.Net Foundation .Net Community Standups, 25 Feb 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: ASP.NET Community Standup - Feb 25, 2020 - Intro to Kubernetes for ASP.NET Core devs

Description

Join members from the ASP.NET teams for our community standup covering great community contributions for ASP.NET, ASP.NET Core, and more.

Community Links: https://www.theurlist.com/aspnet-community-standup-2020-02-25

A

B

Didn't have my opponents today, I just know, we don't that's. Why.

A

I think it's awesome, though, so we just assuming it's.

B

Working I guess on the stream.

A

If it's not working.

B

A

B

A

Good I'm back I apologize.

B

A

Week, Oh schedule.

B

So I the plan, apparently.

A

I didn't know this until I was meant to come back, but I had agreed to come back like three or four days after getting my wisdom, teeth, yeah.

B

A

I found that out like the night before that I had apparently agreed to that.

B

And I was in no state to come back to.

A

Work so I did not come back to work and I'm, always hustling for speakers and I'm, like oh shoot, I got an email yeah, so anyhow yeah. This is great, so this is feel better excellent, good good. This was definitely like. We put the thing at the beginning at the end of the show where it's like, where you can request something.

A

We had a lot of people and during the cloud native show, people were like I, don't I, don't know, kubernetes teach me the kubernetes okay and then what I also made it come up was I saw your Andy, see, show and I was like. Yes, we need this okay, so okay.

B

Great I wonder if I watched your NBC show also and then tweeted about everybody.

A

What I think I actually just deep link to everyone to.

B

Like 57 min.

A

I saw the performance.

B

Interesting yeah.

A

B

Already seen that information I had those opposed to my secretaire publicly, you.

A

Want to watch me talk.

B

To people about it and I know that you block out all the framework names and stuff which is like super lame. By the way you should just like totally.

A

The worst you're the worst, but it was very.

B

Cool yeah well.

A

I in case you're wondering Damien. Yes, did you have any questions for me? Yes, I still love the community.

B

A

Yes, thank you see: okay, okay,.

B

A

The thing and then hopefully my Mac will continue there. Will you okay? So, yes, this is the yes, we are sharing these on the you were list and these the links also include them at the end, don't have a huge amount today, but some pretty good ones. I think john hilton riding on from webforms to blazer. So definitely this is really a refactoring of a web forms application. There's not really much blazer specific in here. This is more setting you up for success right.

B

A

Here he's taking a real-world thing, looking at a search form, written kind of web forms e and he kind of breaks this out and says you know what here he goes and starts, refactoring it the main step sedated here that I like this kind of approach. Well, first of all, going back up to the top.

A

Is he says a lot of time, people look at a rewrite and they'll say: okay, what are we gonna do and then they start doing this big rewrite and then they're halfway through it and then they're maintaining two applications- and you know so. I've been on some of these in the past, where the rewrite never actually shipped right, and so that's a bad place to be so. What he talks about doing is doing a refactor as a bunch of small refactoring steps.

A

So he takes a web forms application and he uses mediator and he just simplifies it so simplifies the structure and then converts it over to using service calls with mediator. Okay, when he's done with that, I'm gonna scroll way to the bottom, he ends up moving from this to actually a lot.

B

A

But this and what's nice here, is he separated it out and he still has a web forms front-end, but there's not it's not really doing much. It's just creating a query using mediator, calling to a back-end. What's nice with that is pretty easy to move that to blazer. So.

B

A

The presentation.

B

Exactly exactly which.

A

Is something we hear over and over again yeah, but we don't often have to confront our mistakes right right and it's, and definitely that was a thing with web forms. Was it made it pretty easy to put a lot of code in, and that is something I do think we need to keep reminding people with blazer. It's a little bit easy to Zizi put too much into the component itself right so anyway, like that Christian wire, so I clump them together, Christian wire, also in our blazer segment.

A

Here he's talking about running asp, net cord blazer webassembly applications as a PWA, so he talks about setting it up. You know which kind of which things you're gonna be putting in your manifest. He actually uses work box, which is from google it's in the Google developer area and it kind of it's an NPM tool with CLI, and it kind of can put all your stuff together and also work build your serviceworker.

A

So then one nice thing out of that is you know, is audit for lighthouse. Now it's really good and- and you can install it as an application- that's interesting yeah! So we are doing, we are doing PWA support in the next blazer preview rubies. So.

B

A

But I will full full is a tough question. Yes, there are actually lots of different ways to do. Pw A's, like there's lots of opinions about what yeah right thing used to do like you immediately went to like we're using this thing as a Service Worker, so it'd be interesting to compare and contrast what's in there with what we're doing in the Box we're.

B

Probably going to choose.

A

I didn't do the work. Steve did the work, but we probably chose something: that's pretty pretty free of opinions or pretty pretty minimal. There's a lot of options in the space. I can't say: I understand them all. Do you think we should get Steve on soon to show off? That's.

B

Okay, there you go me.

A

Too, okay cool one more here or well. We actually have two more on the blazer front. This one is using tailwind CSS with blazer the main reason I'm bringing this in like there is some stuff about how this was set up, but I honestly haven't looked at talen before, and the idea here is. This is a great comparison. This is using bootstrap and you'll notice with bootstrap. We have some classes that are pretty opinionated and and tailwind is more of a toolkit. So.

B

A

I mean it's almost kind of like you're, just setting some CSS class yeah.

B

You know that looks a little bit.

A

Like an inline style but more abstraction to it, yeah yep, so then after this was the main value honestly that I got was like wow. This is neat. He does walk through this is you know when, if you decide you're using tailwind how to do that so he's using with NPM and gulp, so he chose how to set that up, which you're gonna add for your package. Setting up.

A

This is the actual styles that he's using and then he sets up a gulp task and ren-san so pretty mean all right, one more in the blazer world. So this is unka and writing about using OCR with 4 blazer, and here he's using. This is a jur service, and so you know here's a nice animated gif. He shows up loading it and he's actually OCR uploaded images. So this is pretty cool calling back into the azure computer vision. Cognitive service. Is this.

B

Is this server.

A

Side, or is this point, this ones clients? Ok, actually this is.

B

Server side. Sorry, the other one.

A

I had was was client side so yeah. This is server side, so yeah I'm, just making that that rest call over and so cool and at the very end of it he shows a bunch of things showing you know, uploading and you know OCR and then handling in responses, and so some neat cool all right. This is from Andrew Luck here he's writing about if your asp net core, if you can't find a controller and basically debugging application part- I saw this this morning and I was really happy. Oh.

B

Good, do you want it? Do you want to talk about it, I'll? Let you talk about it. Ok! Well,.

A

So what he does talk about here is this application parts yep and that witnessess added. Actually this was added application. Parts I think are a 100 features biochemical.

B

A

Part of its been a core ok, so here he talks about how application parts can be both inside your app or you know like with Razer class libraries can be external and then you've got. They can include. You know, controllers tag, helpers, pages, views and obviously, in some cases, if it's a your pager would or whatever you may not have abuse or may not have a certain thing. So he talks about how their how they're built up and then in his case he wants to actually see where's my controller right.

A

So he goes through and he says there actually is a doc, but he doesn't like it. The doc has a controller that you know lists him out and he says I think that should be a logger instead, so he creates a custom logger and the logger basically iterates through the application parts yeah and writes him out. So then he found out here.

A

Here's his weather forecast controller and it's not found yeah the big, the big thing that we see with application parts that people get people report a bug with routing and say routing must have a bug that to me of like three or four years of working on asp net core is like no, it's never routing here.

A

Your your controller is in a different project that are not including so yeah MVC will look at everything bit severe build time, and so, when you have dependencies that aren't there at Build time, there's some manual job stuff to do this.

