Flatcar Container Linux Tech talks, 11 Mar 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kinvolk Tech Talks: Continuing the CoreOS Legacy with Flatcar and Lokomotive with Andrew Randall

Description

Andy provided an update on Kinvolk’s two flagship open source projects, Flatcar Linux (its “friendly fork” of CoreOS Container Linux) and Lokomotive, a 100% open source Kubernetes distro, with a heritage derived from CoreOS Tectonic. He covered what’s new in recent releases, where the projects are headed, and how they fit into a broader vision of enabling a truly enterprise-grade open source cloud native platform. We wrapped up with a demo by Iago López Galeiras and Joaquim Rocha, including a sneak preview of some new (work in progress) Lokomotive features.

A

Okay, hopefully mike's picking up, I'm gonna talk now about what we're doing at.

A

Kinfolk continuing the core os legacy, what we're doing with flat car and.

A

Locomotive I did introduce myself briefly at the beginning, where's the.

A

Click: oh ok, great, some response with business development.

A

But I also work a lot with existing customers. I'm responsible customer.

A

Success as well, so I work with customers and kind of get a lot of exposure to the.

A

Challenges that they're facing who has kinfolk, where we have an open source.

A

Driven business model, so everything we do is a hundred percent open source.

A

We're very active in the cloud native community built on a lot of expertise in.

A

Linux and security and work with a lot of big companies out there in the.

A

Industry, as well as a lot of small ones as well, a lot of technology leaders and.

A

One of those companies that you see that we've worked with actually probably the very first project that, as a company we worked on, was with a customer called.

A

Core os back in 2015 and back when core os was founded. They they had a mission.

A

Which was to secure the internet and which sounded like a very, very grand.

A

Ambition at the time, and probably still dozen is still a valid mission. I think.

A

Today and but what they meant by this was all kind of the mechanisms they they.

A

Wanted to get out there to enable a security of the internet.

A

Were quite a few really just looking at all of the attack vectors.

A

How can you minimize them by reducing the attack surface of the operating.

A

System of applications by making file systems immutable where they don't need.

A

To be changed lex, let's lock things down with a principle.

A

Of least privilege so creating policies that say what applications explicitly.

A

Can do and then restrict them from doing anything else. That's not explicitly.

A

Granted and so things like pod security policies, network policies automatically.

A

Updating the you know, the number of servers that are out there with old versions of code on that wouldn't be vulnerable if only an.

A

Admin had been able to go and update it. So how do you make that alt automatic.

A

And addressing challenges around the security of the container runtime in.

A

Effect, that's where we're kin fault first started working with with core rs.

A

Because we worked on the rocket container runtime, which was designed to.

A

Be more secure than the docker runtime at the time, so core os also talked about.

A

Another concept which they call giffy or google's infrastructure for everyone.

A

Else- and this was really kind of building on the side here, if we're.

A

Securing the internet, how do we run it at scale as well? How do we run.

A

Applications at scale- and so they looked at at that time at a lot of concepts.

A

That I think to this group now a pretty much everyday concepts of containers and.

A

Scalability and immutability and these kind of things, but back then, would were.

A

Not known at all outside of the hyper scale companies, so you know so it's a so.

A

This was kind of tied together with that idea of building security is also.

A

Building scalability and robustness, and they actually built really a whole stack.

A

Here to deliver giffy and security together and you'll, I'm sure you'll know.

A

Most of these projects, the foundation of it, was the original. You know the what.

A

The company was originally named after core os. The operating system container.

A

Linux, on top of that, the runtime that I mentioned for containers and then they.

A

Invented that cd built on some academic work that have been out there around.

A

Consensus, algorithms, and how do you distribute state in a in a you know in.

A

An environment where you may have machines going away, and that became at.

A

Cd on top of that, they got into container orchestration initially, with.

A

Their own project fleet and then they rapidly adopted kubernetes when.

A

Came out and built an enterprise version of kubernetes really the first.

A

Enterprise kubernetes product and then on top of that they built a container.

A

Registry and a security scanner to go with that container registry, so quai and.

A

Claire and so quite a lot of technology in there- and it was all going great and.

A

You know really, they were driving the industry in a great way and then this.

A

Big fish red hat came along and saw what they were doing. I thought that's great.

A

We can gobble them up and do some all sorts of great things and of course, then.

A

Another bigger fish came along, so that's that's the little bit of animation. You.

A

Get for the evening um and what happened as a result of that.

A

Was you know some of these projects continued and thrived some, some of them.

