VMware TGI Kubernetes, 22 Jun 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: TGI Kubernetes 040: Kubebuilder

Description

Lots more details on this episode at https://github.com/heptio/tgik/tree/master/episodes/040

Come hang out with Kris Nova as she does a bit of hands on hacking of Kubernetes and related topics. Some of this will be Kris talking about the things she knows. Some of this will be Kris exploring something new with the audience. Come join the fun, ask questions, comment, and participate in the live chat!

Episode Shortcuts:

- 00:00:14 - Welcome to TGIK!
- 00:02:56 - Week in Review
- 00:10:50 - Kubebuilder
- 00:13:48 - Installing Kubebuilder
- 00:19:40 - Running Kubebuilder
- 00:24:45 - Creating our first API
- 00:26:10 - Our first `kubebuilder init`
- 00:34:30 - Looking at the files Kubebuilder just created (IDE time)
- 00:37:35 - Why do Kubernetes projects have these hack directories?
- 00:43:56 - Adding a Schema and Business Logic
- 00:54:56 - Integration Tests and Pushing to Kubernetes
- 01:09:09 - Watching our status
- 01:15:00 - Generating our documentation

A

Hey, what's up, everybody welcome to TGI K this week today is Friday June 22nd and we're doing an episode on Q builder. Today, I'm back in the the hot seat or I guess I should say a hot treadmill. Cuz I'm, not sitting down I'm, actually standing on a treadmill right now, I'm doing a TGI K on on Q building, ah looks like a hold on I got some feedback coming back. Let me mute my my screen here: nails for everybody's favorite time of the the week.

A

Let's say hello to everyone, so let's do hi Philippe the first one Suresh. We have Craig Olaf good afternoon from Copenhagen Denmark good. To see you again. It was great visiting Koch at Copenhagen for a cube con earlier this year, greetings from a cube farm west of Washington DC, awesome sunshine today, but raining in Florida, hey everyone, George good to see you thanks for helping out this week, LeMat e great to see you again, George sighs. Can you refresh the notes file? I updated it? Let me do that now.

A

Joe says well, my headphones are definitely working. My TIGI headcase alerts just fired and made me jump. That's awesome, I'm glad to know folks are getting the TGI K alerts and that system is working good, but sorry that it was super loud. My bad, my bad hi from Senegal good to see everyone awesome. Well, let's get started so this week, as I mentioned we're doing Q builder, but first we want to do this week in kubernetes, so let's switch on over.

A

So like always, we now have our lovely markdown file in our TGI K github repository this week. It's episode, 40 we're doing Q builder I'm, your host Chris Nova, and we have our lovely image and I think this is a link to the live stream and, ultimately, the recording folks want to check it out and then also a friendly reminder. We have a pointer pointer here on the home page for a quick and easy access. You can just click.

A

The watch live Now button and that should always go to the the most recent stream that that we're working on. We have some show notes that we'll be updating as we go and I'm going to be working out of this directory and I guess real quick. That really is a good segue of something that I was thinking about earlier this week, which is uh now that we have an open source github repository every time we do a tea tik.

A

We create one of these episodes directories and if you look Joe did this a little bit last week, but now each person has a small directory they can put whatever they want to throughout the course. The episode- and this becomes like their working space for the episode. So I've already got the tea gik repo checked out and we've got a branch checked out, so we can go ahead and update that throughout the episode as we as we move forward so scrolling back down.

A

Let's do some quick updates this week in kubernetes the first one is we have the cluster cuddle release so for folks at home, who don't know I'm one of the organizers for the cluster API working group underneath the cig cluster lifecycle, sig and open source kubernetes, and we just released our first our first tool here. If you see we have two releases and it's this tooling called cluster cuddle, and what cluster cuddle is.

A

Is it's a it's an example of how folks in the community are coming together and starting to think about deploying kubernetes clusters in a standardized way? So it's a it's command line tool. It uses mini cube to start up a cluster, and then we deploy controllers from then on and actually brings up different variations of the cluster API.

A

So ask me: look at the cluster API and, as we get different implementations growing an open-source we're actually going to be able to use this tool to get us up and running and get a kubernetes cluster up and bootstrapped one of the things we were looking at today and, let's see if I can find the issue for it was cloud provider. I. Think it's this one here we started looking at renaming, know the various cloud provider repositories in the cluster API.

A

So now we're having folks like GK he's gonna get one amazon's gonna get one I know: Marco was talking about making one for digital ocean there's a lot of folks who are talking about making this new compute level repositories in the kubernetes SIG's organization for deploying cluster API implementations. So in here, if you want to come and contribute, we would love for you to come in and cast your vote. It looks like vote fora. Number three is winning right now of how we should name these repositories moving forward and how you do that.

A

Is you just read which one of these you? You agree with the most and come down here and give us a thumbs up if you want to vote so we'd love to hear your feedback and we are really excited about seeing cluster cuddle and the different flavors of the controller is up and running an open-source moving forward. So actually, let me add a link really quick right here and I'm gonna cheat and I'm gonna edit, the the markdown directly github.

A

So don't don't tell anybody I'm doing this, but we can come down under cluster cuddle and we can one-two-three-four a little sub link there. So folks at home can get a copy of it. Bam! Okay, coming up! Next, we have this Google platform. Blog GPUs, as a service with kubernetes engine are now generally available, so big shout-out to our friends at Google. Congratulations GPUs as a service with kubernetes is now available.

A

If you, if you were at cute connie earlier this year, you saw a talk that I did with David or on cheque, and you probably also saw his keynote talking about all the different ways of getting tensorflow and cube flow up and running with the kubernetes. So I'm imagining that this is just going to be an extension of that and another way for folks who are taking advantage of gke who are interested in running any sort of GP workload which is sort.

A

You know, machine learning and training your model and and doing all of that like fancy AI stuff that I have no idea how it works. Gk is now a good place for you to do that. So big shout out to those folks up.

A

Next, we have this. This blog assal and read it earlier today. It's called the history of kubernetes and, like I, didn't give it a full thorough, like inspection and go in fact, checked everything, but everything I saw off-the-cuff, actually looked pretty legit and it's cool because it starts way before kubernetes actually was released in 2014. If you guys remember a couple weeks ago, we Joe had helped co-author the blog about the four years and open source, and this timeline actually begins in 2003.

A

So almost a decade before the whole kubernetes thing even came out to begin with. So this talks about borg and how they moved to omega and then ultimately released kubernetes in 2014, and then it jumped into the open source timeline, which is where like open source, got involved, and I started to sort of poke around at this and then alton like turned into what it is today.

A

So this is pretty cool and kind of glues, a lot of things together and you can sort of see all the different kubernetes releases and understand how how all of those fit together. So that's a cool blog to check out as well so up next we have scaling kubernetes.

A

It looks like this was written by Brendan Rios and, let's see what we have here so I think George just added this one, so I don't even know which what we have here. So let's go see when moving more than 500 containers to Kate's. We ran into some some problems, scroll down and read the big titles below to see how they were and how we solve them. Ok, cool.

A

So this looks like a good blog George checked. It out, see what's going on in chat and we can move on cool so so next we have AWS cost savings by utilizing. Kubernetes ingress with classic ELB hold on I'm gonna get some some water here really quick.

A

So this looks like a blog about how folks are able to save a little bit of money using the classic lb with AWS.

A

So this is pretty cool I wonder where this one came from, so an article about an engine X controller. This article was actually the second most popular blog post and it focused on the kubernetes ingress for on-premise deployments cool and it goes in and explains how it works and talks about all the different ingress deployments you can install. Oh, this looks really cool I definitely want to come check this out. So.

A

This next one is a pretty exciting blog I found earlier today called the hefty o Authenticator, which is an open-source tool that we released and how it works with helm. So did Joe did a really great episode last week about all the different ways we can authenticate using eyes gangway and decks, and this looks like an example of how to actually come in and do this with helm, so another great blog there. So what I'm learning is?