A

The whole thing with routing and it's I mean I, remember back to like the asp net MVC route, debugger that, like yeah fil hacker way back in the day, and all that like debugging routing and the the sometimes, if you have more complex routing or if you have some routing, is in your in your you know, it's declarative and sums and your attributes, and all that is its. It can be a little tricky too. So that's actually nice thing with razor pages right. It is it's well and.

B

You you specify, we.

A

Actually have a decent there's a couple tools for that, so Rick Rick, one of our writers, was working on a Rick, a.

B

Duesenberg Anderson.

A

Was working on a page that would show just these are all the endpoints that are defined in application and everything I know about them.

B

What attributes do.

A

They have what's the route template. Yeah I've got a visualization that you can find in my presentations, repo and my name on github yeah, that I've shown in some talks before that is like it will show you the actual tree structure that routing uses yep. So we have code in the product, that's supported and undocumented, but you can use if you want that.

A

Will output a graph like basically a directed graph in a particular I'm, trying to remember the name of the format, but it will output a standard format in the graph and then you can visualize it. However, you want so how do I find out more about that? I have got undocumented unsupported I mean.

B

A

Reason why undocumented unsupported is what, if the format like the format, could change or something like that. So.

B

A

Well, I'll switch over when I finish my hand, so I'll switch over, and you can include that in there that sounds. That sounds pretty cool all right. The last one I have to share, so she had talking about handling errors in asp net core 3-1, so all the all the handling stuff. You know different caching exceptions he's been building out this net learner application, so he shows that and then different error handling techniques with MVC razor pages, Blaser, etc. So, as always, he's writes, he's very insight, encyclopedic posts, so this is just you know.

A

This is worth grab a cup of coffee. You sit down and read through it. So alright, the last one I have is just more informational shop on web, is updated through dotnet core 3-1. So this is the its. This is our like ebook style, and it's there's.

A

This is a full ebook with an application pool application, and so, if you haven't looked at this before I know, we've called this out a few times, but this talks about different things like client-side web technologies and compares your different options and all that, so it's all updated through dotnet, core 31 and again I will be sharing those links out in the chat and now, let's pass over yeah go ahead. I'm gonna, I'm gonna deviate from my prepared remarks. Awesome as I've been prompted. Okay.

B

It looks like everybody.

A

Can see this this loaf I kind of graph here this is this- is basically that visualization that I was talking about. So let me jump to. Let me jump to the code and I can explain what's going on if I, which, if I, have it up sorry for the motion, sickness everyone, let me bring it up envious code and show you all so I've got this app here and this all the magic is happening in startup, so the undocumented unsupported format that I was talking about. There's a DI service that you can get.

A

That's called the DI, the DFA graph writer, which is a lot of Ryan's sort of jargon.

B

A

You can use that as a text writer. Unfortunately, it's text writer and we made a change in asp net core 31 to ban synchronous I/o by default. So if you.

B

Wish you're welcome if you wish to liberate.

A

Your app's route graph, then you'll, have to re-enable synchronous, I/o and right into the response stream. So for anyone who's not in the know, the endpoint data source is where you can find all the routing information, and then this thing will write that graph and then the way that this is actually being visualized is I have got a graph. Is that's the name of the.

B

Foe right so I'm using this.

A

Great package here that somebody wrote called vis, Jas mmm-hmm and that's that's what I'm using to display this and I'm just doing some simple JavaScript here to display it so what's going on in this application, is I've got a home controller with some conventional routes, sort of stuff going on here and I've got a weather controller. That's set up a little bit more like an API controller, but it's using routes.

A

One of the things that I like to do when I show this off is you can add, you can uncomment these. You can uncomment these attributes for the other for these other methods and then, when I run, this I've done comment throughout as well. When I run, this you'll see sort of how we build this graph. So I'll run this again and you're going to see something more complicated than I just showed.

A

Okay, so that's running go back to my browser, so this is what it looks like before. I uncommented that thing you can see sort of the graph of what URLs are going to match. You're, going to see the weather yeah much more complicated now, and the reason why things go all crazy is when we build decisions like what HD.

B

A

What content type we're matching those kind of things all get built into this very efficient graph structure that we use for routing so graph is actually handles. The layout graph.

B

A

This layout I'm not smart enough UI savvy enough to build something.

B

Well, we had worrying about how.

A

Well, it'll scale to lots of rules right, yeah.

B

A

I mean, and so the answer is it scales really well, we test and point routing with like the full route table of every service and Azure combined into one. That's.

B

Right, whoa yeah, so.

A

Man, how are we gonna stress.

B

This yeah I said why don't we just get the they have like a definition there anyway, but because it's used to generate the address decays or it was at one point. Okay, and so we were like. Let's just pull in that basic swag apply, it is swagger, it swagger finals, so you go swagger files and then we generate the rap. We feed that into the system, as they generated me in one roundtable that basically composes every URL for every API. That exposes an editor thing. We.

A

Wrote this to the swag out there that basically gets gets all the api's of Azure and then generates a file that defines manually defines like two thousand endpoints and.

B

We use this as a benchmark, routing.

A

And make big improvements? Okay, as well as really small stuff, because people come in things come in all sorts yeah. Also, we also did the same thing with github is API, which is more close to what we that people are building. It's.

B

Interesting and like I imagine across those, we have a variance of like how shallow yeah, but why the graph videos is how deep versus narrow yeah we.

A

Don't find deploy did.

B

Some stuff too, we.

A

Don't find graphs that are incredibly deep, because if you think about depth, depth is the number of segments.

B

Representation yeah so.

A

We don't really find graphs that are more than six or seven nodes, deep, it's more about wide yep and we find that when things are wide, they're usually are wide because of a branch on like a static string, yeah, which is what's happening this okay. So if you were curious about this, if you want to play with it John, maybe you can share the alkylate it, but it's in it's in my to my github, okay.

B

A

Presentations and then it's in the Austin's objects, I did it. I did a tour with with Carol Sigmund in Australia and New Zealand, which is this was one of the talks that I gave get I made two of the g'day mates or talked to people with poor internet connections about blazer server tour.

A

Okay, alright, so are we ready for some we're ready for some kubernetes where.

B

A

Eddie's it up alright, so we're gonna do we're gonna do very intro today. Yes problem and I'm not sure how far our content is going to stretch. We might. We might have a lot of time for questions and maybe that's good.

A

We're gonna I wanted to I'm sure I'm, trying an approach for teaching and explaining kubernetes here that I'm actually kind of interested in feedback on. If this is a good way to understand, that's it's maybe a little conceptual and we will have some examples and some demos and stuff as well so I'm, just gonna I'm gonna get into it. So about me. If you don't know me, I work on all that stuff.

A

Basically anything in asp net core I've been spending a lot of time with kubernetes and micro services and containers, maybe for the last couple months. What we're going to talk about today? Well, it's an intro.

A

We're gonna be talking about things that, if you're totally new to this, you might want to know I'm gonna, try and give you a lot of concepts and a lot of background and maybe assume that even you're not totally sold on the idea of something like kubernetes or why it would matter to you yeah we're not gonna get as far as getting to super super realistic things today. Just because there's a lot there's a lot to digest. I want to let the concepts breathe.

A

B

Hopefully, hopefully,.

A

Some of the things will whet your appetite a little bit and give you a taste and I'm happy to come back and do more if people want to see more practical than of usage and keep the comments coming in people are saying: they're happy for high-level, ok, yeah, ok, great so I want to do. I want to do a thought experiment. So I did I, didn't I studied physics in college and one of the things that I really enjoyed about physics, physics, Pro is.

B

Physics, all right is.

A

B

A

Physics, when you have.

B

A

Of the things that I always enjoyed about physics is that you could use a set of rules to prove other rules right when you start with very simple rules, and then you can derive very complex rules, you use to solve real problems. I always felt like by remembering the simpler things I could kind of bluff my way through anything into physics. We.

