Cloud Native Computing Foundation KubeCon + CloudNativeCon Europe 2019 (Barcelona), 24 May 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keynote: Kubernetes - Don't Stop Believin' – Bryan Liles, Senior Staff Engineer, VMware

Description

Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io

Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects

Keynote: Kubernetes - Don't Stop Believin' – Bryan Liles, Senior Staff Engineer, VMware

Kubernetes in 2019: Don’t stop believin’. Kubernetes is five years old. Is the journey over? Of course not. What’s next for the Kubernetes ecosystem? In this keynote, Bryan reflects on where we’ve come from and offers ideas on where we could go next.

https://sched.co/MQWR

Join us for KubeCon + CloudNativeCon in San Diego November 18 - 21. Learn more at https://bit.ly/2XTN3ho. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.

A

Good morning so I know some of you were thinking. Brian's gonna sing some journey, that's not gonna happen, but I do want to talk about a journey, and you know us keyboard users. We love the puns, so a little pun for you this morning. So yes, kubernetes in 2019, don't stop believin. So let me talk. Let me let me set this talk. So there's been 31,000 contributors to kubernetes. Oh, my god.

A

That's a lot of contributors. That is a lot of ideas. That is a lot of thoughts. You know that's a lot of time and if you sit in the sig meetings and you see how the sausage is made, you understand wow, that's crazy! If you look at slack, that's crazy, there's been a hundred and sixty four thousand commits. This is a large large project. You know, I can't, and you know, I used to actually I actually had a script that I would go through.

A

This summarize, the commits for kubernetes, so I could actually pay attention because I was busy at work. I! Don't do that anymore. Now, I just look at the release notes, but here's actually a little stat that I found was pretty interesting and all the github issues for all the kubernetes projects. There's 1.2 million comments. You all are a chatty Bunch. What are you talking about?

A

It's very interesting to me, so we were talking about our theme tomorrow this morning and actually this kind of ties into our theme yesterday and it's really is it I'm trying to drive home this thought, and you know it's not a new thought. Actually, it's funny I brought this up. Someone quoted it on Twitter and then someone suggested that Kelsey thought it up. I had independently heard it from Jo beta, but really what it is and and what I'm hearing and what I'm seeing is that a lot of great minds have fought this.

A

This is actually the truth. Kubernetes is a cloud native platform for building platforms. Sounds weird. A lot of people don't think about it. That way, kubernetes does not the end. It's not a product. You know we don't we don't you know unless you're a cloud or or vendor you don't get anything from selling kubernetes. The real power of kubernetes is what it allows you to build, and that's how I want us to think about it. Kubernetes is a cloud native platform for building platforms, we'll go into that just a little bit later.

A

But let's talk about this date, anyone know what March 6 2008 een was it's. When kubernetes graduated it became you know the first CN CF project that graduated. What that meant is that on that date, a year ago, is when kubernetes reached a level of stability where we would say you know what we've shipped we shipped, we shipped we shipped this thing is stable. You should build on it, you should use it, we're that we are actually hitting. We were actually on the trajectory for the peak of the hype cycle.

A

This is good software, so what that meant is that we've hit a level of stability. So are we done, but let me before before I do that this is my mom slide. You know you don't get to put your face on a huge slide every day, so my mom couldn't be here and she's, probably not awake yet, but I still want to. Let her know that her son did it he's on the big stage.

A

Hi mom, alright. So the reason I put this up there is because I have actually have a unique and unique job in the kubernetes ecosystem at VMware. You know, I looked and chatted and work with people to do upstream commits, but really what I focus on one is taking this whole huge project and figuring out how developers can actually use it, how we can write software that is based on kubernetes, so I spend a lot of time.

A

Thinking about what kubernetes is I'm a project that you may have heard of that I spent a large amount of time working on. Is this project called case on it, where we were trying to make it easier to do configuration in kubernetes? This is not a case on it. Talk I just wanted to talk about it up here, but really what I want to talk about is that we've hit we've hit the level of stability in kubernetes.

A

You know, kubernetes point releases are still coming, but you don't hear anyone saying well, you know we have deployments and we have services. You know there. There is a little rumblings of v2 for some things, but we're not hitting the things. We realized that all the all the basic constructs we have. You know they hit 80% of what we're trying to do. It's pretty amazing but I want to think you know. I don't want to stop here. We can't stagnate in this in this realm.

A

You know we can't just say that we're stable and that's good enough. What we really need to do is understand that kubernetes is a journey. It is the vehicle that moves us to some destination and the real neat part is that the destination. Actually it's just it's a concept we will never be to where we really want to go, but we definitely want to have a nice vehicle to get us there. Inclu Bernays in the cloud native. Basically, the whole cloud native world is really what its gonna put us there so I wanted to.

A

Actually this is a kubernetes talk and if you have numbers for some reason, we really like seven. So what I was thinking is there's already been a lot of work in this room and I love. This chart for the reason that it says that a lot of people have put in a lot of work. I have a project on here. Mike workers have projects on here. Many of you have projects on this screen.