A

Throughout the week we've had a ton of great blogs in kubernetes, open source shout out to our friends at arc. They've finally released the 0.9 alpha so come check this out, give it a spin. We keep talking about getting Andy in here, because we really want to do arc. I've done some really cool arc live demos, cube con and anytime I go to a conference. I always make it a point to do. A quick demo live at the booth.

A

So if you want to come swing by the booth at one of the conference's I'm at I'm, happy to show arc for you and we're trying to get like a more rounded out way of actually doing an in-depth T, G I can arc moving forward. But the project's growing and it's it's really exciting cool. And then we have some more reference links down here at the bottom. So real, quick, I'm going to check out chat and the Rhoyne and jump into queue builder.

A

All right, let's see what we have here, we have more folks saying: hi I was hoping that you could quickly talk about ingress best practices. If you have a bunch of services, should each up have its own ingress, there should be a unified ingress for your services. So this is talking about how we map a single interest to multiple services. So, right now the way ingress works. Is you combine multiple, endpoints and point those at different services through different interest rules?

A

The only way I've ever set it up as having one monolithic, ingress and then I split that out to point to different services in my cluster I would defer to Dave Chaney he's done some work on hefty o contour. He can probably speak a little more intelligently about pros and cons of having more than one ingress to point to different services, but that's a great question: Arian Linate Aaron, there's a great article up on that topic.

A

I'll submit EPR, with a link to the article Chris Nova probably has some super cool ideas, though, and maybe she'll give her thoughts there. We go Justin, yeah I, just thought. Through the lessons learned or interesting to this group, yada-yada Aaron says Aaron for what it's our jaws Aaron for. What it's worth I would generally use one ingress controller yeah, that's exactly what what I was just saying as well. George, really quick, a comment on arc. This version of arc has rustic support in alpha and feedback testing.

A

The area would be much appreciated, awesome and then real, quick, I'm gonna crack the door, because I am already sweating in here. It's getting pretty hot in the room.

A

Ok rad. So, let's jump in to Q builder.

A

Cool so real, quick before we go into the the code and actually jump in the command line. This is a good opportunity for me to talk about the kubernetes SIG's organization.

A

So this is a new organization and it's like when I first looked at this about among and a half or two months ago, there was only a handful of repos in here and if you come in, you can look and actually see that we've uh we've grown quite a bit and the way this works now is any repository that a kubernetes cig is responsible for its supposed to be stored in this new github organization, and then all of this code is managed by a corresponding cig and then usually attached to like a working group or a sub project or or something that would come in and actually maintain the code or maintain the project moving forward.

A

So we've got the cluster API. Let's see if we could find that in here, which is the one of the main working groups I'm involved way. The underneath said, closed life cycle and the one we're looking out today is called Q builder, which is underneath the API machinery kubernetes special interest group.

A

So this is the tool that we're going to be playing with and you can find it at github, comm, slash, kubernetes, cigs, slash, cube builder and I think in the same way that Joe kind of likes to go and poke around at ya mole one of the first things I always really like to do whenever I'm exploring any project is just come in and get a feel for the repository and see the way things are set up.

A

I feel, like you, get a really good idea for the the type of engineering behind a repository just coming in and just trying to find the main function right away. So I picked this up earlier this week and I I try not to like learn too much before an episode et gik, but I try to learn just a little bit to kind of get my feet.

A

Wet and kind of get me excited, so I came into the CMP directory came in here, I found Q builder and then BAM main dot go, and this is where I always start. Looking at a project, you kind of understanding what's going on here, so one of the first things I noticed is we have our main function and we have a Cobra command again. We see this pattern all over kubernetes.

A

It's an easy way for us to define like a queue builder command and then add, like various sub commands. After that, you know, things like you build or create, or cute builder generate, and you can sort of define those commands as you need them. As a software engineer, and then you can come in and you can add some text here in the long description and talk about all the different ways you can configure and run your command-line tool, so Cobras a great library for helping out with that down here at the very bottom.

A

We have this run main command and it calls command help, and this is how we enter our program and then, if we come back to queue builder I bet yep. We can find all of our various sub commands here as well, so here we have an in and we have run and we have update and I bet as we go through the documentation today, we're going to come through and actually run some of these. These commands as needed. So first things. First, let's install Q builder.

A

Okay, so coming to the repo: let's go to the main page here and the first thing we noticed as it says, do not use, go, get or go install. You must download a tar, binary lease and create or create your own release using the release program to build your own release, see contributing empty.

A

So to be completely honest, the first thing I did when I found the kubernetes repo here is I typed go get get comm, kubernetes, eggs, cute builder and, of course, that didn't work, so I came back and read the documentation, but it says we want to download a release. So let's do that so releases are here and the way these things work in github is, you can come in.

A

You can download the tarball and extract it and hopefully, if all goes well, you should have an executable ready to go because go compiles to a statically linked binary. So we want the one for a Darwin since we're running it on my macbook, so let's click there and let's grab it and let's jump into my terminal cool, so I can go to my downloads directory and I bet. I can move Q builder and I'm gonna kind of do the cooking show thing, since this is still downloading. In the background. It's a pretty large file.

A

It's a hundred and eighteen megabytes, so I'm kind of curious to see what aisle we're downloading, but I've already downloaded this beforehand. So we can actually move this to our TGI, hey, repo and untie their let's go to go source, github.com hefty OTG egg, hey episodes 40!

A

Let's go here and see what we see: source github, calm, hefty OTG. I k episodes o 48 woo there we go so taro, xcf untie you builder C, like I mentioned. This is 180 Meg's, so I'm curious to see, what's going on here and hey a.m.

A

again from Turkey I'm, trying to keep an eye on the chat and trying to um car at the same time, and let's see what this unturned you okay, so this is me and I'm really grateful that the folks who built the Q builder release did it this way because sometimes you'll you'll untie a file, that's called one thing and it'll untier to a directory called something completely different, which usually, when I'm doing this it's in my downloads directory and if it gets called something else, it's pretty much gone forever because I'm not able to find it.

A

So in this case, thanks for keeping the the naming convention standard there, it's greatly appreciated. So if we go into queue builder, let's list and see what we have in here, we have a directory called bin. Let's go check out bin, and here we have ooh. What do we have here? We have a bunch of stuff looks like this is all pretty familiar to folks in kubernetes LAN, so we have client, Jen conversion, Jen, deep copy default or Jen. So these are all command-line tools at the API machinery.

A

A special interest group has built for us that we can use to generate various parts of our source code and do like common lease bits of functionality. While working with kubernetes go code, we have Etsy D, which I'm curious. Why we have that cube. Api server, also curious, why we have that in our controller manager? Okay, so it looks like we have most components of the kubernetes control, plane and I'm, assuming those are binaries, let's check it out and see if we can execute like Etsy D, for instance yeah.

A

That's that's that CDU last I checked okay, so we have a queue builder, which is the one I think we're gonna be running most of queue. Builder Jen, everybody's favorite, command line tool, queue Bechtel and we have a handful of other ones and it looks like we have a little vendor. Tarball here, checked out locally as well. Hi Mateus good to see you from Berlin uh in a real, quick. Let's check chat doesn't look like we have any major updates. So let's keep cruising along.

A

Let's jump back in the docs and see what the docs have to say.

A

Okay, so it looks like there's a getting started section and then one of the links that I kind of skipped over earlier when we were doing this week in kubernetes, is the queue builder. Folks actually have this really cool queue builder book and the book is actually like a bit of online documentation and we're gonna see how an easy way to generate generate documentation like using cue builder, a little bit later on, listen Jason tiberias, as cue builder sets tests such that they run against a locally running control plane, oh so that makes sense okay.

A

So that's why we have a TV and the controller manager and the API server running so that we can easily test our our code that we're generating using cue builder. That's pretty cool, so I explored this a little bit earlier and the one thing that I started to go through was I did I. Think was installation and setup. No I was this Quick Start Guide here yeah. This is the one I went to. Let me, let me add this to our repo, really quick, so that folks at home can uh yeah.