A

Merged into red hat technologies, some of them went away, so you know quay and.

A

Claire both active red hat, pud services and projects in effects- it's been great.

A

To see red hat, even open sourced cray, because that'd have been a closed source.

A

Product tectonics been end-of-life, but some of the technologies and concepts.

A

Have merged into open shift and have become more broadly used within the.

A

Industry at cd is, is not going anywhere, because that is absolutely the heart of.

A

Kubernetes and it's an action project within cn cf and in fact, it's being.

A

Contributed to cn cf since since the acquisition. So that's a great thing.

A

Rocket went into cn cf and also kicked off this open containers initiative.

A

Which led to the creation of the container deep project which addresses.

A

Some of those kind of criticisms of the original docker run time and and so.

A

What's happened is you've had a lot of momentum. Around container d we've had.

A

Other container run times like cryo being being created and rockets actually.

A

Been retired but you know not that being an issue because it achieved what it set.

A

Out to do, which was the standardized and that interface and make that a multi.

A

Vendor initiative and then core ice container, linux is still very widely.

A

Used but red hat has merged those technologies into a new os called.

A

Fedora core os confusingly quite a different os, even though they're still.

A

Using the name and the the challenge with that is for people who want.

A

Commercial support, you only get it if you buy into the full open shift stack.

A

And even if you don't, then it's quite a different kind of experience. But.

A

Nonetheless, red hat's recently announced that that's going to.

A

Be end-of-life as of may 26th and then, as of I think september this year, all of.

A

The images of core rs will be pulled as well, so so that's clearly reached the.

A

End of the road I mean, I think, if you look at that full stack of what korus.

A

Achieved- and I mean they had a- they had an office here in berlin, so some of you.

A

May have known the folks. Some of you may even have worked there, but it was a.

A

Pretty small office here they had a smallish office in san francisco couple.

A

Of people in new york- and that was it was a few dozen engineers and it's I.

A

Think you know any of us in the development space looking at what they achieved have to be impressed, and you know this is a quote from one of the.

A

Engineers that they had- and I think this is how they all felt- that was pretty astounding what they built and akin fault. We were pretty honored to have.

A

Been part of that original rocket project and to have really kind of been.

A

An extended part of their team and we've kind of looked at some of what they've.

A

Done and want to take some of these technologies forward in a way that maybe.

A

Is a bit closer to the spirit of what core os was trying to do back when they.

A

Set up and and kind of a little bit more away from kind of where red hat's taking.

A

It- and I mean I love this quote here from the same chap, who said you know, we.

A

Really kind of respects what we've done and says you know he trusts us with the korres legacy, so that quote was kind of the inspiration for this talk title in a.

A

Way because we see ourselves taking that core os legacy forward. So how are we.

A

Doing that, as you probably guessed from the you know, the slider looking at the.

A

Things that are being end-of-life tanned, you know we feel that there's.

A

A lot of life in the in the model that they had for deploying kubernetes.

A

And also in core os container linux in the case of locomote, in the case of.

A

Tectonic is a little bit harder because a lot of tectonic was closed-source.

A

There was an open, open source distro from one of the core os engineers called.

A

Typhoon built off of the open source pieces of tectonic and kind of complete.

A

Into a solution and that's what's become our project locomotive, although we've.

A

Taken it on quite a long way from that flat cars a lot easier because all of.

A

Core rs container linux was open, so we've just forked it and you know we're.

A

Taking it forward and I'll talk a bit more about how we're doing that, but it's it's much more of a linear continuation. So I'll start first with flatcar core ls.

A

Had three channels for flat car thought four core os. I called alpha beta and.

A

Stable and we've added something called edge, which has got a lot of people.

A

Excited because there's a challenge with an os like core os, where a lot of people.

A

Who want the very latest? Cutting-Edge technologies want to use it, but at the.

A

Same time now this is a very mainstream enterprise technology stack and how do.

A

You kind of square that circle. Well, you have your your enterprise release track.

A

A train which is alpha beta, stable and, at the same time, you can throw.

A

Experimental stuff into edge- and so this is, I think, got a lot of people really.

A

Kind of coming back to this model and playing with it.

A

The images are out there for all the major platforms and some of the clouds.

A

Have it integrated already and we're kind of working with the other clouds to.

A

Make it a first-class citizen os that you can really easily.

A

You know schedule new instances with that os and we have. We have an update.

A

Service and an update server that hands out these images as well, which which.