A

We have put in so much work in this ecosystem, but there's things that I think that we're still missing and I want to talk about those things for the next few minutes. So I am a developer and I don't consider myself a creative. So what this is is me trying to express some ideas here so kubernetes in more places. This is not a shoe rack. I got this one Twitter when I posted it last week. It is not a shoe rack.

A

It is a server rack and really when we think about when we originally thought about kubernetes when I first heard about it, we didn't think we were thinking about running in the cloud, but we really wanted to run it in our own data centers and that's pretty cool, but because you all can't see my vision. What I decided to do was make a more high fidelity version for this part. So kubernetes in our data centers we figured this out.

A

You know it was difficult because all the server vendors have all sorts of different hardware, and you know it was difficult getting it up there. So storage and networking it's crazy, but then we figured out how to run kubernetes in the cloud and as you all of you probably do I mean vast majority.

A

You probably do you are using kubernetes in the cloud whether it's a managed service or it's a vendor who's, providing it running on a cloud and then also if we're gonna develop kubernetes, don't we want to run it on our desktop and we did work. We had tools like mini cube. We have people who are just running Linux and booting up kubernetes by themselves. We had kubernetes on arm.

A

We have the micro KS project from from canonical, and we have my favorite thing that I will talk about a little bit later kind but running Cooper neighs on our desktop figured that out, but let me actually throw something at you: kubernetes in your data center kubernetes on the cloud kubernetes in your in your on your desktop. How we thought about kubernetes in other places, now I like nice cars, so I put this car up here, but have you ever thought about it? I'm not saying this is a good idea.

A

So if you put this on Twitter that Brian says you should put kubernetes in your car, you know I will say tweet you, but really what I'm saying here is how we thought about it in other places, what what kubernetes looked like in a cloud or in a car, here's another thought. What would kubernetes look like in a store lucky?

A

Let's say you had a huge store and you have lots of computers in stores and what you want to do is you want to actually use the power provided by kubernetes to run your point of sales or back in inventory management, security and things like that? Have we thought about that and even better? Yet what about small IOT devices? What about devices on the edge are we thinking about how kubernetes runs on the edge? Should we change it?

A

Do we need to introduce new constructs, I, don't know, but really what I'm employing you all to do? Is to say, hey where we are works well, but guess what yeah the opportunities are endless here. So, let's think about how kubernetes can run in places that it does not now so, like I said before, I think about developer experience more than a lot of people do and and I have something here and I will say that I love, my cig API machinery, friends.

A

So this is not a dig, but here's one of the problems that we have with kubernetes client go isn't for mortals? Actually, even if you ask the team, you talk about client go and you say that client go is hard to use. What they'll say is that really client go is made so sig EP IB sheet, our API machinery inside of kubernetes works, but that's not good, because guess what there is we want official clients for kubernetes. We want to be able to write software that works with kubernetes api and the issues that we're.

A

Having is that, you know, you have to understand informers and you have to understand caches, and you have to understand things like that, but sometimes I just want to put an object in the cluster or listen for changes. So what we want to do- and this is something this is actually a call to action for you know whether you're working within the six or you're actually thinking about it yourself think about your experiences for your for your developers. We do not.

A

We have to make sure that the barrier to entry doesn't stop people from contributing to the space, and this is another big one. We all know you know, kubernetes is written and go and that's actually a great thing. The the benefits provided by go. It's a simple language. Concurrency is pretty easy. It's we can teach it to. People is great, but guess what there is more JavaScript developers out there. There are more Python developers out there. There are more Java developers out there.

A

There are more c-sharp developers out there and really what we want to do is we we provide. You know this first-class experience for go developers and there are people out there working on clients for other ones, but as I go through, I went through the rust, client and I went through a JavaScript, client and I go back to the go client. All the interfaces are different. We should probably think about that. How can we actually make that better, so now I'm thinking about how kubernetes is a platform? So yes, kubernetes, is a platform.

A

We have this idea of custom resource definitions and really what these allow us to do is actually create new API is for kubernetes. It actually allows us to extend kubernetes functionality without having to change the core itself so with custom resource definitions. We core OS, created this operator pattern which allowed us to merge. You basically make a custom resource definition, which was just a picture of State with a picture with some code that will actually make that code.

A

Make the internal state of the cluster look like the the actual definition itself or the actual resource itself, but I want to actually uh ask the question: is it done? Should we think more? Should we move further because here's something I'm thinking about operators are a great pattern, but the pattern, but really what it is? If you look at it from someone who's not in the space operators feel like a clutch, our crutch is. Are we missing something? What is missing and once again I?

A

Don't have an answer or judgment I'm only saying that we need to think about this, a little bit more. So there are people who are thinking about it. Lots of us have written custom api's for our custom resources for kubernetes my favorite example, and actually here that this conference is up found and their cross plane project. What they're doing is they are actually exposing their exposing resources that are outside of your cluster.