B

Actually, senior.

A

Year this was back when I the math fit in my head, but we we derived thermodynamic rules from quantum long yeah.

B

A

It was like wow. This is crazy.

B

So I really enjoyed I.

A

B

I liked physics, but I, particularly.

A

Liked that aspect I know that you as a student could rediscover anything that the great scientists have discovered Yeah right. Yeah I really enjoyed that aspect of it. So so what we're gonna do today is we're gonna, try and prove kubernetes. Okay,.

B

A

Gonna, try and we're gonna try and start from nothing and say like well what? If, what if your boss came to you and said, I want you to I, want you John or you Damien I, want you to design the micro-services platform that we're going to use for all of our production workloads? Okay, right! What? If knife? Do you mind a night yeah.

B

I, don't I don't mind what.

A

If I said, why do I just use docker swarm? Is this too early to that's? That's your boss would say: you're fired, okay, I need to design. Oh.

B

A

So I want that.

B

Was hypothetical yeah that'd be nice.

A

So what if you had to design a system for micro services and we've got we've got some of the we've got. Some of the following sort of inputs of requirements like needs to scale needs to support many many machines. It means to be tolerant of machine failures. So if the machine crashes in the middle of doing something- that's that's something you expect to happen. We need to be able to deal with that. It needs to support any language or any workload type we want to deploy.

A

So it needs to support, go it needs to support C++. It needs to support Java.

B

A

B

It needs turbo.

A

Pascal, an.

B

Addition, in addition to.

A

That it should support long running servers as well as jobs right.

B

As well as an arbitrary.

A

Jobs and those kind of things- okay, it needs to provide the things that applications need and that's a poor statement of requirement, but somebody would push back on that in a real discussion right. It needs to provide things that people need to be able to write successful applications, so that would be like configuration right needs to provide configuration needs to provide logging should provide. Some kind of metrics capability should provide the ability to deal with secret information addressing.

B

Addressing how do you find other services to them right.

A

All of the application level concerns and I expect I hope you still find this useful, but I expect that most of the people that I'm speaking to you today are probably in this zone like you're in the needs to provide things, applications need zone and that's okay, because there's a flavor of there's a flavor of using kubernetes, that's focused on that a lot of people have done really hard work and all the rest of these things so that you can be a developer and be productive. Does this include things like fault, tolerance and yeah?

A

I mean that's. That's here with the sort of paling and all.

B

Right so feeling yeah.

A

B

Kind of its kind of big vision, areas.

A

Right yeah, these.

B

A

Are not strictly axiomatic, I'm not gonna depend okay, sorry.

B

I went to a massive. It's all. The data analysis, self-evident or unquestioned is axiomatic. That dividends have to be financed. Is the example that sir.

A

Mr. magoo says I'm on very thin ice now so I'm just gonna.

B

Cry it down for.

A

It does support BCE I like to comment: that's real spicy. It needs to integrate with a variety of tools and by that I mean CLI ease. You should be able to build tools for this platform and it needs to be self-healing like we don't we don't want it to be somebody's job. To say the system is down. Remove the computer right, like that's IT from 20 years ago, we're at uoit for for now.

A

So, let's, let's start talking about some of these requirements and some of the options are- and maybe we'll arrive at something really similar to that. I'm gonna I'm gonna spoil some of these, but it's like well.

B

If I'm a developer and.

A

I had to think about scaling and dealing with machine failures. Well, my first concern would be I need to be able to add vm's to my cluster or my group of my group of workers. Right I need to be able to manage and control those VM, so there has to be some concept of a VM joining the cluster. To put it under my control right, because I have to be able to interact with it. I have to be able to put work on that machine. Ii, a control, plane, I need a control plan.

A

I need to know some way that that thing goes down right and and and- and that could mean like well I- have to run something on that computer, because I have to know when it goes away and the typical way that you would do that is you just do a heartbeat right, so the computer is gonna. Tell you hey, I'm, here, I'm happy everything is good and if it doesn't check in like you can expect that it goes down.

A

That's a well-understood sort of way to do that sort of thing right and that's the agent setting page pings back and forth. So we need some sort of master controller who's going to manage the nodes. We need some kind of process running on all the machines and then the master controller can update things and the machines drop in and drop out. So much.

B

Just questioning your use of the 10 VM I'm, assuming we should.

A

B

Attach to VM itself yeah.

A

B

Is required yeah.

A

That's true: it's just a unit.

B

Like it's like, whatever this.

A

Is a good job digging through the requirements right, they didn't say anything about VM, so it could be physical hardware. It could be VM could be whatever you like.

B

Unit of deployment.

A

We're using the word VM yeah, the other. The other sort of thing that comes out of this is well. If we want to be transparent to machine failures, we want to avoid anybody making decisions about where applications are going to run, but we want the system to decide or the applications run, because what as much as possible, because what, if you're a machine? What, if you said, I really want this to be on machine and then machine dies like you're in trouble right right.

A

So you want the system to be the one who decides where things go, so that's kind of how we would deal with these. These goals.

B

Until yeah I'm having flashbacks to before I joined Microsoft, are you designing a micro service.

A

No I'm back to.

B

The Windows 2000 days there was the thing from my club called the dynamic systems initiative. I, don't know how how old y'all are John.

A

Quite all, but.

B

Back then, there was a code named white horse, which was the project, and it was with. This was what the whitby series of tools came out of, so we.

A

Had no choice to skew.

B

The architect, skew yeah and the dip skew. How was an industry and if you don't, if you remember I, was in industry too, but I remember. This is then.

A

That the architecture.

B

Would let you literally model systems like we're talking about you wouldn't say machines and stuff, you would model like I have an app and their app does web stuff, and this app does processing and then the systems architect would model these systems. Let now.

A

B

Then the whole idea that the holy grail of DSi was that you would say, take your app model, which was what the architecture put together for that and then throw it into a system that have been architected using the white horse system architecture and then it would decide. Oh you have an app that needs a web thing. I'm gonna, take that part and put it here. Then you have an app that it like it's like we've been chasing this concept.

A

Yeah, as an industry.

B

A

Decade so absolute decades.

B

And I remember that that effort specifically because there was so much marketing so much hype about it, it never really materialized. Like a lot of the a lot of a lot of came out of it, we got the architecture tools, we've got the diagramming tools, but the actual self-healing self-deploying continuous deployment, self-monitoring aspect that Windows was going after with any app while all the time. Just never really, you know eventuated, you know about what you're gonna tell me: I! Think it's! Okay, maybe maybe it's.

A

We're getting we were there, so the next next set of requirements supporting any language and workload and providing the things that applications need well. Yeah I, agree about. Zookeeper has been.

B

Yeah, it's a great name.

A

How would you, how would you do that? Well, like one approach that you could do that, maybe we know maybe we notice in places in Azure that have done this and have problems with it. You could just install everything you want to run yes on the machines. Yes,.

B

A

You could you can install everything you want to run on. The machines may.

B

A

End up like a they're sending.

B

Up everything, Stahl Visual, Studio.

A

Everything why was I was thinking of I was thinking, sir, is actually to do this, but then.

B

A

You ask the question of well: if I patch, one application do I have to patch them all. So you don't you clearly don't want to globally install SDKs and frameworks and runtimes and tools on on your machines, yeah and then, and then you also have to answer the question. If you choose a strategy like that who's going to do it right right, so there's a there's another way that the industry has sort of staked out here, which is using virtualization technologies like containers yeah.

A

So if everything is a container that you run inside your system, then everything brings its dependencies along with it and you don't have to manage those dependencies or those workloads conveniently what containers provide as far as providing the things that applications need is a really good like starting point, because it speaks to all the primitive things that applications can do so using containers and using docker or technologies like docker that do the same thing you can you can set configuration via environment variables and files yeah, you can do login via console output.