A

Interfaces with the they're built in yeah was core os. Now now flatcar.

A

Auto-Updating engine, so the the host queries, the server. Is there a new.

A

Version available, if there is it, pulls it and updates it. We have also built and.

A

Released an open-source version of what was the proprietary core os core update.

A

Service, and so this is fully out there, anyone can use it.

A

It's called project nebraska. You can find it under confront, github and and.

A

This allows you to do rate limiting of those updates define you know, which.

A

Servers in which groups get which versions, and it also gives you that.

A

Audit trail and monitoring a be able to see which, which updates are happening.

A

And and when they happened or if they failed, what happened and how to rollback.

A

Things like that, so this is pretty cool we'll we don't have time for a demo of.

A

This but check it out it's if you're using core rs it's a great way of.

A

Actually, managing a fleet of machines to migrate to flatcar and talking of that.

A

Migration, it is really simple, because really flat car is just the same as.

A

Caller s container linux right now, it's just a new version and you that you pull.

A

From a different service or just update, the server equals in your config file.

A

And that's pretty much all there is to it. Do you check out the instructions? Is.

A

Like you know, one or two other things that you might want to look at, but it's.

A

Everyone who's tried, it has basically said it's completely seamless and in.

A

Fact there was a debate on this february 12th so just a couple of weeks ago, not.

A

Even that the team at metal put out a blog post, just saying yep, we upgraded to.

A

Flat car, it's you know, it was seamless, it was straightforward, so encourage you.

A

To do that, definitely before may 26th, when all the security and maintenance.

A

Updates stopped on on core os. If you are a chorus user today we even got this is.

A

Stopped press, yes, this is yesterday in the register, so you know you've made it.

A

When you're in the register, and so we've got a nice shout shout out.

A

There, for you know for being the alternative that that, in fact, red hat is.

A

Pointing to as well so so, where are we at right? Now? We we have a dedicated team.

A

As we've actually restructured our engineering team and have a dedicated.

A

Team within that focused on operating systems, security.

A

Led by former aws students program managers sitting right there, so he knows.

A

His way around linux and we've got we've taken over bill's ci we've got, you know.

A

Tests all that stuff working, which is quite a lot of work and edge and alpha.

A

Channels are completely independent now of upstream and beta and stable will be.

A

As well over the next few weeks, so it will really get to the point where we.

A

Can update everything independently and start to kind of really diverge from.

A

Upstream, which you know we have to give them there's only a couple of months.

A

Left of upstream and we have support infrastructure and enterprise customers.

A

With thousands of hosts and that kind of thing, so you know we're set up to.

A

Actually make this a viable ongoing project, you know for the foreseeable.

A

Future talking of the future for flatcar there's a lot of things we want to do.

A

You know call rs had pretty much stagnated over the last couple of years.

A

And we've had that feedback from people saying god well, you know when he when.

A

Are you finally going to get a new version of docker when you're finally going to upgrade the kernel version? All these things we want to do you know.

A

Update the kernel, we want to do a lot more and bpf because that's becoming a.

A

Big thing so bpf with cgroups v2 has a lot of exciting stuff for kubernetes in.

A

It so one of the reasons we want to do this is because it dry will help us.

A

Drive innovation up into the higher layers of the stack as well, but the.

A

Bread and butter of maintaining security updates is, you know, is obviously just as.

A

Important as well for a lot of users and we've also had feedback from some.

A

Customers that they want better debugging tools. You know some portion of.

A

My fleet is crashing, I'm getting kernel crashes, you know how do I diagnose that.

A

The chorus didn't have great tools for doing that. There are some things that we can. We can do and unbundle in you know, without kind of compromising the the.

A

Minimal distribution ethos cuz were absolutely need to keep that, but while.

A

Keeping with that philosophy add better visibility and debugging int into the.

A

Distro so lots of things we're already thinking of and we've had input from.

A

Customers, if you've got things you want to see we'd love to hear this is all.

A

Done in public, throw it up into the github and, as my first bullet point here,.

A

It says we actually want to be really transparent about this and we're going.

A

To be publishing a road map and sharing what we plan to work on and so you'll be.

A

Able to see in the open where things are going, so we really want this to be a you.

A

Know involve in the whole community, so that's that's it for the linux layer. Now.

A

I'm gonna move up to the kubernetes layer locomotive. Oh, I didn't even.

A

Explain who does anyone know where the name flat come flat? Car comes from.

A

Because, okay, we got a couple, but so it's yeah a couple of train enthusiasts. It's.

