Kubernetes v1.32 with Frederico Munoz

This is Text Strong tv. Hey everyone, welcome back here to Techstrong tv. You know, I'm really excited about this next interview.

I have been since we originally scheduled this one. I want to introduce you to Frederico Munoz. Uh, Frederico is the release lead for Kubernetes.

That is a lot in there, but let's welcome him to the show and find out a little bit about what he has to share with us today. Frederico, welcome to Text Drunk tv. It's great to have you on.

Thank you very much, Alan. Thank you for having me. So, as I mentioned, your release lead for Kubernetes, but you know, how does one actually get to be the release lead for Kubernetes?

What does it really mean? Tell us a little bit about your journey here today. Uh, I mean, a fair question and one that I sometimes, uh, also think about it, uh, because it's, it's not something that I, I, I would think if you would ask me two years ago, uh, I would say that, that that was impossible.

But, um, so I, I guess it's also tied my overall journey in technology. So let me just go through that a bit. Um, I've, uh, started then in technology in general with programming and everything else, um, with the rise of the, of the Linux and the free software movement, uh, around there in the late nineties perhaps.

Um, so I, that was the thing that brought me in. Um, it, technology, et cetera. My actual, uh, academic background, uh, is in archeology and then, and then physics.

But I started with archeology. So, um, in, in a way, free software and the open source movement that, um, uh, uh, start. There was my gateway into everything around computers.

Um, I, I started with as a programmer, then I moved into, um, a platform and Unix system administrator. Um, and this, uh, in a way ties with the, uh, with the rise of Kubernetes later on. Um, why, because, well then I, I, I, I, I joined, uh, IBM, where I've worked on a lot of different stuff from IOT to solution design to quantum computing.

Currently, I work at the SaaS Institute as a cloud architecture lead for Europe. But, um, the reason why this is important is because, uh, this background in Linux and Unix in general, uh, actually made, made my, um, going to Kubernetes and the entire cloud native, um, uh, uh, landscape and ecosystem that started just around the announcement of Docker, right? We have that, uh, uh, fundamental, um, that, uh, fundamental aspect.

And then obviously Kubernetes as an orchestrator. Um, it's built on Linux as well. So, um, I started to work more and more with, uh, with, with, with Kubernetes as a, as part of my day job.

Um, and a couple of years ago, um, I've decided to just like I did when I, I, I joined the Dian Project, for example, and others to, uh, join Kubernetes as a contributor to, uh, because, uh, Kubernetes is the second largest open source, uh, project out there. And that's what I did. So, uh, the, the, the, the more direct path to be being be, uh, being the release lead was like two years ago, two and a half years ago when I joined, um, seek Country ex comms and, and other, um, uh, communities inside Kubernetes to just help out and do my, my my part as a volunteer.

Um, when you start doing that with some consistency, you then obviously meet people you trust people. I joined, uh, the 1 25 Kubernetes release team as a shell just to learn how things, uh, were, um, were done. Um, and I was, uh, consistently done, um, a member of the following releases, 1 26, et cetera, until, uh, last release.

I was nominated as a one, the ities 1 32 release lead, much to my surprise. But in the end, it's a result of this, uh, path, the short one that started two years ago, but also a result of a longer, uh, I will almost say life, um, journey around the, uh, software and open source with a cloud native twist. Excellent.

So you mentioned this, uh, there's a new release of Kubernetes. 32, and it's called Penelope. But, you know, we were talking off camera before we started, right?

When I first became aware of Kubernetes, it was actually at a DockerCon, I think, in Austin, Texas. So not even the first DockerCon, maybe the third, second or third DockerCon. And, um, you know, while the news was all about Docker off to the side at one of the booths or whatever it was, I saw them demonstrating Kubernetes.

8. And I remember, and people were so excited, right? And you gotta remember at this point, there wasn't, it hadn't been crowned the king of the orchestrators for cloud native or for containers, right?

You had Docker Swarm, you had rancher, Mesos. There were, there were several different ways that looking at how to orchestrate containers. And, but yeah, the buzz around Kubernetes was, was palpable.

com already, security Boulevard. And I said, you know what? We shouldn't use the name Docker because they have it, uh, trademarked.

Maybe we should look at containers. I said, but everyone there is talking about this thing Kubernetes, but I don't think it's ever gonna catch on. It's too hard.