A

You can talk to other services using DNS and networking that are programmable. As a docker and other tools like that, have these programmable Network backplanes and the the path is sort of trod at this point through the industry in things like making DNS updatable or making networks, little bridges, and things like that that are programmable, that can control traffic or or proxy traffic. So we can, we can. Basically, if we stick to primitives all right, we can easily provide everything that applications need by using something like docker and basically you have to figure out.

A

You have to figure out for your framework like okay, so I've got environment, variables and files on disk as tools for configuration and how.

B

A

Use of those from my my web app or whatever I'm deploying ok, but it's a task of it. Basically any framework can solve right everything, you're saying the parts are there and we need a framework to yeah.

B

A

There, or rather like any technology that you pick up and decide to use today, can do console logging right people when.

B

They talk about Eddie's, say.

A

Kubernetes is technology neutral docker? Is technology neutral care what technology you're using got it? It all works the same way here: okay, that make sense yep and then, as far as integrating with with a variety of tools like you need to Nick as Nick is making me laugh. I'm gonna have to stop as far as integrating with a lot of tools like what.

A

What is the way that we know to build things that are toolable, that you can put the UI on top of, or you can put a CLI on top of well, you would run web services, you would run rest api's mm-hmm. So if your control plane and all of the things that you're you're putting in your cluster- like let's say all the things you're deploying all the things that you're running can be represented as a REST API like you could write tools for that, you can write CL eyes for that.

A

You can write you eyes for that. So how do you manage applications? Well, a crud or STP I can manage application right right. Okay, the really interesting part of this that I think we're probably not gonna have time to delve into today. Maybe if we do four of these well well, I think it would be just that.

B

We get if we do.

A

Three of these we'll definitely get to I. Think. Yes, how do you extend the feature set right and the answer is at least in Cooper Nettie's case you don't have to have a single REST API. You can teach a REST API that it can talk to other REST API, which.

B

Is something pretty.

A

Cool but you want an extensibility model. The key is that you want an extensibility model for the control plane itself, and you want that extensibility model to not require you to put code inside the control plane. Okay right, so you want to do X. You want to support extensibility, but you want to be external to the actual control point right: okay,.

B

A

Don't want to have to like write code and go compile it into kubernetes and then redeploy your own build of kubernetes. You want to extend the processes from out our hand, side yep, and then how do you be self-healing right? If something is failing, we want to be able to restart it, and the answer is you want to limit the one of one of the strategies that you could do use do?

A

This is I think up till now that you want to describe your deployments and your applications and things like that in terms of commands right commands like deploy this thing, which is like a docker file, yeah.

B

Which is like a.

A

Docker file or, like a like a docker, run sort of yeah. You might think that you want to describe things in terms of commands, but one of the problems with doing commands is how do you self-heal on commands right? You ultimately want to represent.

A

Like do you replay all the commands that happen like it's kind of a confusing question yeah, the other problem with doing things based on imperative commands is how do you check in the source of truth about what's supposed to be in deployment, so you said that the key word there is imperative versus declarative.

B

Imperative versus declared so to you so.

A

For the system to be self-healing and all that you need it to understand, you don't just want to give it a bunch of commands.

B

Yeah tell it I'm a shouldn't: it shouldn't be a big.

A

Script right right: you want her to result in some kind of state, then you can store and you can represent and.

B

A

Way the way that kubernetes does this is it represents all your objects in terms of their actual state in the cluster and.

B

A

Does a technique called desired state configuration that people are familiar with right where you represent all this, you represent the desired state. You represent the current state and then you run a series of processes that iteratively change the world to match what you desire. So if you say, I want three copies of service a running and there are two: then you would do a loop of that reconciliation process and say we should have one more of these.

A

Please add one more now that that little point here that says this concept is called reconciliation and talking about making the desired State or making the real estate match the desired state. That's the hard part, like that's the really hard part of all of this, and that's when we start to talk about what are some of the actual tools that kubernetes gives you and how do they work and why are those things useful? Okay, because there's some really complicated stuff? That's in this last bullet point yeah.

B

Unfortunately,.

A

A bunch of other people have worked on that we don't we don't have to. We could just use it okay. So to summarize this, what have we built? Well, we have an agent running on each node kubernetes, that's called the couplet okay, we have programming programmable networking, that's the coop proxy and core DNS. Those are the kubernetes components that manage that we have a REST API, that's called the API server, which is not surprising. We have a highly available data store to represent our data and support failover. If nodes go down, that's called a TD.

A

We have. Oh, that's Edie, yeah, it's kind of hard to say it's not yep. It's like Etsy, but like right, I, don't.

B

Know how they named.

A

It there's a decision maker for allocating services to machines. So how do we decide what runs on what right? That's the.

B

A

And then there's an there's, a reconciliation process for resources, that's extensible and that's the controller manager and any controllers that are deployed so I took this from the docs from the kubernetes Doc's, and you can see all these things represented there. Basically, we've got on each node, we've got it and it could be the amaz or they could be physical machines. You've got a couplet and the coup proxy running which do networking and the couplet is responsible for ok. Let me actually start the things that are supposed to run on this machine.

A

Let me actually report back to master controller. Let me report back to headquarters that that out just crashed. Ok right, you think about it. Like somebody who's kind of spying on you, looking right.

B

It's like a kindergarten.

A

Teacher for your server right, yeah somebody.

B

A

An eye on the kids and making sure they're not going too crazy, ok as well as telling them when it's nap time or when it's time to you, know, go outside and play so I mean in a way it's like putting these small focus, little daemon processors or whatever in in like distributing now throughout the application, yeah exactly and then those are talking to the API server, which you can sort of think of as like the master of state or the holder of all state, and then basically, you can think of those communications is like the couplet watches the API server and says: oh, my node is up to I'm supposed to have another copy of this.

A

Let me start another copy of that on my note, ok and then it reports back to the API server. Ok, I did it it's running so by watching the API servers, it's like pub/sub, watching the server or pinging or.

B

A

H to be streaming actually, oh well, it's.

B

A

Streaming yeah a lot of cases, but ok again, you can do streaming over rest. You can do streaming with the API server, any basically any kubernetes api in terms of their rest. Api s, you can talk to you- can talk to them over like get a streaming result, so negligee nick's, asking from the diagram that it looks like the api server might not be highly available. So this is something that I actually was trying to find more information about. As part of preparing for this talk, I think the answer.

A

The answer that I have heard is for people that deploy kubernetes on their own machines. So a non hosted deployment like. If you didn't rent a kubernetes cluster browser or from google or from AWS, then there are ways to deploy highly available. Api servers with failover I. Think if you follow the default set of getting started, steps to.

B

A

Your own cluster I think Nick is right and you will not have a highly available API server. Ok, my understanding is that for either at least that the API server is not running in your cluster. It's sort of bridged to your cluster, so some of these components, if you're using a hosted kubernetes like from GCP or from Azure from somebody like that some of the most critical stuff is actually going to be running somewhere else, and it's going to be not even managed by you like you, couldn't even like mess it up. That's right right!

A

So anyway, that's that's sort of the summary of like if you tried to work backwards from requirements to something like kubernetes right. This is where you get to in terms of the architecture. Well,.

B

I like how you did that, because.

A

Otherwise, you look at it and it's a huge list of all these different things, and it's like for this there's this part in this, but a huge like glossary of yeah things. But it's like it's not clear what they're all doing here.

B

So we're going to we're going to.

A

I'm going to go into some more practical examples and we're mostly going to focus on the things that are on this slide today, because those are the things that developers touch okay like when you develop applications, you're, not usually thinking about what happens if the machine dies. Hopefully that's not something you have to think about all the time right, you're, usually thinking about how do I get configuration? How do I do logging? How do I just write my app code right so.

B

We're gonna talk a little.