A

It's the what it's the wagon on a on a train on which you put containers. So if.

A

You want to take containers off a ship right. Everyone's everyone sees all the.

A

Ship metaphors, but the whole point of containers is their intermodal. When you.

A

Put them on a train, they go on a flat car. So it's it's the thing. The top of.

A

Which you put your containers so when you know that it's an awesome name only.

A

All the world view that so we're going to go from the flat car to the.

A

Locomotive how we drive in kubernetes forward, so so, what are we doing with.

A

Locomotive, first of all, it we're not making any kind of custom version of.

A

Kubernetes we're not doing an open core thing, because this is kinfolk. We don't.

A

Do open core it's all open source with we're scouting around what are the best.

A

Of breed open source technologies that you have to put together to really.

A

Create a viable kubernetes deployment and putting them together and enabling.

A

People to keep keep up with the upstream releases that are coming out.

A

Rapidly kind of integrating them testing when putting together trying to make it.

A

Consistent, so one of the things that you find is when you actually go and deploy.

A

Kubernetes and you put in all the different pieces that you need like.

A

Storage and networking- and you know and monitoring, and all these other.

A

Components that you add in it starts to get very messy, because there's a lot of.

A

Different things you need to think about: well, let's try and make it as consistent.

A

As possible as to how you manage all of those different things, it's fully self.

A

Hosted same as techtronic was, which is this thing that core os really.

A

Introduced- and we kind of take that further, I think than it has been before.

A

And sensible secure default config, so we really believe in trying to beit make.

A

This secure out-of-the-box, you know kubernetes out-of-the-box, is the.

A

Opposite of secure, so so, let's try and actually kind of make it such that.

A

People don't shoot themselves in the foot quite as much with regards to security and then and then the last thing is you know we want this to be a.

A

Platform for innovation, not just like pulling the the you know, the old.

A

Projects from three four years ago that are all proven but actually make this.

A

Something that we and the community can try out and deploy new things with so.

A

And- and this may be a whole range of things we first of all, though the kind.

A

Of areas that we really want to innovate around security and monitoring, we have a.

A

Project called inspector gadget, which some of you may know which allows you to.

A

Apply various bpf tools into the cluster. A lot of those are around security and.

A

Monitoring- and you know- we've- we want to do a whole lot more in that area as well.

A

One point I'd like to make here in terms of the philosophy of how we think about.

A

This and I've seen this with with some distros that it's basically the distro.

A

Is basically an installer, and maybe it bundles in networking and a couple of.

A

Other things, but for other other, really key infrastructure components. They just.

A

Say, oh there's, some upstream, the amyl from the project sure we're compatible.

A

With deploy it just like you would your app and effectively what they're saying is you know the the kubernetes layer is the.

A

You know is the boundary between infrastructure and applications, and you should run your infrastructure components in just the same way. You run.

A

Your app, you know, five times a day, updates or however, frequently as soon as.

A

It comes out upstream there's a new version, just pull that in, and I think in my view, this is actually broken, because I think the the.

A

The right layer is that is to is to think about what is the the.

A

Infrastructure on which your developers and deploying applications so that that.

A

Is kind of the it's the organizational boundary between the operations team and.

A

The application developers and it should also be a boundary in terms of how you.

A

Think about the components that go into into your your overall distro, so we.

A

Think of locomotive as the full distro you know the containers plus the.

A

Kubernetes kernels, the containers is like the infrastructure containers or.

A

The components are like the the bundled user space tools that you get with a.

A

Linux distro. That's that's my opinionated view anyway, that what the.

A

Stack looks like is you know we actually separate out this idea of the compose.

A

Components that you deploy from the underlying kubernetes engine, so the goal.

A

Is you can run on top of any managed service and obviously we have flat car.

A

So if you're, using our installer with flat car and kubernetes, it's going to.

A

Put those together the self-hosted, so we use blue cube thor for deployment and.

A

This is how we've kind of gone beyond what anyone else has done up to now, we even kind of go through this phase of moving cubelet from running standalone.

A

On the host to running inside a pod- and I think cora has tried to do that and.

A

Never actually got that published, but we've got that working which kind of.

A

Blows my mind to think of that and then once you do, that in place updates. If.

A

Your infrastructure components plus the control plane. You know just pretty.

A

Simple and we'll do some demos for you in a minute to ship to show that.

A

The the the common component framework here is the idea. Is we that whatever.

A

You're deploying whether that's a you know an open, ebs, plug-in or prometheus.