A

That's my like getting ready for the weekend wallpaper there so that folks at home can find this if needed and again I'm gonna cheat and edit this file directly. Don't tell anyone. I reference like this one in review: let's do I'm just gonna. Do a smaller title here: documentation used in the episode.

A

Let's add a link, yeah Jason says no. It's pronounced cue cuddle. So actually, speaking of cute cuddle, I was thinking how much fun it would actually be to write a command line tool and call it verbatim cube cuddle just to like mess with people and make it like do like something else, maybe an abstraction layer on top of québec tool or something and just to like really really described in the troll there a little bit but anyway, that's just me being evil. You don't really have to do that.

A

um So anyway, there is the Quick Start Guide for folks at home. Let me go ahead and commit this, and if you want to follow along file cannot be edited. I bet somebody came in, oh, is it here, I think I think I have another browser up anyway. I can add it after the after the episode so anyway, coming back to the Quick Start Guide. This is going to create a project, create an API and run an API and to give you an idea of what we have going on here.

A

I have a hefty Oh QuickStart cluster, which is this is the the second time I've deployed a QuickStart cluster. Usually I use keep a coin, but I like to kind of just set up just to kind of, like sanity check that everything is like working as expected and see. If there's any big major differences between there are the clusters that you can deploy.

A

I want to do a cop's cluster soon, because it's been so long since I've actually deployed a club, a cop's cluster in AWS to see what's actually going on there under the covers there as well. But anyway we can just do a quick cake. It know everybody knows I Elias, k, 2q Bechtel, and we can see that I am running in us.

A

West too, in Amazon I've got a handful of nodes up and running and let's just go ahead and see what name species we have namespaces, perfect Q public keep systems and default and I bet. If we check pods there should be no pods running okay. So this is a pretty bare-bones cluster that we have here.

A

David said: that's a statistic, thought Chris, George, says I'm taking notes, offline and we'll push the repo. You can just keep going awesome thanks for helping out George. That explains why I couldn't edit it Joe Thompson says it was officially announced at cube con.

A

It's canonically, cube, control, well, I think Brian grant says he created the tool, or at least like in architect of the tool, and he says CTL and that's where a cubic ORN came from, which was forcing people to explicitly say the last three letters of the command-line tool, but I think the truth of the matter is the world will never know how to pronounce KUB CTL.

A

Okay, so back to our kubernetes cluster. Here we don't have very much running back to us. Quick start wim KaBlam I said it I, always say it I told myself, I'm not gonna, say it. Then I always end up saying it. Installation installed up okay, so this is already starting to look familiar to the operator framework episode. We did a couple of weeks ago when we created a kubernetes controller as well, which it looked like it is vendored in ours.

A

I'm, sorry not vendored in depth, but shelled out to depth from the go binary so I'm willing to bet. If we look at the code behind queue builder, it's doing the same thing as well: let's go ahead and check and make sure DEP is installed and it is and I think I can do. Def version I want to say yep and that's going to tell us that we're running 0.4 0.1, so we have def installed. So that looks good next. It says download the latest release the releases page.

A

We just did that you checked out the directory and extract the tar and move and rename the extracted directory to use your local key builder. Okay- and it looks like here on, the right- is a step-by-step instructions on how to do that. So here's the curl command to download the tar ball that we downloaded from github here's our tar command to untie it, and then it looks like we move that to user local key builder, and then we update our path here at the bottom.

A

So I've already done a little bit of that when I was a pre loading queue builder before we started the episode. So if we go to user local cue builder, here we can see been and here's everything we downloaded that should be in the go source. Let's just compare really quick github, calm, hefty Oh, TGI, K episodes, 40q builder bin Wow, okay and those look the same and now, let's check out path so echo my pass out.

A

I hope, I, don't have anything vulnerable in here, no path, information and then here at the end, we have use your local cue builder, then so now, just to do a quick sanity check. Let's go to my home directory, another quick UNIX one-liner for folks at home, who don't know CD with no arguments brings you to whatever you have defined in your home directory. So you don't even have to do that.

A

You can just do that and save those, oh, so precious extra, two keystrokes in case you're, wondering so anyway from here I, should be able to execute everything in my path. So I should be able to run a queue builder and poof. The binary executes okay, but he yells at us- and it says: okay queue builder must be run from the project root under go path, source package, current go pass, users know the go. Current directory users know blah, so it's yelling at us that were not in our go past.

A

So, let's go to home directory, go source, github, calm, hep, tio2 gik. Let's try to run it again, cube builder and I'm gonna mute myself, really quick, because I keep getting those audible bells. How do I do sound preferences out? Poet mute? Let's see, if that does it that way at least I'm not going to be blowing you guys up with my audible bells. The entire time so anyway should be able to hit Q builder now and poof. Now that we're gonna go pass, Q builder execute okay.

A

So that's installing q builder now, let's create a new API. So the first thing we want to do is we want to create a project Q builder, an it domain, kate's IO and then real quick, because I want to kind of jump ahead a little bit here in a second we're gonna create a resource and in this example they use group ships, but I would like to do something in other than ships so that we can easily create this new idea.

A

That's relevant to TGI K I used the operator framework Lyle onstage at gopher, con EU in Iceland, and there we created a puffin operator that actually ended up echoing out various ASCII puffins on the screen live on stage. So if anybody has any ideas for what would create a resource on feel free to drop them in chat now, and then, when we get to this next step here in a moment, we should just I, have some ideas, but anyway, coming back to creating a project.

A

So the first thing we want to do is want to do queue builder and it domain cased on Io and what I want to do is I want to create a new directory me open up my terminal in TGI K episodes 40 for us to do all of our work today and then we're going to push that up to the episode like I mentioned earlier.

A

So here, let's do a make dirt we're gonna call it live, and that will be like our working directory in our working directory, where we can put all of our go code and know that we can have a main function there so going into live, and we should have nothing in here and now we can run our Q builder Nick command, Q builder in it domain gates that I. Oh, we all know my favorite domain. We're gonna, do fabulous staff, which is great I use this all the time for live demos.

A

If you ever get a chance to check it out with me on stage, it's pretty fun, okay. So what see what's going on here? So the first thing it says it says initializing project structure, so it creates a CMD directory which again that's a common convention.

A

We see and go, and especially in kubernetes land, which is usually the entry points of the program we have our peaky, J or PKG directory, which is where we're gonna put all of our internal, like guts of the program, everything from the actual API definitions and the different versions, all the way to a specific business logic or controller logic that will be writing downstream and then the very last thing it does is it says, updating vendor dependencies, running depth and sure so here I bet we're showing out to DEP and I bet.

A

If we did like a check on my file system, we could see. Actually you know what let's do this: let's split the zoom out there Oh already missed it.

A

Ok, I was gonna actually see if we could find a depth processor running at same time, but it looks like Dobbs already exited okay, so let's readjust baby I'm there and there and let's scroll up and see what we have it says: updating vendor dependencies, running death, ensure updated vendor dependency successfully and next to find a resource with queue builder create resource which, let's see, if folks, had an idea for a resource.

A

Yet, if not I might just do puffin again, because I really like the Icelandic puffin to be honest, Aaron says fabulous Afghanistan, which is true. That's where the a F suffix comes from the Afghan Afghanistan country, they're, not cheap, but if you ever get a chance to go check it out, you can actually find some pretty cool something AF domains there Joe Thompson says another fun trick is CD which goes back to the directory.

A

You were last in and that's in the POSIX definition of CD I'm gonna see your CD and raise you one more job Joe. So here, if we do list- and we can do change directory back and list again, you'll see we're flipping back and forth between our two directories, like Joe just said now check this out.

A

If we clear and we do get checkout new branch example branch, we can do git branch through an example, branch and I can do get checkout and it goes back to the branch I was just previously on, which is pretty cool with now we're back on the queue builder branch so get, and a lot of other programs also respect the command that a lot of folks don't know as well. So that's like you're at Chris, Novick Oh today get checkout cool. So we have another comment and chat here. It says, need help.