A

Let's say it's a database or anything else and they're using CR D's to basically say that if you can create, if you can create objects in your cluster, you can actually have any resource type. You like I love this pattern. I think we should be thinking about it more when we're doing when we're actually developing our software.

A

Another example of a CR D that I really like is cluster API Jo beta came up here day before yesterday, and he talked about this. So what cluster API really allows you to do is say: hey cluster I'm gonna, tell you how to mutate yourself, I want you to be bigger, I want you to use different machine types. I want you to upgrade your version, and now we can actually tell kubernetes how to upgrade itself. This is an amazing thing. It doesn't matter if you're a cloud provider or end-user.

A

What this does it gives you a single API that actually allows you to actually change your cluster. This is a big deal but, like I said earlier, operators aren't quite there.

A

What I really want to get out is I want people to think about building on or building with with operators and CR DS we're really we're just building on kubernetes we're making kubernetes do things that it couldn't do before and we're giving it the actual code to make it do that, but think about this and here's just an idea, and if someone wants to actually sit down and build this idea, I will be your first customer I want someone to take Postgres the idea that the API, the storage layers and and actually envision what would Postgres look like as a cloud native application.

A

I have basically a Postgres, I, say I want you to install in this cluster I give Postgres my cluster credentials and it installs itself. It figures out redundancy. It figures out how to distribute things and that's way more interesting than operators, because what it does is it's only one piece of software and I: don't know how we're gonna get there. I have some ideas, but think about that. That's pretty crazy, and if we have more stuff software, that is this cloud native API aware.

A

Imagine how much power we can get so before earlier I said kind I want to shout out two men elder. He is literally my favorite developer in this space and the reason I like kind there's two reasons. One is: if you create a kubernetes change before, how do you test it? How do you prove that it works over time? How do you integrate this piece of code? Well, what you could do is you can spit up a cluster Oh? How are we gonna spend up a cluster? Oh I, don't know we can use mini cube.

A

You know it doesn't work here or it might not work here. What kind actually allows it stands for kubernetes and docker and what it allows you to do is say: if you have docker running you can literally spin up a cluster and not just a cluster. You can spin up any version of a cluster that you want and even better. Yet you can create multiple clusters in my favorite piece is that you can create a cluster with multiple notes.

A

This is super powerful, think about on your desktop whenever you're running your software on your desktop I, and you want to simulate something in production being able to run multiple clusters or clusters with multiple nodes allows you to get closer. This thinking, these kinds of projects, not in the core of kubernetes, but on the on the peripherals of kubernetes, are the projects that are going to take us to the next level. Kind is actually I think is, is literally my favorite thing, so actually I want to give a shout out to Ben.

A

I know he's here: I saw him the other day. Shout him out. This is really one of the cool software. We have out right now.

A

So um another thing that I like talking about is declar configuration management, include Burnett, ease, I like talking about it, because whenever I say it, I sound, really smart declarative configuration management and kubernetes, but here's another thing. So one thing that I've seen out here and I want to preface this again. This is not a talk of snark. This is the talk of love and I. Really just want to push this forward.

A

I see you all, as my children and I want to see, you run and ride your bikes and not fall, but um we're we're in configuration management we're still not hitting the bar. There is great projects out there like helm, but here's the problem with helm. The problem is, is that helm is basically our cloud native analog to Debian packages or rpms great in their own realm, but think about this?

A

No one is writing a whole bunch of tooling or or or trying to basically rig rpms to install and and own their own life cycle with other tools. What we did is we built tools like cfengine and ansible and chef and terraform. That's what we need to think about. So we do have stuff out there. You know customize is now in there, but we can think further. There's lots more analogs that we can have in this space. So on the state.

A

I know, sada Lee talked about this I just wanted to put this on the big stage. This is bad advice, don't run stateful workloads in kubernetes.

A

The new advice is, it depends, and the difference is is I'm, saying no I'm changing the know to maybe it is still complex, there's a lot of moving pieces, but we need to understand that we kubernetes, because it is stable, has made these things possible and even though they're not easy, we need to think as a group and as a community how to make this stuff easy, because it definitely can happen so another piece here- and this is actually the most important thing of all- is community.

A

We have to build our community, we have to maintain our community, our community of of corporations, of the foundation itself of our end users of our maintainer z--. We have and power. We have to do that. So one thing we need to do is we always empower the end-user, that's how we default.

A

The second thing we do is we turn those end users into contributors, and then we turn those contributors into maintainer and that's how we do this so I want to leave with just a little summary: don't stop believing I am so happy to be able to do a talk like this because we have come so far and it's crazy I can't even see where we would go next because there's so much surface out there.

A

So here's my summary: if someone wanted to take a picture of that and say what did Brian talk about for 20 minutes, that's what I talked about so last slide and I'm, not gonna sing, but I really want to say that that song go listen to don't stop believin. Just think about this. We really can't stop and we won't stop and now I'm, quoting Diddy, that's awful, but this is where we are, and my time is up. So thank you all and I'm so happy to see you thanks for the invite and thanks for coming.