A

Or contour or something yeah, the the way you configure that feels exactly the.

A

Same it's not like you have to go and learn a new tool. Each time you add in.

A

One of these components and as we really simplify the the configuration and we're.

A

Trying to write on all the different platforms, the the architecture this this.

A

Is kind of the the visions where we want to get to so that so at the moment, with.

A

We have this local ctl tool and that actually bump for a version of the tool.

A

That bundles all of the upstream component binaries we're going to have.

A

That as an option for air-gapped deployments, we're also going to have the.

A

Ability to to download there's separate components based on a manifest, but.

A

There's a there's, a top-level config file, which defines defines your cluster.

A

Info defines the version, and then you can either bring component configs into.

A

That, or have them in separate files, whichever works best for how you're.

A

Managing it, those are define, those are defined using the hash, leak or hcl know.

A

Which is saying where you program tell us terraform a declarative model, pretty.

A

Simple, I kind of I think it's a little bit easier than the amal, but you know.

A

It's it's similar and then local ctl takes takes that declarative, config and.

A

Applies it to the cluster, if it's bit, if it requires it to build a cluster it'll.

A

Do that for your terraform and boot cube if it requires it to deploy things into.

A

The cluster it'll create helm, charts and apply those using helm and, in fact, even.

A

The control plane itself is deployed using helm, so that's kind of a new thing.

A

As well, I think and then and then obviously you can talk to the cube api.

A

In for some things as well, so so that's kind of the the top level architecture.

A

The components that we've built in a being driven a lot by work that we've.

A

Done with customers so far, and what we're hearing that this list will grow.

A

Being see, there's a whole load of things that we already have out of them box from authentication to storage, including his sf and open ebs, metrics.

A

Backup auto-scaling networking we've got a nice site, a site.

A

Vpn, which is which can hook into the aws vpn service, so you could have say.

A

Something running on bare metal or in packet, which sets up a vpn using the aws.

A

Is built in vpn service flatcar linux, update operator is, is a neat component.

A

Because it allows you to coordinate those automated updates of the.

A

Underlying os, so if you think about it, if you have a container os, that's doing.

A

Automatic updates that could really mess with your kubernetes layer if it's not.

A

Done in a managed way, so it allows you to actually coordinate the underlying os.

A

Updates with the with kubernetes, so it actually drains the node before a bit.

A

Before it reboots the box with a with an os update and there's a lot more things.

A

We want to work on as well and add in so overall what we're trying to get to here.

A

Is a user experience? That's got as much of the simplicity of using a gke type.

A

Service out of the box, but that you control and you manage and it's open.

A

Source and you can play with it, however, you want.

A

You know command line for managing clusters and components in the same way.

A

So the way you would update components is the same way as you update your.

A

Control plane with a much simpler config than you've ever had to do up to now.

A

With the self self install kubernetes and also a dashboard, so monitoring.

A

Control and you might say yet another dashboard.

A

Yes, we've built yet another dashboard and you might say, are you completely.

A

Loco and we think not because we did, we did go out there and looked at all of.

A

The solutions that were there- and you know we have a number of things we.

A

Wanted to see- and we you know- we'd have hack we'd- have preferred to have picked.

A

Something off the shelf: if it was there and did what we wanted, we wanted it be.

A

Open source, obviously, and that that kind of goes without saying we.

A

Want it to be an active project, we want it to be independent of a distro, so.

A

There are some where it's very tied together with that specific distro, so.

A

Still open source, but it's very hard to kind of tangle it apart. We wanted it to.

A

Be on a modern, up-to-date framework, you know the ui world front-end world moves.

A

So fast and keeping up with the latest security and latest performance and all.

A

These capabilities is, it was pretty key to us. We were also looking for it to be.

A

Modular and extensible, we have things we wanted to plug into it and obviously a.

A

Nice, clean, modern ui, and we wanted it not just for monitoring. So some of the.

A

Dashboards out there allow you to monitor, but don't allow you to kind of.

A

Perform operations as well and, and so looking at all of these, we said, there's.

A

There's nothing out there that does this, so we're actually in the process of.

A

Building one that does and you're the first group of people, we've ever said.

A

That to in public we haven't this, isn't this, so this is.

A

Still a little bit stealth, it's still a bit behind the scenes. We will be. You.

A

Know we've we've got some work to do to get at the point where we feel comfortable releasing it, but you know you'll be the first to hear about it. So.

A

Where we're at today with locomotives, so all of all of the cluster bring up tear.