A

2000, 1806 queue builder must be run under the project route. Go pass, so I think how we were able to solve that again. I'm, really sorry! If I mess up the pronouncing your name of dinner, was you navigate to your your go pass on your local file system? So we were able to navigate into our go pass and actually get the cue builder binary up and running. So let us know if going into your go pass, is gonna work for you, okay, cool, real, quick. Let me grab some de coke.

A

Okay, um let's see Aaron says it's goes helpful, build run time that doesn't work like any other tool. So ever you have to do related everything related to undergo pass yeah and then, if you're curious, what your go pass is you can actually do an echo go pass? It will actually tell you where it is and another interesting nuance to go, that a lot of folks don't know is you can actually set your go pass to more than one directory. You can actually do like an export.

A

Go pass, is equal to your old, go pass, : delimited and then you know, like some other. If we had root directory go, but you see that some containers a lot we now can check out or we can now can echo out our go pass and actually have more than one go path that we could work in. So if we went to slash, go which do I have nope victory, slash, go see you a bang-bang in going to slash, go. Let me see the bash really quick.

A

Now I bet I can run Q builder as well. Oh, it's not in my path, but if I did user local Q builder been cube, builder I bet it won't yell at me. Oh oh! It's because I lost my go path here. So let's do export because I came up to route. Let's try this now Q builder I lost it. What happened here, I'm in flash go. I, probably lost my export so anyway, I'm going off in the weeds here: okay, let's go back to our directory.

A

Go source, github, calm, fto, tgia, I'm, gonna be so good at typing this by the end of the episode episode 40, and then we want to go to lie right, cool, alright, so back to our documentation. Okay, so we've created our our project. Now we want to create our API, so we're going to do cubo to create resource group ships version v1 beta-1 in kind is equal to sleep and we might do kind, is equal to puffin, we'll see so the kind this is again. We talked about this a little bit in the operator framework.

A

This is just an arbitrary idea, that's important to you and your your organization and building whatever API. You want to build so kubernetes api czar are super handy and I guess I could rant a little bit here about.

A

Why were you been doing this and why queue builder is even relevant but we're creating a CRD and then ultimately running a controller just using a different tool tool chain with different paradigms than the operator framework that we did recently and in this case we're defining a new arbitrary idea, that's important to humans and we're going to write some sort of reconciliation loop to go and actually respect that. So in this case, we're going to do. Group is equal to birds because we're gonna have a kind Puffin.

A

So let's do birds and version is equal to V. 1, theta, 1 and kind is equal to puffing yeah olav got it as well so yeah, let's go ahead and execute that BAM queue builder must be inside your projects, route I thought, I messed up my go pass. Let's do export it go pass. It's equal to users, go back, go let's, try this again cool, ok, so either the way that queue builders, parsing to go paths or the the syntax I.

A

Just prescribed is no longer valid in the go programming language, one of those two things that's going on so I I reset my go pass, reran it and now we're back. Ok. So, let's see what happened here, we have creating API files for you to edit. We have edit your API schema, so it's created some go files for us here.

A

It created our controller for us, that's pretty rad and we can go in and we can edit our solar functions so package controller, puffin controller that go and we have a controller test which is really handy so already we're starting to see like we're, generating tests. We're taking this to the next level and we're getting a lot of the boilerplate taking out of us are taking care of us as we start to define api's and write controllers for them in kubernetes, and then we have this hack sample, puffin dot yamo.

A

We all know how much we love looking at ya mole, so we should totally go check that file out and it says, generating code for new resource regenerate after editing resources files by running queue. Builder bill generate ok, cool, so I think what's happening there is it's actually going to read our go file and then generate our yeah mole for us. So we can do this queue builder, build a generate command here to regenerate that yeah mole as we we make changes to our underlying software.

A

So it's next install the API, run the controller and create an instance with this command here and then we get to québec to hole, apply hack, sample puffing Bat Yam, oh okay. So that's what we just did when we created our our resource. Let's go and start poking around the the directory here and actually see some of these files and start looking at some ease files, IQ builder just created for us. So let's open up my IDE we're here in episode, 40. So here's our cube builder tar ball that we entered earlier.

A

Here's the actual tar ball itself, here's our live directory and then everything in here that's kind of a funny color is something that we just generated with queue builder. So we have our go package lock in our Co package of Tamil that was created so that we could use step looks like we have a docker file here. Let's see what's going on in this docker file, this looks big, so we pull golang.org 9.3, which is a pretty new copy of go. We defined some environmental variables.

A

We download some test framework binaries we mark a few thing is executable. The API server key Bechtel we copy in the go work directory. So we're saying that we want our work directory, which is episode. 40 live to be our working directory in the container when we create the container and then it's copying, overpack command and vendor, so that our code can run, and it says, building test the API, so it says, run see, go enable equals this hero, goose equals linen Gorch equals amd64.

A

So this is the proper way of singing these two environmental variables. It's goose and Gorge. Rob Pike has a YouTube video on this. If you actually want to go and check it out, but goose and Gorger handy because the go compiler will cross compile for you automatically out of the box, which means running on Darwin I could go and potentially create a Linux binary.

A

Just by setting goose equal to Linux and Gorch equals to AMD 64, which is pretty handy because I from a single computer, you no longer need to like compile on a container or compile on a virtual machine or go over to a different computer and compile you can cross, compile right here locally on your own machine using goose and Gorge. So Philippe Martin says what is the meaning of these hack directories in Cades projects, which is really handy? Oh, you got any wine okay, so Joe third present got me.

A

This is really exciting, so I'm rainbow house wine and so I'll- probably wait and celebrate after TGI K for my my house wine here. So so, thanks Joe for my rainbow cam behind I'm live on T, G. Ok, so anyway, goosing gorgeous handy. We do a go, build. We check out the controller manager and then we run main go.

A

We copied the controller manager into a sin image which is now running in Ubuntu. So this is a pattern that that you see commonly in dr. docker files. Now that allow you to to build using this go late container and then I actually stick everything in this thin container afterwards and actually generate your image here, while building everything in the container above it, which is pretty handy and then we just copy it, and then we actually run the controller manager right here so so yeah. So that's a docker file.

A

So, let's see what else we got and then I want to answer the question about the hack directory in one second.

A

Ok, so Felipe says what is the meaning of these hack directories and gates projects. So let's actually go and look at some of these and I can try to answer them because I think they kind of mean a lot of different things to a lot of different folks. So the first project I want to pick on is kubernetes cops because I wrote a lot of these files, so I feel like it's safe for me to pick on them.

A

So if you come in here into kubernetes cops, we have a hack directory and these are all shell scripts. So these are just miscellaneous tor scripts that we would use in kubernetes for everything from building a binary to deploying into kubernetes, maybe like running some sort of integration test or e 2e tests, or even just like validate that like before somebody tried to push a commit that all the headers are actually there. So again, these are just like common tours for software engineers we actually go and we look at kubernetes kubernetes the hack directory here.

A

Oh I did kubernetes kubernetes yeah, ok, we're here in the hack directory. Now that was cool. How that just worked? um You can sort of see it's even even more busy chore shell scripts, everything from validation to verify, go imports to verify, go fund that just again make it easier for software engineers to manage and work out of the the repository here. Another thing that you'll see in the case of our let's go back to write IDE is you'll, actually see different bits of yamo here. So this is the first bit of yanil.

A

That I said we were gonna look at, but this is actually what's going to be encapsulating our project that we're working on and then later getting deployed to kubernetes, so TLDR hack is for whatever you want it to be. You can use it for anything and it's kind of a hacky directory, which is why it gets called hacked and I think it gets called hack and I.

A

Think if you want to drop a hack directory in your project, that's sort of like your I look at it as like the junk drawer, just like throw anything you want to in there and use it for anything. You want to ok, so joe says, hack directories was something I added to the early Cades repo and a set of scripts I. Think we borrowed it from docker at the time, and now it's just totally like Yolo drop anything in the hack directory and call it a day.

A

But here we see sample, slash, puffin camel so without further ado. Let's look at some animal. Oh, this is really anti-climactic, but still really powerful. Okay. So all this is. This is actually just defining a a single instance of a puffin customer resource, so we say the API version is Birds that fabulous that F version 1 beta 1, it's of type puffin or of kind puffin.