A

Bit about using kubernetes and practice, putting we're gonna run some apps and we're going to talk about what those things are and I'm gonna sort of gently introduce like a few more concepts cool, so we're now gonna run something for the first time here and let me show y'all what we're running. I've got a really simple little espenak core application, and it barely does anything at all the thing that it does. That I want to draw your attention to right here. Okay, so I was gonna, ask. Is it like a worker template or.

B

Is this is this is created.

A

From the empty web template just.

B

A

Like some little rap handlers, who I got has some stuff here, so I am going to get for you, I'm gonna get the HTTP connection feature and the important reason why I'm doing that is so that I can tell you what the local IP address is. Basically what address what address was connected to like okay,.

B

A

Am also going to tell you the hostname of the machine that I'm running on and I'm gonna tell you, the I'm gonna tell you the value of a particular configuration key which we'll play with later. Okay. As far as the things that I have to do to kubernetes kubernetes, this I have to have a docker file.

B

A

The first time we've ever used that it is not I.

B

A

You're prepared to hear it a lot as far as the things you need to kubernetes this, you need to produce a docker container yeah. You need to push that container somewhere. It's not enough to have it on your machine, because your cluster is running somewhere else. You need to have that write it somewhere else. Okay, and we'll talk about options for how you can get started, playing with kubernetes yeah.

B

A

Well, but I'm gonna dive into showing you some things before we do that so I have I have built this and pushed it. Can you show that docker file one more time.

B

A

I'm trying to see if there's anything new, that's different I.

B

A

So I hope not okay. If there is my no.

B

A

I was just you were saying at the kubernetes that, as of negative, not kubernetes,.

B

Yeah, this is a regular docker.

A

Fergana, so if you, if you haven't seen much docker, it's a little bit complicated and you kind of have there's a little bit of a yuck factor with dr. I'm.

B

A

But it's it's powerful and useful, so we deal with the yacht yep. This is a script I'm not going to run it because I've already pushed this, but I'm, basically just building and tagging this image: okay and I'm not going to change the app during these demos. My NBC London demos has more where I do actually change that.

B

That was another question. Somebody said what.

A

Presentations, do you recommend it to go deeper and so? Okay.

B

A

London there's a ton of great material on kubernetes. If you want a very dotnet focused presentation, my NBC London one is going to go further and faster than we go today, which is a little bit more of a survey. Yeah there's a ton of great kubernetes content out there, they're good books, I would I would heartily recommend the azure as a KS workshop, which you can do even if you're, not using Azure, you can do most of even if you're, not using Azure I would super recommend that okay.

A

But let me let me play with this a little bit yep and the first thing that I'm going to do is I'm going to deploy this application and we're going to see it run so I'm gonna do something that I, don't I'm gonna do something that I. Don't super recommend that you do. This is not the terminal I wanted. This is the terminal I wanna do I'm gonna, do something that I. Don't super actually recommend that you do so.

B

Blow that up a vanilla but I can blow it up even.

A

More so people that are kubernetes mavens out there might might be distraught by this, but that's okay. This relax I'm trying to introduce the concepts new, so coop control is looking online. The communities I'm going to be doing a lot of this stuff at the command line. There are other things but I think it's the most basic way to learn. I'll show some of es codes UI as well, which is really good. Okay and then the other thing I want to mention is a m-- on Mac. You don't have to be on Mac.

A

You can do this on Windows. You can do this on anything. I think that the experience of using kubernetes from windows is pretty good. The experience of trying to run in kubernetes on your Windows machine might not be good. Okay, I've heard people having success in doing it with WSL I had not personally tried. Ok, but running from Windows is fine. Controlling it from Windows and stuff is fine yeah, so I've got a bunch of yellow files here, so f is Dutch.

A

F is from file okay, so I'm gonna deploy how you said to force my git changes. Yeah.

B

What F, let me just use that all the time? Yes, yes,.

A

B

Serious so oh oops.

A

Sorry I got that wrong. That's great! That's really embarrassing! Yeah, there's thing wrong: coop CTL apply apply, is the command for do this thing to the cluster. Okay from this yellow file got it and so I'm gonna go ahead and do apply and it says pod kubernetes app created, so I can then say: coop control wait did I space out. Did we see what that was.

B

Here we will see, we will see you.

A

Will see coop control and I'm going to say get pods because I just created the pond and you can see that I've got kubernetes app, one of one ready status, running 15 seconds and so on? Okay! So let's, let's flip over and let's look at that yep. So going back to I'm gonna close some things that are in the way in the way sorry I opened a bunch of different windows and.

B

A

I have far more windows than I need, even though I'm not using with those okay. So here's my pod definition and I have done like the most minimal dumbest thing possible. Okay, this doesn't super, look like things, you're gonna use in production yeah, but this is how much this is, how things look like when you gonna strip them down to the core, so I'm going to talk about. What's here, you can basically ignore this for now.

A

Okay, think of think of this stuff up here at least for right now, is kind of like your public static void, main of kubernetes right, like you need to you need to declare you need to declare some stuff here, but the real meat and potatoes of these kubernetes objects is always inside this spec field. The specification of the desired behavior of the pod as the dock so elegantly say. Okay and in here all I really have to do. I keep hovering stuff for emphasis, and it brings up the tooltip right.

B

A

B

Stop use my words.

A

All I really have to do to run. Something is all I really have to do is say: I want a container. You have to give that container a name mm-hmm and you have to say what image you want it to be, and so this can contain multiple containers. So.

B

Pods like a unit deployment, so this could.

A

Contain multiple containers, okay and that's. The reason why we talk about deploying pods instead of deploying containers is that a pod is like a container ish right. Okay, it's usually a container, but it doesn't have to be just one. Okay, the container could be our pod could be three containers.

A

There are concepts for pods called like containers and shutdown containers, so you could say: I want to run this container before you start my container to prepare some copy, some data right with something like that right and I want to do that before my app even starts. That sounds really handy, so you do. You could do all kinds of stuff yeah to initialize before and what's what's cool about.

A

It is that your containers that are running together will share across a space and a memory and a networking space, so they can interact with each other through networking without that traffic having to go off the pod, and things like that, so those.

B

Are some of the reasons why.

A

You might want multiple containers, but we're sort of keeping it basic at this point. Okay,.

B

Electric havoc asks get a back-end website; we usually pay them in a pod or put the web pod put the web part different to the time you.

A

Would usually put them separately and.

B

A

Reason why that would be my default explanation is you would probably scale them separately right all.

B

Right and you'd have to bound volumes of them on go yeah.

A

Thankful they have different requirements, so it's.

B

Gonna have bullion mounts and whatever else or is the website.

A

B

Even stable yeah.

A

Right but there's a lot of flexibility built in right, so, like you know, is that answer always true I think it's probably true for a standard web app, yeah somebody's, going to think of a scenario like it's really.

B

A

So there there's a lot of flexibility built into the system for how you do these kind of things, so some so that's kind of cool. Now, if I wanted to talk to this pod, how would I talk to this pod is I'm running I, said I'm running a website. I can find out more information about this, so if I say, coop control get pods and I can say output wide. This will just give me more information. Okay right, so you can see. I've got I've got a little bit more information.

A

I know what note it's running on. I know what the IP address is and stuff like that and I've done a little bit of networking before so I see 10, 244 149 and they say that's not a routable IP address. I can't talk to that all right, because I haven't said anything about wanting to talk to this pod I. Just.

B

Said make me one.

A

B

Kubernetes has.

A

Made me why, okay now, if I want to talk to it, there's a couple ways that I could do that. Let's say that I wanted to poke this thing for the purpose of like troubleshooting. One of the ways that I can do that is I have features built into coop control, for that, so I can say, coop control, port forward and then I can say, pod slash and then the name of the pod which is kubernetes app and then what comes after that is the local port.