A

Down a bunch of these components or implemented sported on packet, an aws and.

A

Bare metal- and you know the fact they're there is proven, because we have actually.

A

Got locomotive deployed with a number of customers in some of its kind of early.

A

Work-In-Progress versions, we don't yet really feel comfortable fully going.

A

Calling it kind of a v 1.0, but we get we're getting close to that point and.

A

There's a number of things we're going to do over the next few weeks. You know.

A

Getting the versioning update upgrade process, our customizations support.

A

Components, dashboard, obviously getting bad integrated and then just like doing.

A

All of the work that it's going to take to get what we have internally and doing.

A

Kind of one-off things for customers into an open-source project with.

A

Documentation with you know, governance processes and things.

A

Of that that we can we can put out, but you will see that happening. I mean.

A

Literally, over the next next few weeks, the teams working really really hard to.

A

Just get our first stake in the ground of the first public, open source, repo.

A

That everyone can get their hands on so so that's. That summary, is how we think.

A

We're continuing the chorus legacy, we think spirit lives on in us. We're.

A

Working very hard to make flat car container linux the platform that.

A

Everyone wants to deploy kubernetes on to core os was super popular. We think we.

A

Can kind of rekindle that and take it forward.

A

And we're pretty excited about locomotive. It's you know it's. It's our.

A

First attempt to to to really kind of push forward the envelope of kubernetes.

A

And and it's a community project, it's open source. We would love for people to.

A

Get involved as well so with that I'm actually going to hand over for the demo.

A

To so you can meet a couple of our engineering directors, we are going and.

A

B

Hi, so I'm yago, I'm director of the engineering in the cloud division, a team.

B

Fault and have a cheat sheet, so I don't forget anything about my demo. So, let's see okay, can you see that all the way to the back.

B

Yeah good cool, so yeah. I wanted to show a bit.

B

How simple it is to install a locomotive, cluster and so yeah what you need to do.

B

Is just run local cto and in this case this version- oh one. This is important.

B

Because we'll later we'll be updating cluster installed, but of course this.

B

Takes a while, so I don't think we have much time so what I did was I pre.

B

Install the cluster, so we we save some time, but of course you need some.

B

Configuration here you cannot just do plus or install, and then your cluster.

B

Your your dream cluster will appear, so we have some config file and I'm gonna.

B

Show you that to you now so yeah this cluster is running on packet, so you can.

B

See here that we are running a packet cluster, so this is the hcl language that.

B

We use to configure a cluster as any set and yeah you set up some the usual stuff.

B

Like that, your cluster name, the number of controller nodes- and we call masters.

B

Controllers so yeah you can have one like in this demo because yeah I didn't.

B

Want to do an h, a cluster for a demo, or you can have three or more for your hd.

B

Needs so in production, you should really do that. Yeah, you configure our dns so.

B

Right now, we support two two ways: you can use route 53 or you can use external.

B

Dns and bring your own dns yeah in this. In the demo I chose raw 283, yeah and.

B

Then other stuff, like the facility and also the os version flat pair, of course,.

B

And in this case I chose the edge channel because you know I want to do.

B

Demos on something that is probably breaking so yeah, I'm like that yeah and.

B

Then all that stuff, like the public keys for your servers and some.

B

Network stuff yeah, apart from new york controllers, you also want workers, so we.

B

Have this concept of worker pools, so you can have different worker pools like if.

B

You have a pool for storage or a pool, that's you know, has a lot of cpu.

B

Capacity, so you can choose different kinds of distances for your for your.

B

Worker pools, in this case I have two workers and again I'm using edge yeah. I.

B

Said why before so yeah, this is a cluster, and then there will be the.

B

Components that annie was mentioned before, and so we want to show you how.

B

Easy, it is to set up some components, so the configuration is here, for example,.

B

For a certain managers, we have to configure the email that it's used for.

B

The certificates and the namespace so yeah, that's pretty pretty easy for.

B

Contour, which we use for ingress yeah, you don't have to specify anything. So it.

B

Just works like this: we use metal lb in packet to get a load balancing because.

B

Packet doesn't provide a load balancing service like aws or any other cloud.

B

Providers because it's a it's a bare metal cloud provider, so we use metal.

B

Beam- and here we just have to specify what's the public ip, that you will use.

B

For your load, balancer yeah another metric server, which doesn't need any.

B

Other configuration so yeah, that's that yeah, so the cluster is running. I can.

B

Show you cube ctl, I don't know, guess what such a to see that something is.