A

We give it a made a data name, and we say the suspect is equal to fubar, which we're going to define all the different various fubar key value pairs we want to for a puffin downstream and actually reconcile that in read off event. Moving forward cool felipe says: cool thanks for the explanation and abdennour says great you're using multi stage, dr files. So again, this is just a cue builder. Did everything for us?

A

This is just me, exploring the directory that cue builder already created so yeah, here's our sample, puffin, yeah, Mel and I bet we could come in and we could actually create a new copy will say: puffin, 2, dot, yeah, Mel and we'll give puffin yeah Mille. What did we call it yeah when we were in Iceland franceska had the best name for a puffin ever and I want to use the name. I can't remember what he said. Oh, he said you should name a puffin muffin, so we have muffin the puffin.

A

So thanks for in task for naming our puffin here several weeks later on TGI K. So now we have two puffins we can play with, as we start to build out a controller which will be handy because we'll be able to deploy them and you take them independently of each other. Ok, so let's go back to our documentation.

A

We've created our resource, we said it self group, Birds of type puffin and let's see what the dock site here says: optional edit the schema or reconcile business logic in the package api's and package controller respectively, then run queue builder generate, and then you can read what is a controller and what is a resource which we've kind of defined those loosely and if folks have questions just feel free to ask and Chad and I can draw a picture if needed so I want to run this queue builder, generate command and just see what what goes on here.

A

So there's cute builder, generate let's clear my screen: no wonder if there's like a verbose mode because not spitting anything out right now.

A

This is why I always think logs are important, because whenever a program takes a while and like doesn't do anything, is there no verbose mode? No, there's not I, always wonder like what would you do behind the scenes there so Aaron says: if you have time later, could you speak us to when? Is it appropriate to create a new kubernetes custom resource type? Aaron says please yeah, so creating a new kubernetes custom resource I.

A

Think for me to answer that question like what I would do every time, I I want to define something in kubernetes that goes above and beyond a simple deployment, in other words like just installing it contain you're, just running a container or just routing traffic to a container isn't enough. You know I need to take some sort of other action like I need to be able to declare something or ultimately want this to be responsible for managing something.

A

That's when I started, saying to myself, I wonder if the CR and a CR ID is going to make sense here so again the TLDR. When simply the point is not enough, is when I start to scratch my head wondering if it's, if it's time to write a controller or an operator and then again I'm going to quote Joe a controller is a very specific declarative piece of software that reconciles state and an operator is just a domain-specific controller. It's just a controller that is very specific to doing one thing and doing it well.

A

Okay, so Aaron says that's what's going on here from what I understand yeah so anyway, back to our Docs I, don't know what queue builder generate did but I think we're gonna run it later after we mutate something. So maybe we just need to make a change and then we'll see some output there, okay, so back to our documentation.

A

Well, we want to add some business logic. Okay, so it looks like the docs here say: oh well, clearly they tell us to open it in nano, so we're gonna have to do that. Live on tgia nano w package, api's shipped sloop type stock, go I'm, not gonna open it in you know. I'm gonna use my IDE here, but we want to open up sleep tight go so, let's see that's going to be in package. Api's birds, v1, beta 1, oh and it's not sloop, because we renamed it to puffins.

A

So we want puffin types dot go and if you look at the way the directory structured here, we have all of our different api's. So we could do create different projects and resources, one of which called Birds. We could do another, one called dogs, we could do one called mountains and then we could have different instances inside so in this case we have puffins.

A

So this is like the core of our app edit, this file in all caps at the top. So this is where we're going to come in and define command P abdennour. What are you trying to suggest I do with command P P? Is the font not big enough?

A

Maybe let me know the desired outcome here and then I'll see if command feeds the rate, command, T's and then suresh says what is a trade-off between operator framework and queue builder for creating a controller when to use which circumstance um so we're still learning a little bit about queue builder. But I'll try to answer that question. A little concretely, cuz I did notice some some differences along the way and we're gonna kind of discover them as we go through.

A

Actually deploying one of these controllers do kubernetes, but the TLDR is the operator framework is very specific to writing and operator, whereas the queue builder framework is a little broader scope.

A

It says you can create a controller and then, from that controller you can sort of make that do whatever you want so like you could totally build an operator with this, but this is a little closer to using raw kubernetes code, whereas the operator framework actually gave us this sdk abstraction layer on top of the raw kubernetes libraries to make writing the controller to writing an operator easier. So queue builder is going to be a little closer to the core of kubernetes, and the operator framework is going to kind of be a layer.

A

Above with this like abstraction that the folks at core OS, Red Hat, put called the operator framework SDK they put between you, the software engineer and the internal kubernetes libraries. So it's it's a matter of convenience versus stuff like for functionality, so you're sort of exchanging one for the other there so anyway, I hope that explains. What's going on so here we have our puff spec insert additional spec fields, desired, state of the cluster, and then here yep. This was what I was expecting run key builder generate to regenerate code after modifying this file.

A

So it looks like we have the same thing going on down here in puffin status and it says again run cue builder generate after we update our status. So let's do status first and let's say we have a message and we'll call that a string and then here in the puffing spec we're gonna do will have different colored, puffins, so we'll say: color is also equal to a string and then I think we need to define this jason here, where we're gonna run into problems downstream.

A

So let's actually copy we'll copy this one and tweak it real, quick for color and for message. So here we want to say color, which is lowercase, and here we want to say message, also lowercase and then of course, if those aren't there, we want to omit them from the JSON. So that's what this omit empty is doing for us doing for us here and then down here at the bottom.

A

The very last struct we have defined is a puffin struct, with a reference to a puffin spec and a puffin status and you'll see this pattern all over kubernetes that we have a spec in a status. Those are two very important parts that defining a kubernetes object here. They are defined in, go most folks, usually see them in the animal, and here above we can define spec and status and again we can put anything we want in. Here.

A

That's going to be important to us and our business or our use case so like to take a cluster API example.

A

We would have a cluster spec or a machine spec and in the case of a machine spec, we would have all of these direct these directives here in our ghost struck that define everything that we would want to define in a machine so for the cluster API, we would have things like operating system version or operating system type or the size of the machine, or maybe its IP address or, like all of these wonderful configuration bits that would define what a machine is in the case of our example.

A

Here we're just doing puffin and we're giving in a color just to have a quick demo and then, of course, for our puffin status. We have message which would just be some message about what our controller is done behind the scenes cool. So let's go and let's now run our queue builder generate and regenerate these docks because we can mutated our spec in our status here so cube builder, generate.

A

Again, the dreaded silent weight command what's happening, I should have done a git commit, so I could see what changed uhm, but I bet if we go into hack. Actually, let's do this in my IDE I bet if we go into the hack directory and we look at our sample, no that hasn't changed.

A

Let's go back to the docks and see what's what's going on here. Ok, so it says to edit our types and then also to edit our controllers, so maybe the queue builder generators expecting us to to edit both before it actually mutates anything. So, let's go see if we can edit the controller as well. So that's gonna be down here in package controller, puffin controller go! So let's see what this thing has to say again at the top. We noticed this giant edit.

A

This file, this file was created by queue builder, create resource for you to edit controller implementation. Logic for puffin resources go here. ah This is already getting exciting. So here we have a method and if you remember, we learned about functions and methods. When we did the operator framework, T, GI, k and a method is just simply a function that is attached to some data structure and it's usually a pointer to that data structure and then usually, a method is going to mutate.

A

Whatever you have, the whatever data structure is attached to you so right now, it's not really doing that. So it's kind of breaking idiomatic C, but we can. We can definitely change that moving forward and in here it says, insert your code here, which is where I think it wants us to take action and make some sort of mutation, and it looks like we do.

A

A real, quick log, printf implement the reconcile function on pufn pufn controller to reconcile, and then it passes an R, our name here, and it returns no, and then we have a controller struct okay.