A

I want and the remote port that I want and I just remember this, because it's.

B

And that's pretty.

A

Standard to like docker syntax.

B

Yeah, it's pretty similar.

A

Right so now, I'm forwarding from 5,000 localhost to 80,000 and kubernetes I mean that pod use and I will bring up cream. So I can hit that keep posing things that I don't want. So I can go to localhost 5,000. Then you can see I've hidden.

B

That so it came through on localhost.

A

Because of this, what it's reflecting is it's reflecting the IP address of my machine, which is where I hit it from and I can see that the host name is kubernetes app, so the the pod actually gets its own host name yeah, it self reports, the name of the pod and my configuration value is not anything because I haven't set it to anything. Okay, that's for later so.

B

That's something.

A

That we can do for a pod, let's explore a couple of so that when.

B

A

Hit control-c there that that that.

B

Cancels that canceled.

A

Okay, I'm going to show a couple other tricks here because you might be saying I'm confused by this and I don't understand. What's going on at the CLI and I gosh I can't remember all these things right, which I think you can. You know if, with practice, I look a lot of stuff up with kubernetes stone. The reason I zoomed out is for you to see this icon here and.

B

Then I'll zoom back in right. So if.

A

You go get the kubernetes extension 4 vs code, I. Think the kubernetes extension 30s code is excellent. I recommend anybody doing doctor or courier Nettie's go get these extensions and use them. They're great I've got a couple clusters here. You can ignore those. Let's focus on this NDC demo, one and let's look at what's here so I go to work, loads and I created the pod and I can open.

A

My pod and I can see kubernetes app in there and if there are multiple containers, this would say like one out of two out of two ORS right that because, but there's just one and I would see more if there were multiple containers running here. So what I can do with this? In my context, menu- and it's a little bit hard to see, but I can do a lot of stuff here in the context menu. That's useful I can create a terminal running in this pod because it is literally just a remote Linux. Oh.

B

Right I could just have a terminal.

A

In here, if I want to yeah I could manually do port forwarding, which will save me from having to type all that Sun right I could show the logs. So if you manually port forward, does it pop up? Some sort of UI is if.

B

I do port forward.

A

Here, let's see, I know it in the terminal or something it tells. You.

B

A

The choose the form of your destructor.

B

Choose reports you want to so.

A

One of the things that I use this for all the time is I always get the command wrong for logs is following the logs. Okay and this app isn't logging anything. So you just seen what came out from startup right. You can follow the logs. You can do some filtering here and they've got a nice little UI, for it I'm pretty sure that the debug option does not work for dotnet I, think it just works for JavaScript at this point.

A

Okay, I'd love for them to fix that, because it'd be cool yeah, but there's really a lot of mileage that you can get these really simple tools for interacting with it, because you could just log into it like you, can just log into it and see what the files are on disk or what the barnard variables are.

A

Things like that, Wow, okay, the other thing that I would sort of highlight, is and I'm not going to explain this, because it's it's complicated, but over here, just by clicking on this thing, I've got like way more data and you can see that this kind of looks like the manifest that me Road, but not really, there's a lot more right, so we said Rhino our kubernetes app is my is my image. It's got the name of that and there's like a whole lot more stuff here, yeah where'd that come from and and remember.

A

We said that it's a big rest API right mm-hmm. So this is literally the vs code tools, making a gap to that REST, API and saying get.

B

Me, the representation.

A

Of this thing in yellow yeah, you can do the same with Jason as well. Okay, you could output this in yellow format, and so you can see basically everything about the state of this object just from sort of going to it. So you can see in a lot of these cases like this image pull policy, always that represents a default being applied to this I didn't.

B

Specify this value.

A

I could have right, in which case I would see what I specified here. Likewise you're, seeing some sort of like kubernetes magic here right like there's a there's, a volume mount there's a file share here with a service account and a data token, and it's read-only and okay. So you can see a lot of the there's a lot of guts here and we're not gonna, try and explain all of them right now, because that's not an intro for beginners.

B

A

If you want to inspect the state of something, if you want to see, what's going on your best tools for figuring out, what's going on with something in kubernetes, I did not mean to do my chair. My.

B

A

Tools for seeing out seeing what live share is actually good to get somebody else to look at it, but.

B

Your best tools.

A

To see what's going on, are you can you can just click on this thing and see what it's see? What it's doing you can also go to describe and describe is going to give you a little bit more of a curated view of. What's here, that's just llamo what the Oh Zahraa man, it's just you.

B

Know which makes it better again.

A

But the key thing about described that's really useful when you're troubleshooting, something like if you you're doing kubernetes experiments and you're falling on your face and you're angry and you're gonna like smash a Windows computer yeah like what you could do is do a describe and then you'll see all the events that happen and chances are. If something is wrong, you can't understand why the thing.

B

Is just not working yeah see something.

A

In the events that explains why, okay, so it's like, if you try to deploy something and kubernetes, can't access your container yeah kubernetes, can't access your container, then, like you're gonna, learn about that here, yeah right! Okay!

A

So that's those are some of the really really basic debugging tools and diagnostics tools that you're gonna want to have if you're doing this, so let's get I keep going to the things that are not useful. Let's get a slightly more realistic, I'm gonna. Remove this there's a time check: okay, okay, I'm gonna speed up a little bit, but I'm gonna, I'm gonna go ahead and remove this and the way you do that is, you say, delete and then you can pass in the path to what you're doing here.

A

So kubernetes app, pod, ya, mole delete and it's going to delete my pods, so you could still they leave my one pot. You could delete you by name. Probably you.

B

Can also delete by.

A

Name, if you it's kind of considered good style, to use the files to delete.

B

Something if you.

A

Use the files to create something: okay,.

B

A

Might have more pods right, it might have, or you might have more stuff, so speak up more stuff. Let's have more stuff yeah. So I mentioned that this isn't really the practices you would follow if you're somebody doing kubernetes like you, probably won't directly use pods and the reason why you won't directly use pods is that they're a low-level permitted. Ok right so pods are like the sort of like fabric of the thing. There are lots of higher-level concepts that create pods that you use instead, so one of those higher-level concepts is called the deployment.

A

A deployment is generally the thing that you use when you have a long-running service. There's another there's another high-level concept called the job, which is the thing you use when you have a short running service. Ok, something that's going to do work and go away. There's also one called cron job, which is like well crowd.

B

A

Timed job that I want all right so, like all of these kinds of things, are useful for their own purposes. So deployment is basically like a building on the primitive of God.

B

A

Then, like giving some well.

B

A

Of how long it's lived and.

B

We and we talked.

A

About the concept of we talked about the concept of reconciliation or controllers, and things like that well, pod is built into kubernetes at a very fundamental level. Mm-Hmm deployment is built into kubernetes at a slightly higher level or because there's things running inside the cluster that turn deployments into pods. Okay. So let's look at our deployment and I want to say that from here on down, you can ignore everything, that's below the dashes for now from here on down right now. It basically looks the same as what we just did right. We have a name.

A

We have an image. We have a container the thing that I've added to this is I've defined, a container port, 80 and.

B

I'm basically saying.

A

Well, I expect to get traffic on 80 from outside the pod, make it available for just networking. You know okay, but then there's way more stuff yeah, so the signal-to-noise ratio has gone way up now that we're talking about a deployment instead of a pod and I'm gonna unpack I'm gonna, deploy this you'll notice. I've said yeah.

B

A

Gonna deploy this and then we can talk about why. That is why there's so much more signal-to-noise and maybe a little bit about what those things mean so you're creating three replicants.

B

A

Am so I've got a deployment I'll go ahead and create that and I've also got a service which we'll talk about in a little bit. So if I say cube, control gets pods again. You'll see that there are now three okay and they've got kind of a salty name right, like somebody's put a gooood in these names here. So that's one area where doctor is a little more friendly because it calls them like swashbuckling pirate or what yeah a memorable name yeah, but.