B

Running yeah, we have a lot of a lot of pots here, yeah, but as much as I like to.

B

Come online, I also like to see the clusters. You know bore you I so I want.

B

To show you the yeah, I think first public dashboard locomotive, dashboard.

B

Thing and here it is boom. Well, we have some round things and then something.

B

Some clickable things: that's awesome, yeah. So here you can see what you would.

B

Expect from a dashboard, like your overview, cpu memory usage, how much? How.

B

Many parts are running successfully: 100% yeah god yeah at some events that.

B

Happened in the meantime, you can see your workloads. Yes, a full-size replica.

B

Sets you know your deployments, your pots. Of course you can also filter by name.

B

Space, so we don't have anything running in the default name space, but I don't.

B

Know I guess in the decks namespace we have running web decks running so yeah.

B

Things like that yeah, I think your game will mention it.

B

Later but this is a very early, you know we're showing this in a very early stage.

B

So there's still some rough edges to polish yeah, so what I'm gonna do next is.

B

I'm gonna install a new component on the communities cluster. It's the flat car.

B

Linux update operator that andy mentioned before so, let's see that.

B

That's not running right now it is not so there's not nothing about flat car.

B

Here in pots, so yeah, let's do that so in local ctl, we're still gonna use.

B

This one version you can do a component list and this is already available.

B

Components that any showed in the previous slide so yeah, let's, let's just.

B

Install the flaca update operator- let's add it to the class reconfig so we'll.

B

Add a new component, but her linux update operator and it doesn't require any.

B

Configuration so we just leave it like that, and now we'll do component install.

B

And so it doesn't go through all the components to update them. Let's say.

B

Directly operator yeah, so it's now installing a component and hopefully it.

B

Will succeed yeah it successfully installed yeah.

B

Wrong tab, so let's refresh and yep now it's running so we.

B

Have free agent pots that run on each node of the cluster and then the.

B

Operator itself that you know manages all this yeah so yeah. What's next yeah.

B

Next thing I want to show you is to bring your attention to the version of.

B

The cluster, so it's a v1, 17.1 and yeah. If you go to the notes, you can see that.

B

All the notes are running that version yeah one 17.1, it's 1901 117 number one.

B

And this is actually a nice opportunity to show you how the cubelet is.

B

Self-Hosted, so if you go to work clothes, you can see here that, if I stretch for.

B

Cubelet, it's a demon set so a normal cuban area subject and in the pots you.

B

Will see it running there too yeah and also I'm gonna, do an update of.

B

A locomotive version- and this will update both the kubernetes version and.

B

One component, I chose one at random, it's dex in this case. So if we search.

B

For dex here you'll see that it's running version 2 2000. It will be.

B

Updated in a minute, so I will start the update process now and since this takes.

B

Some time I will hand it off to joe keim's.

B

Which, who will who would talk to you a bit more about the design of the.

B

Dashboard, you do a couple of more demos, so, let's start this update instead of.

B

Local ctl or one, let's do a google ctl, oh and I'll. Do a cluster install.

B

Because our install is that important, so you can get just install again and if it.

B

Doesn't update it will it update, will probably change the name later because.

B

Install doesn't really match this, but yeah. As I said, this is a really early.

B

Preview, so let's just do that and you see it's using terraform behind the.

B

Scenes to actually do the update and then it will use home to actually update.

B

The the control plane and everything so, let's see when tara from plan shows you.

B

The plan we'll just say yes without because we're living on the edge.

B

Yeah, let's do that and and introduce you kim.

C

And I'll press enter cool.

D

Hi so see ya so, and you already explained to you why we.

D

Chose to create yet another dashboard yeah and that's what I'm here to talk.

D

About but first I have to give you a disclaimer that this is like highly.

D

Unstable code written in javascript and running in a cluster that is not high.

D

Availability, that is updating in the back on the background and doing all.

D

That, while at a resolution that we didn't design this dashboard forth, so if.

D

It goes wrong, just you know, don't judge me yeah, so one of the things that that.

D

Is very important, I think, in the in the reason why we chose to design a new.

D

Dashboard or create a new dashboard is that we wanted this to be like, of course,.

D

The the dashboard for lokomotiv, but also generic one and sorry the way that we.

D

You know, I think our managing to do this is by by allowing front-end plugins so.

D

That means that if you want to use the dashboard for your own kubernetes.

D

Distribution or something like that and you don't want to use the same plugins.