A

So we saw a very similar pattern in the operator framework. A couple of weeks ago, that said that we had I was called a handler, but now we have a reconcile, but regardless both are attached to this sort of arbitrary struct that we could do whatever we want so like here. Maybe we could have a directive that will called status bid and actually just do status.

A

Let's just do this we'll do this log and we'll set that to a bool and what this is going to do is after we log something we will flip this bit.

A

So this is going to sort of act as an internal cache for a controller as it's sort of reading and dynamically pulling configuration from CRS as it's running up in kubernetes.

A

So we can use that here and we can do if DC logged, if not VC logged, so the way go works is, if you don't set it a boolean value. It defaults to false so by by the default case here will be, will trigger this. If statement and we're gonna say if BC logged I'm, gonna move this up here will log it, and then we want to flip our bed.

A

So we'll say: BCE log is equal to true quote, and that means we should only log once there and should I lose my my log statement, my login port I think I did let's see what we can do here. Let's undo some of this and go back to where we were because I don't know what logger they were importing, so we'll copy paste and then we'll come down and ctrl K delete that there we go.

A

That looks a little bit better and we can do BC logged is equal to true and come over here and have this over. So again, we now have our controller cache in memory that we can define whatever we want, and we can use that here in our reconcile to reconcile that it's needed. Okay, um so we've made a mutation as the docs have instructed us to: let's run our generate command again. Oh great, we got some output. Now, okay, I could not read package. Api's directory defined, API versions.

A

Go routine, am I, not I bet it's cuz I'm, not in the right directory. Let's try this again yeah okay, so that would be my first bit of feedback for the queue builder folks, which is, uh we totally threw a panic. We had a go routine crash, just cuz. We were in the wrong directory, so maybe a little bit of a validation check at bootstrap as we entered the main function and as run time kicks off for the program would be helpful there. So, anyway, we ran queue builder generate.

A

Let's go see what the docs say to do next, we did that and it says pushing so. It says integration testing run. They generated an integration test for your API, so go test package and again this looks like it's gonna assume that we're running in the top level directory here. I, don't know if it says that anywhere here in the docs, but that might be wise to say and also have the program remind us in case. Something goes wrong along the way.

A

So, anyway, back in the top-level directory, we gonna run go test, see what happens so. These are just regular go tests interesting and it looks like my CPU and my fans just turned on for my CPU cuz. The go compiler just really started warming up, so we're gonna. Let this go and compile I'm gonna go up, some diet, coke and maybe some of my rainbow wine here. Why do we wait for our tests to run?

A

Have you Pratt everybody.

A

So this is the integration test for your api. I actually haven't ran this yet okay, so that wasn't too horrible I was expecting not to take a little bit longer, but it looks like we have a normal output of go tests here. It'll, let us know if it run, ran okay and then, if you see a question mark that just means that it recursively checked out a package do not find any test files and just moved on so anytime. You see okay, that says. Yes, we found test files.

A

Yes, we tested them everything's, okay, and every time you see something other than okay or a question mark, that's usually a sign that something went wrong and if we actually check the exit of our the exit code of our previous command here it exited zero, which means yes, our tests have passed so go back to our documentation, so we were able to run our go test package. Everything passed and it says now: let's actually go and publish this and running kubernetes.

A

The exciting part Arin happy pride week, thanks Aaron I wore my really nice plaid just for a pride this week. It's so anyway, we have docker build command. So, let's see what we have here, it says: docker build F the name of the docker file, t a tag and let's go ahead and build ourselves a docker image so coming here, OOP, that's long. I'm gonna cancel that because that's not gonna work.

A

I accidently got my new line at the end of that command and then, of course, UNIX picked up that new line and executed my command, but we're gonna change that a little bit first. Okay, so we want to build our dr. file dot controller. So that looks good and now, let's actually call this something else. Let's call this.

A

My docker hub username, which is Chris Nova and we're gonna, do cube builder example: I already have a puffin controller, so we can't use puffin controller again and then we'll just say: let's do latest we're. Not that looks good to me, so this should build a docker image for us. Let's see what happens, sending build context to docker Damon and in the now it's running that docker file we looked at earlier. You see a lot of these commands that we looked at earlier in our docker file are running now and we're building our container image.

A

So while we wait for that, let's go what's next Q build recreate config controller image and it looks like you pass in the name of the controller and the name cube ships, so I bet I, know what's happening here. I bet, Q builder create config, it's gonna actually go in and create our key animal for us to deploy our newly created controller to kubernetes.

A

Waiting on our docker image, I really want to run this command, because this is gonna be exciting. Cuz, it's gonna generate yeah mole. This is the wine it's already kicking in.

A

Let's see you can I split my screen zoom in on this one, okay. So, while we're doing that, let's get our next command ready here, so go source github comm have to do TGA episodes 40 and going to live okay. So now we can get our command here on the right, where we're waiting for our docker image, which is Q builder, to create config controller image. Let's name this rename this rather to what we just called our image for building, so that was Chris Nova, slash, cube builder example.

A

Tag latest name cube. Ships will call it cute birds. Why not so I bet? This name is gonna translate to a namespace or something downstream. When we actually generate we're gonna deploy a kubernetes, um but I, don't think we, the code should assume that the image is on the local file system. I think this is just gonna generate Yamma, with what we name it. So I bet this is safe to run while we're waiting for our docker image to build over here, omni the left side.

A

So let's go ahead and on this yep sure enough, it worked and it looks like our docker image is done as well. So it says config written to hack, install that yeah Mille. So here again we see hack being used for hacky things and I bet. If we go look at install that yeah Mille, we're gonna start to see human or kubernetes Reis is defined in our in our you know, file and let's open that up.

A

This is hard cuz I'm like zoomed in here, so we want to go to hack, install that yeah Mille uh yeh. What would it be? Tgi K if we weren't looking at Cooper, not easy animal and talking about it? Ok, so the first thing we have is: we have a namespace and my assumption was correct. We are actually calling it the cube birds namespace, but it looks like we're prefix e naught or suffix suffix in that with system here.

A

So we now have a new system: namespace called Q birds, we're creating a cluster role, that's open to the API group that our API is defined in and we're opening up all resources in all verbs to the default namespace here and then we have a cluster role, binding, that's going to map to our service account, and then we have our CR d defined in here as well.

A

Ok, so this is different than the operator framework, so this puts all of the resources into one and then I think you just run this generate command and it'll read eight this as needed. Where's the operator framework, pull of HTML apart in two different manifests, is what they called it so again we're seeing insert of these same concepts in kubernetes defined in ran in different ways. So, some to one person it's called a manifest to somebody else.

A

It's a hack, install gamal, ultimately they're all kubernetes resources to find so interesting paradigms to see how different engineers solve the same problem. So here in our CRD, we have a name. We have our spec, we define the kind is equal to puffin. We have more than one puffin, it is plural puffins, although I'm curious. If that's that's true, do you have a flock of puffins or do you have a flock of puffins? Is it like it's like sheep and fish, so the scope is namespaced.

A

We defined some validation for our CR D, which is pretty handy. This is a really cool paradigm for folks who are interested in really like mastering C IDs and CRS and kubernetes, you can actually define like required directives and required fields and actually validate the CR that somebody's pushing up for your CI D actually needs an arbitrary set of rules, so you're actually to almost be you're, giving us an opportunity to define a schema for your CD I.

A

Don't think we saw this in the operator framework, so this is cool to see that this is defined here in queue builder, and it gives us a starting place where we can start to validate our CRS as they come through last we have a kubernetes service which, again this is just like a network layer on top of some sort of deployment or staples out or something, and it's going to tell us. You know labels that we can map to for various other primitives downstream and last but not least, we have. This is interesting.

A

We have a stateful set, the stateful set I bet. If we find our container containers yep here it is so this is what I typed in the command line earlier was probably pushed down and populated here for the image now I'm curious. Why were using a stateful set for our controller here?

A

I would want to know what was the the idea that we had in mind when we said we're gonna create a stateful set instead of a deployment which they're fairly similar I mean you can deploy a container with both the stateful set or a deployment. There's just some different paradigms about how they're scaled in the guarantees about where the containers are going to be ran with which nodes are running which pods and how those pods are named so curious.