B

They're they're way, they're.

A

Techniques that help you with that I'll spend a little bit of time on so one of the things that's cool about this or I guess why why we have all this stuff is the secret is in this template node mm-hmm, because inside of this inside of this deployment is contained. What was in our pod yeah, basically right, I thought that was familiar so.

B

If we, if we paste this.

A

Here, like every I added the container port, so we can ignore that button. Everything else about this, except for the API version and the kind it's literally just the same stuff, all right as the pod and the reason is pretty simple- is that the deployment is literally the template element of the deployment. Oh, it's.

B

Literally, a template.

A

For making pods yeah right and the thing is you don't specify a name here, because you can't have three pods with the same game right a lot right. So what you specify instead is something called a label which we'll get to in a moment. Okay, but the way to sort of think about a deployment is a deployment is like well for each replica I'm going to evaluate this template and I'm gonna make one right.

B

A

If I'm gonna make one of these I have to have a selector, the selector tells me how do I find the pods that belong to the deployment yeah.

B

A

What are what are labels? Well, you might have figured it out labels or key value pairs that are for matching and querying and Association okay. So one of the things that I could do is I could come here and I could say a coop control get pod and I could paste in one of these names all right and then I'm. Just talking to this pod, but I can also say: coop control get pods and I could put in a label, could edit the label and get all the. So let.

B

Me say, pod or like it here is Milton.

A

Remember I remember pogs.

A

From kids in the hall, no.

B

No I'm Milhouse, oh.

A

B

A

B

A

Are helpful for you for querying and matching, and things like that, so if you had like let's say: I had 20 applications running here, so you can have like you're saying there with the labels. You can tag them in different ways. So.

B

A

Associate you can group multiple yeah.

B

A

So what are the common things that people do with labels is, let's say, you've got an overall application, that's made of like 10 or 15 services. Yeah right, you could factor that application into you've got one application with 10 services yeah, you could give them all a common label yep right and then you could give each one its own label as well, so that you can query for how are the 10 things doing here or you.

B

Say how is each individual.

A

B

A

Ones in the identity, service, yeah.

B

A

B

A

It's really an arbitrary grouping construct for you to do what you want in the case of a deployment you end up with this boilerplate that you'll see in basically any deployment. Ok, I have to define a label and you have to define a selector. So the other thing that's in this file that I talked to haven't really addressed. Yet it's service, a service. You can think of a service as being like a networking. Only concept, okay,.

B

A

Represent something that you can route traffic to mm-hm and they're. Their job is to create DNS entries, ok, so within the cluster, if I'm running in kubernetes and I tried to talk to kubernetes app, just the hostname kubernetes app that would resolve ok right, there's also a longer form of like kubernetes app dot service, dot, namespace.

B

A

Local, like there's, there's rules for right.

B

A

A variety of different dns names that now exist because I've declared this service. Ok, there's a bunch of different types of services, so this is a load balancer service, which means use the load balancer of your cloud to route traffic to this service. Oh ok, right which this is this again, is not a practice that people super do in production. It's useful for experimentation and for demo, so load balancer isn't built-in directly abase its load. Balancer is built into kubernetes, object, model and description.

A

However, if you took, if you went to the kubernetes repo learned it built it deployed it, it would not include a load balancer got in. You could set one up. You could set up nginx as a load balancer or something like that yep or H a proxy as a load balancer, but typically, if you're using a hosted, kubernetes you're gonna get your load. Balancer is going to be whatever that cloud platform provide. Okay, so one of the things that I can do here is I, haven't been stalling from time. I promise.

A

What one of the things that can take a while is to provision.

B

Usually, in a jug.

A

These days takes about 30 seconds, okay, so one of the tricks that you can do now that I've got three replicas is I can curl this.

A

So what this script is doing by the way you can see what it's doing, I'm using coop control in this script and I'm saying get the kubernetes app and then get this specific value using adjacent path: query: okay, so I'm using the CLI to get the IP address, basically and then I'm using curl to make a bunch of requests to that IP address, and you can see that we get responses from different IP addresses and different host names, because the load balancer is doing what it says on the ten it's doing the load, but it's doing a load balance yeah.

A

Another cool trick that we can do with this is I have got a special like sort of VIP endpoint in this application. So what was my I? Have my I? Don't? Have it I'll? Get it again, so control get service? Let's get my external IP address, so we can go to this, whereas my browser we can go to this and we're now on the public facing version of this website. Let's do something special: let's do something that only only us cool kids know. This is the right place of them. Cube control, get pods W.

A

Now we talked about watching and we talked about being able to respond to stimuli and make changes yep, so we're now watching the status of these pods. That's what W is okay, instead of just getting the ones and I've got a special VIP endpoint in here that's called crash. What's.

B

You bet it does.

A

Enhances the enhances.

B

The experience amazing right.

A

The hopes of scale yeah.

B

A

Help it seal so.

B

A

I hit this crash endpoint. You can see that.

B

It completed because I.

A

Forced it to exit I, didn't.

B

Actually make a.

A

Crash I made an exit so.

B

A

It exited it started a new one right and I must have. I must have crashed two of them, because Chrome I think.

B

Roll has repeatedly gonna crash.

A

Them now he drives.

B

A

So I guess, but it is, you can see the sort of self-healing nature of this that these things crash and when they get restarted because it says no there's there's got to be three right.

B

Okay, there's got to be Zen, so you keep restarting.

A

It'll keep trying to put you into the desired state now you're. Seeing another thing here, which is part of the thing that's complicated, is that smart people worked on is that we're now going to back off we've.

B

Said this has crashed so many times that I'm.

A

Not going to keep restarting it every second, because that just doesn't mean.

B

Is this is very similar to its fast fail protection? Oh as.

A

A similar exponential.

B

A

It has a similar restart creature. Then it has.

B

A phosphate protection where, if it keeps crashing it'll, stop yeah I, don't think it backs off quite a in a complex way. I think it just please don't stop to vibrate rise within a certain period. Okay,.

A

So, let's make let's make things so we've seen at this point: we've seen services, we've seen deployments, there's a there's, a standard form for all of these resources right and- and they look kind of like this and I've expanded this a little bit but you'll notice that all of these pre-burn Eddy's re4 resources, all sort of have the same form when.

B

A

Yellow- and this is you can sort of think of this as like the standard prelude, so everything has got an API version. Everything has got a kind. Everything has got a metadata section. The things that you can define in the metadata section are like the name, the namespace, the labels. Labels are key value, pairs for people, and then you have annotations. Annotations are key value pairs for machines. Okay, why are they different? I? Don't know, but the idea is that labels are things that are human, meaningful annotations might not be human I.

B

A

Were saying, labels were for people and I was like wait. You're using people's and.

B

We're not going to label people yeah we're.

A

Gonna be better than okay all right on this podcast, but that's kind of this form of these things and them they all have a spec. So basically everything about the pod or everything about the deployment goes into spec, okay, so the metadata is like naming and that yes,.

B

A

The minute and the spec is what it actually does yeah and all these things have that form where you can kind of just attach key value pairs to whatever okay. So let's do, let's do more, because, because you've heard that there's more, we can do with kubernetes. So.

B

A

Got another deployment here where I'm doing a little bit more, so what I've changed here is I am setting I am setting an environment where I was waiting for that I kept seeing those empty and I know.

B

It's frustrating right, so I am setting.

A

An environment variable here now, in addition, I'm defining, what's called a readiness, bro, okay, so readiness probe you people may be familiar with the health checks package from asp net core.

B

A

You get by default in kubernetes when you just have a process running in a container is kubernetes is watching the entry point of the container to see if it crashes or if it exits or so on and.