You know, famous last words. Um, well, you what's wrong, but, uh, That's the Well, that was there, right? Popular Yeah.

But on popularity, yes. So, but, you know, so, but Kubernetes has, over that, I guess it's been almost 10 years, Kubernetes has grown from being the, the orchestrator for containers to really, the heart, the beating heart of, of the whole cloud native stack. And the cloud native stack today is in many ways the compute stack for, for modern architecture, right?

Whether, whether, yeah, absolutely. Whether we're, we're talking about a microservices based architecture, serverless, um, containerized. It is at the, at the heart of it.

And what I find remarkable is here we are, you know, 10 year anniversary of Kubernetes. I think it's eight years, or eight and a half years since version one. 32.

We're still in version one, right? It, it's not where you see other releases, you know, they do a ma two major releases a year with a bunch of miners. 32.

So I think it says something about the, the steadfastness and deliberation in what goes into a Kubernetes release. It's not something, you know, where we whip together a recipe for this month or this quarter. 32, Federico, talk to us about how are these releases planned?

Is it purposely, obviously it's purposely that we don't do these quickly, or, you know, by happenstance they're very well thought out. Talk to us a little bit about that process. Sure.

That's a a very good question. And, uh, just as an intro to, to that, and based on what you've, um, also mentioned, uh, there's a, an interesting timeline here, uh, as well, because, uh, Python was, um, was released in 91. Docker was actually announced at a Picon in 2013.

Uh, and then, uh, in DockerCon 14, which is in, uh, 20, uh, 14, um, Kubernetes was really, was announced. That was the first announced, uh, announcement was at Docker Khan in San Francisco, uh, 2014. Um, so there's this like Python, docker, um, Kubernetes, uh, being presented in each others conferences, which is interesting.

And as you mentioned, obviously, uh, Kubernetes has this past in Google as Borg, et cetera. That's almost pre history by, by, by, by this stage. Um, but you are right in that there were, uh, there were men, other competitors that still exist, but competitors, I wouldn't call it competitors, but other viable solutions, uh, that, uh, fall within the same space.

Um, certainly it's true that in the last decades, Kubernetes has risen as, uh, as the most common choice in terms of container orchestration, but not only it's as a core piece of the, of, of the development platforms everywhere, um, and around the releases themselves. So, um, the first, uh, one L released was somewhere in 2015 actually. 2, which were the ones that, uh, started, um, the now anniversary of Kubernetes.

1. Uh, I think that in the last, uh, in the, in the next year, it was also, uh, one or two releases. But, um, uh, along with, um, so as time m moved, uh, there was an, an, an adoption of framework, um, well documented and streamlined and, um, uh, uh, structured really project, which is, uh, essentially what we use today.

And what is that? So we have three, uh, Kubernetes, um, releases per year, uh, in, in, so one's, one at the start of the, of the year, another one right there in the middle. And like 1 32, 1 at the tail end of the, of the year, uh, which, uh, explains the, if you do the math, obviously from 2016 to now, if you add up more or less three releases per year, we get to the 1 32 number that we, that we have.

Um, so the, the way that the release works, it's actually, it is interesting and, uh, it's, it's the results of a continuous improvement process at each release we do, uh, retro meetings in order to improve the documentation, the process to make things simpler and make things better for the next release team that will pick things up. Um, I will, I will point out, uh, point out that the, the people in the release team are also, uh, volunteers, which are chosen and nominated by their peers, including through a, uh, open shattering program. Um, and how does a release work?

Well, we, it, the release that we have a release lead with some release lead shadows. We have a release team that contains people that will lead specific sub teams, like communications, like release notes, like enhancements, um, et cetera. Uh, and, uh, we have more or less, as I said, three months there.

There's a, a timeline that is published at the beginning of the, of the year, uh, of the, of the release. And, um, there's a call for enhancements. So any six special interest group, Kubernetes has several different, especially INTA groups like, uh, seek notes, seek app Network will, uh, identify the enhancements, the features that they think are ready to, um, be incorporated into the specific release, either as an alpha feature, uh, better feature or as a generally available feature.

Each one of these will come with some specific requirements in terms of the feature documentation, tests, um, et cetera. And, uh, the, the release proceeds with the, uh, inclusion of enhancements, the code freeze, uh, making sure documentation is there, making sure that the communications is ready, making sure that there's no bugs or regressions tackling any issues that might arise. And after that three months period, there's a, obviously during that period, there's, um, several release cuts, alpha one, alpha two, release candidate one, et cetera, until we reach the release day and there's a Kubernetes release that will incorporate all the new features that were identified and chosen by, um, the, the respective six for, uh, that release.