A

Why stateless that was picked here so anyway that looks good spec replicas is equal to one so yeah. This is going to create a single pod with a single container running as a stateful set in kubernetes with some are back roles with a service in a new namespace called queue, Birds system and I bet. If we go to our docs yep, it says: let's push up our image and then we can do a cube back to apply. So, let's copy our docker push command.

A

Sorry I got my new line again. Docker push will do Chris, Nova, slash cube builder example. Colon latest.

A

Cool, so that's pushing up.

A

Aaron brought up a really great point in chat that says: there's no PBC's used here, yeah I, don't think you have to use a persistent volume or a persistent volume claim, while you're using a stateful set I. Think you just see that a lot, because stateful sets were sort of engineered to make managing state with persistent volumes in PVC is easier, but I don't think it's a hard require, like car meant that you have to use one so again that sort of alludes to like this question of.

A

Why did we pick stateful set I'm wondering if there is a technical reason that or maybe missing out on so that might be something interesting? We can follow up with the API machinery. Folks on Aaron says it's probably just shortcuts to stateful set to simplify the case where controller needs a PVC, very, very reasonable.

A

Explanation, expert explanation: there so good observation, everyone, okay, so we were able to push up our docker image and now let's go ahead and run this apply and see what's going on in communities, okay, so we're here in our top level directory. Let's make sure this is what the docs tell us to do. Next yep copy this paste here, perfect. Okay, so we created our namespace. We created a cluster role, a poster role, binding CRD.

A

We created a service in a stateful set, and this is all in the cube bird's system. So let's do Kb I learned this one from Joe watching a CGI cake. A couple months ago, KB is equal to Q. Bechdel minus n cube words system. Let's do this in double quotes, so now I should be able to KB get P Oh BAM, okay, that was cool, so Q, Birds, controller manager and I bet we can get logs for that doing. Let's do KB logs and copy the name of our pod here.

A

Oh, do I need a space yeah.

A

uh No logs, what's going on huh I, wonder if there's something going on with the cash that I implemented earlier, but we should be able to to see some logs pop out as uh oh, you know what about to happen. You know. The bet is because we haven't created a CR. Let's see if we have any CRS for us here at ease so to list our CR DS, which is this the schema. Basically, we can do TV get CR, D and then I bet we can do KB get puffins, no resources found okay.

A

So let's go back and remember earlier we defined it to puffins.

A

Let's look in our hack directory here um we have sample, we have puffin animals, so let's apply that puffin and let's check our logs at the same time and see if we can get like a side-by-side going on here. So we'll do our logs on our pod here on the left. So I didn't actually clear that now we have some logs and we'll throw in some black space there and then over here we can do K apply minus F hack was it it was sample. Puffin died, yeah, BAM.

A

All right, so we created CR and our controller loop that we just defined. We just wrote this live on TGI K. Was it able to actually go in log this this file for us or not this file? But this is the string for us and then I bet if we ran puffing to it should not one. It has the same name so I'm curious how this is gonna work. It may or may not, actually log a second line. Let's see what happens here, okay, cool, wait! No! It didn't deploy. The puffin muffin is invalid.

A

Oh because it's capital, derp, derp, derp. Sorry, we can't name a puffin with a capital muffin.

A

This would be cool cuz. This will demonstrate our cache okay. So now we do take it listed over here we have KB to find KB get puffins.

A

We have muffin and we have puffin example. We have two of those defined, but if we look at our logs, we only logged once so. This is a great example of how we were actually able to come in here in our controller code in a controller go and we were able to say BC log. So now this is a.

A

This is demonstrating that we can make all kinds of interesting things happen and go, and we can flip pigs and we can have logic or we can hit other api's or we could like do other things in Cooper, not using. We could change the behavior of our controller. Also queue builder gives us this really great opportunity to to come in and define the various spec instances that we're updating.

A

So, let's see if we can redeploy this in this time, updating our status along the way- and this is what we did with the operator framework a few weeks ago as well. So this way you're kind of getting like an apples to apples comparison of the two. So let's go back to her Docs and then, though this would be the last thing we do, but you can actually generate the documentation with queue builder and we can actually pull up like what looks like the swagger generated Docs for our API.

A

Let's see if we can update our status first I wonder how we're gonna do that um simple controller example. So, let's just grep for status, nothing controller walked functions, creating events to be see status anywhere in here. This might be like episode 2, because this is looking pretty noisy, there's a lot going on here. So we can watch watching controller resources which I think we're already doing. We can watch created resources, arbitrary resources, so yeah. These are just all different ways.

A

We can change the behavior of our controller based on our RC RNC, our ID here and then real quick, we'll two things. First, one is Olaf confirms. Puffins is the correct plural for puffin and Jason says you could add, dash dash controller type equal to deployment, and it gives you a deployment rather than a stateful set.

A

That's good to know interesting, though, still that it defaults to stateful said it's not what I would have expected out of the box, so curious, behavior, here and again in differences between two two frameworks that are sort of relatively doing the same thing, which is helping us, get see ours and see our IDs and controllers, or operators up and running in kubernetes. So, let's see here, let's go back to our hello world and, let's see OOP, not stackoverflow.com, know we want the status okay. Well, it's not shown here.

A

Most resources will split their fields into a spec in status. Let's see what this pancake controller is doing so here in reconcile. We take our pancake client and we call get I bet. We can do this and do an update. Let's see I, don't know what's going on with this air, though, let's I'm gonna pull this over to the side, so you guys won't be able to see this and see if we can't live code and use my ID to help us out here. Okay, so I bet we can do BC.

A

Is there a client here, a puffins client?

A

And then we want to do puffins, K, dot, namespace and then, let's I'm curious, to see what functions we have here, because I bet this is going to be like a goodie box here yep. So here we can do an update awesome and then we want to pass in a I. Think a puffin is what this thing takes: let's go, look and see, update and takes a V 1 beta 1, puffin, ok cool! So let's go back to our controller and do we have a puffin here in memory?

A

I, wonder how we would do we have to get it first, yeah. Let's do this, let's get our puffin and then we'll update it.

A

Get so the name here is going to be K dot, name and we're gonna do v1 get options, cool, okay, so then we're going to say, puffins it's equal to this does also return the error that it does.

A

We'll just log here, printf air, do V and we'll give it a new line at the end. Okay, so all I'm doing now is I'm trying to get a puffin in memory so that we can update a status and we can actually show that we're operating on both of the puffins we have created. So hopefully we can do this now: BC dot, puffin, client, puffins, K, dot, namespace, dot, update and then we're gonna pass in our puffin here and we're gonna, say: puffin dot status.

A

Dot message is equal to we operated on this puffin and we'll say the name of our puffin, actually we'll say the color of our puffin.

A

So we want to do a food. That's print out here cool. So this should let us demonstrate that we can actually take action on arbitrary CRS that are defining our kubernetes cluster using queue builder here. So here we want to do puffin, dot, speck, color, cool puffin status, message, Bam, Bam Bam, and then we update it. Okay, so we're gonna get the the specific puffin that we're operating on and then we're gonna update its status in print its color.

A

So now, let's do Q builder, let's do up Q builder generate and then a push I think we need to do it generate again. Let's just run it for good measure.

A

And where did I run, those docker commands earlier?

A

Okay, so here my daugher build, so we can get this one queued up our docker build okay, so we're gonna push a new image up to our docker hub registry and then will redeploy that image by killing the pod. Let's actually, let's check the yeah Mille and make sure this is gonna work, usually the way I do. This is I killed a pod and the kubernetes scheduler will redeploy the pod with the image pool policy set to equal to always but I. Don't know if that's the case here. So let's go see and find out.

A

So how we're gonna check is we're gonna come check, our install that yamo that we created earlier and I bet we can search for pull policy does not defined. It does not define it so we're gonna run it and see what happens. I think staple sets are gonna default, always but good question. Okay, so.