B

A

That that thing is done right, but if you want to be able to detect failure conditions that don't don't result in crashing the process, you need a health generally something more active at a health check. So typically, what people will do is HTTP there's another way. You can do it. If you don't want to have an HTTP server. They also have a built in feature where it can just connect to a port. So you say you open TCP port.

A

Whatever haven't you accept connections on it, but don't do anything with those connections yeah that will work as well. Okay, they have built in support for that, and this is called a readiness probe. They have another thing here called liveness probe and the difference between these two is that liveness probe means don't shut me down. I'm, okay, oh right, that's otherwise! All readiness, if you're, not ready, it's going to kill you.

B

A

Readiness means I'm ready there.

B

A

Different concepts: okay, so liveness means like if your liveness probe fails. That's an app crash situation; okay, whereas readiness.

B

A

Routing traffic today right yeah.

B

It's about routing.

A

Traffic, so you know so some of the things that you can do that people have done is you can have a readiness probe that says I need to read this big file or I need to initialize all this data from the database. Mm-Hmm and don't start don't start sending me traffic until I've until I've read all that data okay and healthcheck supports that. We've documented that pattern.

A

You can have two health checks: endpoints that have different criteria, for instance yeah, okay or you could say you know above a certain level of CPU usage, stop sending me traffic, which is not something I'd recommend there are better ways to do that. The other thing that I'm doing here is a mapping in some configuration. So kubernetes has a configuration system. You can define configuration separately from your deployments and your your your services and things like that, and then you can map it into your deployments and services.

A

So in this case, I've got I've got a hard-coded value here. Yeah somebody was asking about that. Like okay, what about secrets and yeah.

B

A

If I, so one thing that I could do is I can say value from and.

B

A

That it comes from a config feet, config map, that's this config map key rather.

B

A

The configuration concept there's also a secret concept, so this is an example of a config map during a DS config map over here, and you can I've literally put a JSON file yeah. This is an example of a kubernetes config map. You can also have what's called a kubernetes secret now the big difference between secrets and config maps by default is that secrets support binary data configuration does not so configuration.

B

A

To be text secrets, support binary data like your HTTP certificates, and things like that. Okay in the default, kubernetes distro secrets, will be basics before encrypted for security. Oh wow yeah. That's all.

B

A

There they're literally base64 encrypted and that's that's a joke like yeah.

B

It's true face thing: people base64 encoding does not give.

A

You Syrians.

B

A

Same as encryption, but they they're base64 encoded to prevent accidental disclosure. Okay, so if you run a command- and you see a secret right, it makes.

B

Like because you can query them, if you have the.

A

Permissions there is a permissions model for kubernetes. That's getting a little bit more into advanced and admin topics. Yeah there is a permissions model. You could say my developers don't have permission to see the secrets. Okay, that's something.

B

You because I wasn't gonna secrets, be non-binary too well. My.

A

B

Without knowing is that well, everything's binary, ultimately yeah.

A

B

To put text in there just put finer I.

A

Think somebody was making a joke, I think.

B

We'll accept the.

A

Secrets, no matter who they are, so those are some of the kinds of things that you can do if you're using a hosted kubernetes instance like a IKS or a GK or whatever Amazon's call it BBC gas, they typically have integration to use things like key bolts like a.

B

A

Vault or their equivalent as the as a backing store for secrets, I believe the AKS workshop touches on that I'm sure that the documentation is out there if not I, haven't.

B

Personally done.

A

It so those are some of the kinds of things that I'm doing here. There's a lot more, you can do you can set, you can set resource names and thing or you can set resource limits and things like that, which I just spent a lot of time talking about with some people. So, let's see what it's looked like, we need to see that environment.

B

Over it, everybody needs to see.

A

This so goop control apply. Let's apply our config first there's a neat thing that I'm not going to demo here, which is that if you try to because I'm using a volume mapping for config, if you try to deploy this without having applied the config, it will halt, it will basically say no I'm, not ready. So I won't throw up an error when I do it it'll just sit there and not do anything and then I think after a certain amount of time, it actually goes to a failed state. What I call this I call?

A

This I have very creative when I created these things deployment more, please more one of the things that you can do because it's desired state configuration I, didn't delete my deployment. Oh right, I don't have to yeah. So one of the things that I can do is I just deployed over it and says kubernetes app configuring. If I watch the deployment by the way get deployment, because these pods don't know the history right, EJ, God,.

B

A

A version of that yeah mo sort of describing their behavior. If you look at the deployment, for instance, the deployment actually has the concept of the deployment actually has the concept of how many of these are up to date. I wasn't.

B

Quite fast enough to.

A

See this in in flight yeah, but what's actually happening here, is because I set the health check to run twice with the ten-second delay. Mm-Hmm, it's actually taking 20 seconds to bring up each one right, so you can see that what we had four things running to are up to date, then we had three running and two are up to date. Now we have three running and all three are up to date: okay,.

B

We are available.

A

So deployments as a higher level concept, give you a little bit more information about. What's going on? Okay, whereas pods represent a specific version, yeah.

B

A

Let's, let's see if we can hit that do I still have my IP address. I, do not so, let's, let's curl it so now we have a very cool and configure value. Let.

B

Me see if my other thing was working.

A

Here, which I hope is working I was trying to mess with logging via config and I got it working, but I didn't try it twice. So we're not sure if it's really working and what I wanted to see is I. Basically, in that config file, I injected a bunch of configuration for logging to make it like super verbose mode. Yeah, let's see if we actually, let's see if we actually got it yeah, we actually got it. So we've got all this super verbose logging happening because I.

B

A

That config file.

B

A

Of the things about config files as well, that I would I would sort of mention. How do you choose between a config file and environment variable? Well, files are a little bit more painful to use because you have to write more Emel and you have to choose what path they're going to, and then you have to put code in the app to read that thing right. We're we've been to adding a feature with an asp net, and it's not clear.

A

If we'll do it or not, but we've been talking about the idea of using environment variables to injects config file names so that there you have an easy way to not hard-code what files are going to read from inside your application right.

B

Right so basically.

A

Out of layer of indirection, the reason why that's important is environment variables are defined when the process starts, so you can't change an environment variable while the app is running, but you can't change a file and most of our file based config providers know how to do that. The other thing is environment. Variables are less good for secrets than files, because.

B

A

Who has access to the box? Can read the environment variables? It is five o'clock, yeah yeah, all right.

B

Well, I think we're definitely will.

A

B

A

Thank you very much. Yeah yeah I.

B

Might have to actually watch this.

B

I was watching I think.

A

B

Other folks yeah, based on what they said, we might do the same thing. Yeah.

A

So so some ideas that people are saying that seem like good would be to do another one of these that now we've got a conceptual there with an introduction. Let's do let's do a walk-through of something more practical yeah. It would be great to see to like yeah, with like an asp net core app and using some of that integration and, like you know, maybe different tiers and a data back end.

B

A

B

A

Do you write the apps, knowing what they're going exactly.

B

A

Yeah very good, okay: we said that in unison that was great, very, very cool, very cool and.

B

A

Part of the time where everyone asks, if we have the dramatic zoom out working yeah.

B

A

We do have that I can't move animatic very well.

B

Thanks everyone thanks I, really appreciate the strain today. I want to give a shout out. Oh there's, two people I was gonna to give a shout out to the one person who was watching on mixer I've been waiting it holding it in the whole time, and now there's two people watching others.

A

B

I'll give a shout out to the two people who are watching a mixer. That's very nice of you to join us. I will see what we got plans for next week. You know what something so I'm.

A

Gonna be in Romania next week.

B

Well, then, do.

A

You want to do it so maybe.

B

One miss next week next week, but.

A

Well mind: you have medicine you're.

B

A

I think you got us interested about the Steve Sanderson with.

B

Wasd go the Steve Sanderson.

A

Show yes, thank you. Bro.

B