can go look at online.

It says it doesn't have to actually be, it could be any store that you use as a single source of truth for your plan, your desired state. So long as it has versioning, multiple users, non repudiation and all those valuable things that tell us this is a safe place to store your information, your plan about your running systems. But modern GI UPS includes people using other stores as well.

People are using GIT with Jfr now. Yes. Uh, OCI stores, I see people using it with Vault from HashiCorp with S3 on Amazon.

It's a little bit more complex than just one physical store, but you have one single view of the truth and that is your plan. And GI Ops is automatically managing systems based on that plan. It is not putting pull requests in GI No.

And, and, but to many people it is. Right. And that's why I kind of asked you and that was the money line right there.

Well, I posted a blog post at the beginning 2017, and it was called Operations by Pull Request. And that set off that whole pathway of misunderstanding. I gotcha, I gotcha.

Let ask you another question before we jump into config up. You know, the CNCF has two, I do know the CNCF. They got over 200 projects in here now.

Is it really that many now? Yeah. Yeah.

You could get lost in the pavilion over. Yeah, you Could. And that's only part of it.

But, um, and and they're, they're good about this, right? They have numerous service mesh projects. They're managing that.

You know, our first blush can be somewhat competitive. They have at least two major GI ups. Right.

Argo CD and Flux Right. As the creator of Flux just between you and me. Forget all the people watching this.

Okay. I'll just, Does that kind of p**s you off a little bit? Or you think they're handling it well?

Having both Of this? My fault. It's your fault.

Yeah. Why? Well, I'm not sure how, how I should say this.

One of the motivations when we started the CNCF, I was involved in the creation of the CNCF. Craig Mackey from Google was talking to me and he said, I, I wanna put Kubernetes in the foundation. And I said, well, I've just been a pivotal and we were creating the Cloud Foundry Foundation with IBM.

Yes. I've just had some recent experience of foundations. I think I can help you with that.

So he wrote down some rules and I said, look, it's really important to avoid the mistakes of cloud found past foundations. Yeah. Cloud Foundry made a couple mistakes.

Apache made some mistakes, eclipse made some mistakes. They All do. That's life.

But the big elephant in the room is OpenStack. You remember the all of the hope we had? Yes, I do.

Everybody thought, wow, this is gonna change things. What went wrong? It's, they've got it kind of back on track now, but for years they got lost.

Well, but let, let's be clear. For those of you I did don't realize Open Infra, which is now part of CN ccf That's right. Was actually the OpenStack That's right.

Foundation. That's right. That's Right.

The bigger thing swallowed the smaller thing. Yeah. With acquisitions.

So where OpenStack went wrong is they put, they said it's all gonna be one open product, one stack, and we'll have committees to define how everything fits together. And the more you do that, the, the harder it becomes to actually do any real work. Yep.

You just spend your whole time talking between different committees, Meaning that's by meanings. Anyone who's been working at a university will have seen this. If you've been in Europe and you live the European Union, if you'll have seen this Not-for-profits, All of that, you know, and you have these people who love that way of working and they make it worse.

Yeah. We call them bureaucrats. Mm-hmm.

So I said, look, what we're gonna do is we're gonna be project first in the CNCF. We will not assume that we know better about the future than anybody else. So we will not pick one thing and say that's the winner over another thing when it's clearly far too early.

OpenStack did that and they ended up creating a mess around the whole networking layer, for example. And they got held hostage by a couple of projects. So we said, we are too stupid in the TOC to figure out which one's gonna be, we're not gonna pick winners.

We're gonna let them figure it out in the community. Let the market, Let the market work it out. I'm such a big believer in that.

It's a democratic open model. Bit of capitalism, bit of democracy. And then the ti time will tell over time some things will run ahead, some things will fall behind and we won't make dumb errors like open telemetry.

The open telemetry story, I confess, I thought that was the dumbest proposal I'd ever heard. Clearly the team weren't sure what they were doing. And it was a very, very ill thought out set of standards.

But actually we also said as the TOC, but these guys, they have a right to try. Yeah. Let's let 'em have a go.

And they turned into one of the biggest successes. It now fi run the C ncf. Like OpenStack Open Telemetry would've been shut out the door on day zero.