A

Back here, let's go to our docks while we're waiting for this and get ready for the last thing, I wanted to show you all moving forward, which was generating our documentation, sorry trying to talk and pull up dogs here at the same time.

A

Okay, so we want the book there's a book here and then we did Quick Start and then BAM. So this is what the last command we're gonna run is as soon as we redeploy our puffins controller here. So let's get our pod command ready. Okay, get pio, actually go to take it Keo and we'll do a KB delete, Pio BAM and because this is a stateful set, it should recreate it.

A

I, don't know with with or without a new image if it's gonna, repol or not, but we'll see here on the same note with deployments it'll automatically by default. Spin it up on a new node with stateful sets. It'll stay on same note, so that's the behavior that we should expect to see here because we're running a stateful set instead of a deployment and we're still waiting on our our docker, build, which I guess real, quick I can just do a quick, compare the operator framework build time versus the queue builder.

A

A framework build time which was I was able to I think push three or four docker images live on stage within like 35 minutes in Iceland, and this is taken pretty long to do here. Auntie tik anyway, Jason says from the kubernetes Doc's stateful set, is used by default to ensure that only one instance is running okay, so anyway, that's pushed so or not pushed but bill. Rather so now we can do our docker push control. Our docker push get that up and running.

A

That's pushed we already exists, we're almost done there and now let's delete our Pio and KB API, and we probably could watch this, but it's taking a while for determinate container creating we could do describe on it, but it's still going to be called latest, so we're just gonna have to check the behavior just to see. What's doing it look like so anyway, we can do KB, get Pio, look at the name of it and we'll check our logs here on the Left.

A

Implement the reconcile okay, so the logs look good, that's still the desired behavior. Based on the change we made. The real question is gonna, be if we do a que, get puffins. Let's do oh yeah, I bet and then I bet. We can grab 400 right here. We operated on this puffin, but we did not defined a color, so I bet. If we check our yeah mall here in hack, sample puffin guy animal yeah, we could do. Color is equal to blue will name.

A

This puffin um I need a good name here: well, hot we'll name it after Charlie and culturally and let's apply puffing da animal and see if we can get our status. Stop dating actually show the the color of the puffin here. So we want to do a apply: minus F sample, puffin IMO. We should have puffins birds, fabulous knife, Charlie, Frieden, I can't get puffins cool and we have Charlie, which was created 70 seconds ago and I bet.

A

If we, this is like a big drumroll moment, take it puffins, oh yeah, mole type to grab case-insensitive charlie there we go name is equal to Charlie and right here, color is equal to blue, so we were able to make a loop up and call charlie okay.

A

So anyway, let's tie all this back together, really quick and then we'll look at dachshund I'll get out here, sorry about 20 minutes after the top of the hour okay, so we were able to actually go in and define various see ours which are single instances, single tens of our CRD, which is a type or a kind puffin of group birds, and we wrote a controller for that using queue builder and the controller has this reconcile function here now notice that when we did the reconcile function, we actually had to use the kubernetes native client workflow, which is this little line here and this little line here and again.

A

This is like a cou Brunetti, as like convention like if you actually go and look at the controller's or you interact with any of the API machinery code. These patterns are an earth client go even these patterns are gonna, be very familiar to you and the operator SDK before on the other episode actually gives us a thin shim on top of that, so that we don't actually have to drill down and do the namespace and everything like that, it kind of does that for us and then we'll watch given a certain interval.

A

So this is a little gives us a little more flexibility, but we're exchanging that flexibility with complexity. So it's a little more complex, but you can still get the the same functionality that we got with the operator framework, so we were able to get a unique Puffin and we were able to update a status and actually pull a directive out of the CR and work on that here in go code.

A

So in my mind, once we're able to do all those things we can come in and actually get this to do, whatever you want and whatever is important to me and my software organization, so that's sort of like the the end in round robin, if going from yeah mole to kubernetes and indigo code and deploying and what that whole workflow looks like so last but not least, let's check out our dogs and let's check out our dog stocks.

A

So it says we can run key builder, create which we did all of this already, and it says we can just run in queue builder dogs. So let's see what this does you builder air ducts.

A

Well, I'm waiting for that sneaking some wine.

A

So it says it's running: go run package generated, open, API command main go: oh wow, it's doing a docker run index dot, HTML saved owed even wrote, JavaScript for us. Yes, I love. What I don't have to write JavaScript and it says reference stocks written to Docs reference build index.html, let's go in here. Oh, that was cool. We actually got to see it repopulate live here on my screen. We know how this Docs directory that wasn't here before and then in Docs we have reference and then we have.

A

Where was it? Let's go back to my terminal in reference. Oh, we had build index.html. So let's go and let's look in build, oh wow, so this looks like it's a pretty complex website that we're gonna be generating here and then we have index.html. So this is a cool tool for folks who are running on a Mac and there's also a Linux alternative. The Mac version is just called open.

A

Linux I think it's WG open and that's just basically like open this file, using whatever the default opener and your GUI would I would use that your GUI would use sorry about that, so we can do index.html.

A

So let's do open Doc's reference build index.html resize, and we have now generated our documentation page that we could host online likes to get us three year even running in kubernetes, and we actually get documentation for our controller as well. So this is pretty handy because you know within the course of what an hour and twenty minutes here, I've been able to create a custom controller, create a new version and scale.

A

My custom controller define a CR D defined three implementations of my CR TV using CRS and then actually generate API documentation for it as well all using Q builder, which is pretty handy. So here we have our API version, our status. We have puffin status, we have our puff and V 1 beta 1, so we could scale and create more api's as needed, and then here we have color.

A

Excuse me, which is one of the directives we added to our puffin spec earlier, and it says, insert additional spec fields, desired: state of cluster import and run ok. So this is the comment that we see above I bet. If we go here, we can update our Doc's really quick. Let's see if we can do this so episode, 40 live, let's go into package, let's go into api's Birds view 1 beta, 1, puffin types above above color here, I bet.

A

We can do delete this line in this line and add a comment that says this: is the color of a puffin set this to any color string? You like save that jump back in our terminal. I bet we can do a cube builder.

A

Was it docks, let's see and then rerun our docks and I bet we'll be able to update our color directive there, which shows how you could easily create your documentation using go docks and then push that to your your documentation site. Just by making a simple code change.

A

Cool, so let's assume back out and get out of here a little bit while we wait for our documentation to build and let's do our open command again scroll down. Aha, we have color. This is the color of the puffin set this to any color string. You like so now we're defining arbitrary directives, we're defining documentation for it. It's really nothing. Our docks were starting to really like real software engineers. Here this is pretty exciting.

A

This is a really slick tool, as somebody who interacts with a lot of the lower-level kubernetes libraries like the ones we interact with today. This is kind of comfortable for me. I think I'm going to probably use this to create arbitrary controllers moving forward, because it gives me a really really powerful springboard into getting a lot of this boilerplate out of the way and the documentation here seems. Oh, it even goes in and talks about all the different fuel definitions.

A

Okay, so the documentation here is pretty slick, but yeah I think you know we'll do the recap that I do every time. I, do two gik will say everything one time that we did and then I'm gonna I'm gonna get out of here because it's you know we're now hour and a half in and I got a mountain to climb this weekend. So yeah we were able to create a directory live on T gik.

A

We were able to generate a lot of syntactically correct, go code that compiles and we were able to define an arbitrary API, different resources and different groups and version those using the fabulous F domain name. We generated the ya mole to actually run the controller that reads from these CRS and C IDs in kubernetes.

A

We were able to deploy to kubernetes and we were able to update our controller with a new copy of software that we wrote live here on the air and that controller was able to update a status and read an arbitrary directive that we defined in go doing all that and then pull up our documentation. I think this has been a great episode. I'm really impressed with key builder hats off to the folks in API machinery for putting this together.

A

For us, this is not an easy task to do and if you've ever tried to mess around with doing this from scratch or copy and paste in another project before this is not easy. So this made it really great and easy for us. So thanks for coming, everybody I hope you all have a great weekend and we'll see you all next week on to gik Thanks.