Excellent. That was fantastic. 32 13, of which, you know, based upon the, the kind of background you just gave us 13 of graduated to stable 12 are in better, yeah, exactly.

19 are still in n Alpha. But if you had to pick out what are the big highlights of this release that our audience should really be looking for? Give, give us your top ones.

Yeah, absolutely. It's a better off taste. Um, I would say that there are a couple of interesting ones.

Uh, I think that the way that dynamic resource allocation, which is, um, the, the feature and the, the, the capability that enables Kubernetes to, um, manage things like GPUs, FPGA, so not only memory and ram, but uh, also such devices, which is very important for, uh, many different workloads, including AI related and ML related workloads, obviously. And in this release, there's a, a, a, a new implementation, and there's also the deprecation of the old IMP implementation. And this new implementation actually plays, plays, uh, well with the A cluster autoscaler and, um, overcomes a lot of, uh, some of the limitations that the older implementation had.

So I find this interesting because I think that this startup workloads are continue to rise in importance, and it shows our, our Kubernetes release, um, incorporates, um, improvements that deal directly with, uh, requirements that exists in the, in the, in the out there, right? So this is certainly one, um, I think that, um, there are, um, other interesting aspects as well. I think that we have, um, some good stuff around improving the resource management.

Uh, for example, we have the memory, uh, manager going, uh, ga so generally, uh, available. We also have the, some updates, some a new feature around the ability to determine, uh, resource limitations at the pod level. So creating like a shared, uh, uh, limits that can be, uh, shared by multi pods, uh, multi container pods, which I, I, I think it's, uh, very useful, uh, in, in terms of streamlining and avoiding overcommitting.

So, and a couple of them around this overall, um, improvement team on, on, on resources. Um, there's one thing that isn't specifically a, uh, a, an enhancement, but I think it's fundamental to highlight here. And 1 32 is the first release that has achieved 100% conformance testing coverage.

Uh, and this was something that was done, um, during code freeze, and it was, uh, it was actually the, the PR that did this was merged, uh, live during the Kubernetes contributor summit, uh, during North America, uh, because of the relevance that it has. And why is it important? Uh, performance testing is what makes Kubernetes big Kubernetes, right?

Uh, uh, it's, it's a se series of end-to-end tests that more or less determine the, the answer that a Kubernetes platform should have. Um, we, up until, obviously when this started, uh, the coverage was, um, below 10%, it's been improving, uh, and trying to cover all the, the endpoint through time. And in this release, during that specific, uh, um, moment, there was 100% coverage, which means that, um, uh, the, the, the very way that the, the testing and the performance testing is done, as has reached a level of maturity that, uh, should make us all product Excellent.

Fred Rico, we're, we're over time already, but for people who wanna go get, there's a lot more information than we were able to cover here in 20 minutes. For people who want to get more information, what, where's the best place to go? Kubernetes io slash blog.

Um, they have the, the Kubernetes release article there that covers them. The, the, the, the features, the main statistics around the numbers, uh, the theme, uh, as well. Everything is there.

And we have a series of, um, feature blocks coming out. Some of them have, have already been published that go into a bit more detail into some of the specifics, uh, of the new features. For example, the, the memory management, um, aspect that I mentioned before as a specific blog entry in at Kubernetes io.

Excellent. Credico, you know, people don't realize what really makes the open source world go round and round are people like you, who, you know, you don't do this for the money, obviously it's volunteer work. Some people luckily, you know, do get paid.

We're starting to pay maintainers now. Yeah. Which I think, you know, CNCF does.

And that's a great thing because the amount of time and effort, it's really, you don't do it for the money. As I said, it's a, it's a labor of love, if you will. And, um, indeed people like you who donate so much of their time and effort into this.

Thank you. Thank you for all you do. Thank you.

Thank you for Thank you to the team. Appreciate it. Yeah.

And we are looking forward to great things from Penelope. Thank you for being on Shark tv. It was a pleasure, Fred.

My pleasure. Frederico Munoz, muni's, uh, release sleep. Kubernetes here on Techstrong tv.

We're gonna take a break. We'll be back in a minute.