It, We'd still be in horseless carriage, still Be horseless carriage. But you know what? So I have a lot of friends in the VC world.

Good. And one of the favorite things VCs love to do over drinks. Let's talk about the ones they just missed on Uhhuh versus the ones they thought were Sure.

Fire. Yep. Yep.

And, and how bad, how badly they miscalculated. Absolutely. Part of it.

It's part of it. But I, you know, as you sit here in retrospect, you do the best you can. Mm-hmm.

And, and at the end of the day, I do believe the market is the great sanitizer. Right. It doesn't care about your mistakes, it doesn't really, It's organized around incentives.

Yeah. That's the most important part. Absolutely.

Alright, let's pivot over to config hub now. Yes. Yes.

New company. New Company. I've never met a, a founder who wasn't extremely passionate about what they're doing.

Oh yeah. I'm passionate. Where's your, what's your passion on this one?

Um, we think that we can really change how operations works so that it is automated, managed, and safe as, as it's now gonna explode in the world of ai, right? Mm-hmm. What is missing is nobody has a good modern configuration database that tracks what every application is doing.

What what's stated is all in keeps it up to date automatically. So you just go look at that database. You have a fresh view.

This is what's happening out there in the, in, in my world, in your, and if something is wrong, you can change it. And if you're about to do something wrong, it says, hello, you're about to do, make a mistake. My, my policy detector has picked that up.

Are you sure you want to do this? 'cause it's gonna blow this up and you can stop it from happening when things do go wrong. The tool will also let you see which particular YAML file was responsible for screwing everything up for you.

You can find it, you can find the right line and fix it. That's a beautiful thing that we don't have today. What we have instead is people talk about yaml Hell yeah.

Configuration hell we see configuration sprawl, lots of systems, lots of tools, lots of files. Nobody can find anything. Three day outages, data loss, and now bring in ai.

You're putting this agent inside your organization. It could anything could happen. Agreed.

Agreed. We're gonna get rid of all of that. So given your history config hub, open source.

Nope. I knew you were going to say that. All right.

Why? Um, well one reason is market's much tougher now. 2025.

It was, it's different. It's Yeah. In Past, yeah.

You cannot do two things at once. If you do, you're gonna fail. Mm-hmm.

Remember Docker? Yes. Vaguely.

I think When you run an open source business, you're having to win on two fronts. You've gotta get the best open source project or the second best third will not be good enough. And you've gotta succeed commercially as well, which is a whole different thing, different set of features, people, sales, marketing, all of that.

And then you've gotta make 'em work together. And that's a really tough set of challenges. And they've got a, timing's gotta be right and so on.

And also, you've gotta be super patient. You've gotta spend three years making your open source project successful, then build your first commercial thing, then a bit more open source. It's a 15 year job.

Yes, it is. Okay. And I, I don't know if you've noticed that it's not The most Efficient.

Yeah, it's not the most efficient. And where do you want to be in 15 years? Well, on the beach.

Yeah. So we are starting out with a, with our commercial offering first really, really highly focused on user success. That's the big play go to go to the product.

It does some things for you and you can tell us if it's working or not. It's a SaaS, SaaS first product. I've teamed up with some fantastic people who've worked with Brian Grant, who was one of the originators of Kubernetes out of Google.

And Yesper Jergenson, who was the lead at Heroku Sure. Rolled out the Twilio platform. So we're bringing together a really good set of folks around this problem.

It's very exciting. I love it. We're almost outta time, but for people wanting to get more information on Config Hub, what's the website?

com. Easy enough. Yep.

And then how, how would they engage here? Engage. Follow me on LinkedIn, get in touch that way.

com or you can, uh, sign up for the product. com. It'll tell you where to sign up.

Those are the ways to engage. I love it. You'll come back, we we're gonna talk more about the history of GI Ops and other things, but we'll do it on the Tech drunk tv.

Okay. Thank you. We don't have all of this around us.

Alexis, Thank you very Much. It's been a pleasure. Thank you.

Little History, right here on Text Drunk tv. We're gonna take a break. We'll be back.

We've got day two coverage coming at you. You're watching Text Drunk tv.