D

As we do, you just have to not include them, and and that's it so we have a bit.

D

Of isolation in there so yeah one of the plugins that we have it's it's for a.

D

Project, it's for part of a project that I think animation called the specter.

D

Gadget and if I go here and things are working seem to be working, you see that.

D

There is a an icon there, so this is coming from this plugin. That gives you.

D

Traces or trace loops, so the concept of a trace loop here is like a straight so.

D

It's like you know, lisa system calls, but with minimal overhead.

D

Running in the kernel and yeah, we have this installed in locomotive in this.

D

Case and when you click here you should see you know, let me stop traces, but.

D

Let's say that you are running a demo and they don't work and or maybe they're.

D

Even the plot died right, so if the pot died, you cannot come to this view.

D

Because you cannot access something that does not exist, so we have another view.

D

Here called traces and in the traces, as you can see, on the status here, this.

D

Lists all the pots or actually all the traces for all the pots that we had so.

D

If I go here to any one, and it will of course work now- you can see that you.

D

Know these are the traces so again the system calls. So the idea here is that if.

D

You have a pot that suddenly died and you don't know why you can go to this.

D

View and you can go and say: okay, you know, let me see what happened. You see.

D

The system calls things fill and, and you can of course, go here download the log.

D

Send it to whoever is responsible for fixing it and yeah. So that's yeah. That's.

D

Pretty much, I guess, one of the most important aspects of of the design we.

D

Chose to do this is only one plugin which we we will have many more so like.

D

I said you know, if you don't have the respective gadget plug-in, then this.

D

Whole menu and view, and in the route here will not be available and yeah and.

D

So so we hope that this makes it very flexible for other community communities.

D

To join in, let me see what else I could say.

D

Maybe I'll just hand it over back to so yah go because I see that it's already.

D

Dated so yeah, that's it! Thank you! Yeah thanks for kim so yeah. We can see.

B

That this is updated, so it's one. It's now one 17.3. This is the control plane.

B

Version but of course the cubelet is also being updated. So, let's see how.

B

That's going yeah, so we have this cubelet here. It's.

B

Still 117, not one, that's fine, because it does a rolling update over all the.

B

Couplets, so maybe this one is updated. Yeah one 17.3. This one is not updated.

B

Yet but yeah it will be in anytime soon, so so yeah. This shows a bit how the.

B

Update works and yeah. It mostly worked even with only one api server, so yeah.

B

That's what I wanted to show with the update? Well, let's see, what's what else.

B

Oh yeah, I wanted to show that how also text got updated. So if we go back to.

B

Work loads and we filter with the x there. It is, and it's not running version.

B

221 so yeah a huge improvement in minor version yeah anyway yeah I can check if.

B

The notes are updated, yeah, okay, the controller is updated already. This is.

B

Updated so I think everything should be updated.

B

Not yet but yeah it will, it will be updated eventually. So I think that's it.

B

For the demo, and now we wanted to show you one last slide before we let you eat.

B

More or go- and this is yeah the slide about the cloud native reject, so this is.

B

A conference that we organized and it it's happening always right before.

B

Coupon and it's for all the talks that get rejected, so we would like to take.

B

Those talks because they're very good talks and the rejection rate for coupon.

B

Is pretty high and so yeah we do this right before I become next one in.

B

Amsterdam in march 28- that's yet not not far in.

B

The future so yeah, please check it out and and join us. If you can it's a it's.

B

Really great, it's a really great event: it's a smaller and it has a very nice.

B

Vibe, so you can actually talk to people and not just be there lost in the crowds.

B

So yeah thanks everybody and yeah. That's all. I have.

B

Yeah thing: if you have any questions, I can try to answer or defer. Yeah.

B

Yeah so, as andy said, the versions are currently.

B

Hard-Coded on the binary. So if you download a new version of the binary.

B

Will have all the versions hard-coded there, but our plans is to move to a.

B

Manifest way so with one version of the tool you can, you know upgrade to a new.

B

One because yeah you don't want to download everything every time you want.

C

Update yep, so we support a pixie boot and we support bare metal installation.

B

Through I pixie, so you would you just have to you know.

B

Point your servers to the well. We we actually run matchbox also buy core os.

B

So you can actually group just your servers there and then they will get the.

B

A pixie payload and they will install flatcar and then they will install.

B

Locomotive on top and it will be running it's a service that allows you to.

D

Basically inject ignition config, which is blocked or digital server, yeah thanks.

B

Okay, that's not more questions we can wrap up here. Thank you.