Red Hat OpenShift OKD Working Group, 14 Oct 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: OKD Streams Operator Build Pipeline using Tekton-Luigi Mario Zuccarelli,Sherine Khoury,Diane Mueller

Description

OKD Streams Operator Build Pipeline using Tekton

Guest Speakers: Luigi Mario Zuccarelli & Sherine Khoury
Moderator: Diane Mueller

https://okd.io

A

B

Briefing on the newest and latest bits of it, initiatives that were that are taking place in the okd working group. um Today, we're going to talk about the operator build pipeline using tecton. If you don't know what okd is it's the community open source distribution of kubernetes that powers, Red, Hats, openshift and okd streams- is an initiative within the okd working group that we're doing in conjunction with red hat engineering, to build pipelines for building okd, and this has been ongoing for the past I.

B

Don't know four months now, as we have two streams: Now okd using Fedora, core Os or fcos, and now we have a sentos stream core OS that we're building okd with and have our first MVP release of that build coming out this week and part of the openshift and the okd ecosystem relies heavily on operators, and so today we're going to talk about the operator build pipeline that we've created using tecton.

B

That's part of this okd streams, initiative and Luigi zuccarelli and Shireen Corey are both going to walk us through all of this and so um Luigi. Take it away. Tell us a little bit more about okd streams and we'll have q a at the end.

C

Okay, so by web introduction, my name is Luigi, as, as uh Diane's pointed out, I'm working on the CFE team and I've. Just let Shireen introduce herself quickly as well.

A

Yeah so I'm I'm, Shireen I'm, also with the red hat uh openshift Team I'm working on the customer, focused engineering team, so same team as Luigi.

C

Cool okay, so um we're going to be introducing the operator build pipeline using techton, and um so, uh as as Diane pointed out, we have this notion of okd streams and she's she well. She put it really well there not only do we build FH cos as the underlying operating system, but we also now have s-cos, which is also coming out in the MVP, and so we've coined it as okd streams.

C

um So it gives us the notion of building multiple pipelines to realize okd and all the the necessary operating system that supports it um and then tecton and tecton pipeline requirements.

C

um Tecton is a kubernetes open source project for cicd pipeline builds. It's really container based it's fixable, extremely easy to use. In my opinion, maybe someone might not agree but very easy to use it's extremely flexible, as I've mentioned, as as the great thing is they and we'll, we have a lot of um at the end of the slideshow.

C

We'll have a lot of links for you to go, have a look at, um but the the actual the power of Teton is the tecton tecton Hub, where you've got a a sort of plethora of of tasks that you can use and Implement at the box without even changing a line of code, but you can obviously customize and use it as as needed, so um just a brief architecture, overview of what we're trying to sort of uh sort of build here um and and why we've used tecton and what we're actually trying to do here.

C

So on a high level, I've got a a architecture diagram just just to point out the different pieces that are used in the actual um build pipeline. So we have our pipeline runs over here. um I, don't know! If you can see this yeah over there is the pipeline runs, we have a pipeline objects and we have task objects.

C

Well, the toss really could be like something like a gits. Clone um builds a golang, um Builder golang program, put it into a Docker that type of thing. This is what we mentioned by tasks and then the supporting infrastructure. We have um sorry here. We have config Maps conflict maps are sort of objects within openshift that store key value Pairs and there we store the credentials used for pushing images. We use canonco the container Google tools container.

C

The tool is really really cool, powerful, it's it's, it builds and pushes the images and then obviously we have a PV persistent volume and persistent volume claims and then obviously all the necessary rbec for us to realize the um uh all. The all the different, rule-based access controls permissions for the tasks.

C

So that's the the actual architecture, the overview and what I.

B

Wanted to highlight here is.

C

um Just the different objects that we're using within the um within the operator pipeline tree you'll have a you'll, have a look at the top there's environments overlays in crcd, and we make heavy heavy use of customization customization is a is also an open source tool that really helps with um deploying multiple objects or multiple manifests so with it is plugged in using the cube, CTL command line.

C

So instead of saying, for example, Cube CTL mine apply minus F, you can say Cube CTL apply minus K, and it knows that it must use customization and the customization will run through and deploy your full application.

C

um So yeah that sort of gives you a high level overview. um What I do want to iterate here and and sort of explain is that this is a POC.

C

It is a sort of MVP, as Diane mentioned in the beginning. um So it's very naive and very opinionated at this point in time. uh We there are a lot of caveats in here, but what we wanted to show- and what we want to do um highlight is that we're able to build operators and um and and then deploy them into the operator Hub. So more about operators, LM, bundles and operator have them.

C

The operator really is a is a a way of actually um putting your your SRE, your site, reliable Engineers work into a a programming language like golang, where we can actually then take all the um stateful assets, the order that we deploy, something in and and actually deploy that programmatically and then, as well as looking at the the the desired state that we want.

C

um The deployment to be in and the operator will run in a in a reconcile Loop and make sure that we have the desired state that we want. That's at a very high level. Again, there's lots of good information on um that. We'll point you to at the end of the slide that you can read up on operators.

C

Olm is really the the operator life cycle manager and that makes sure that the the operator is um that we're able to have a life cycle, in other words, if I bump the version of an operator I need a way of seamlessly upgrading that operator, and we can do that via Olin and olm, bundles and but the Olin bundles.

C

We'll talk later um about and you'll see in the demo, where we, where we create the operator, the operator might have a an agent that it needs and like we call it the upper end and so we'll need to bundle the operator, the operand and then put it into a catalog. And then the catalog is what we talk about operator hub and in operator Hub.

C

um They are operators that you can't access, because in openshift these are subscription based. So in the red hat Marketplace, for example, there could be um tecton and Argo CD or some really cool database, but you can't access it because it is a subscription based and it's in openshift. So we want to build these pipelines to allow um sort of the community access to build operators and and catalogs and push it into the communities operator Hub so that we have these operators available for okd um and uh well.

C

That leads me on then, to uh handing over to Shireen to do a a live demo. I believe Sharon.

A

Okay, so uh what I'm gonna show you here is how to um build an operator using the pipeline that we've we've built on a kind cluster. So to start off what is kind kind is kubernetes on Docker, so basically it's using a Docker containers as nodes and building a kubernetes cluster.

A

On top, you can find that actually on six kubernetes IO, it's very, very easy to install and it's extremely lightweight, so it's very, very adapted to developing basically can create a cluster by doing something like kind create cluster once you've you've installed it I already have one so get cluster. My cluster is right. Here doesn't want to detect my finger anymore.

A

That's fine, okay and get cluster sorry. So uh we we've got a cluster as you can see, I'm using sudo here and that's simply because I'm not using kind on Docker but kind on podman. So, if you're interested in that, we can also send some uh some links to to tell you how to tweak that now, on top of kind, we've installed the tecton controllers so to install that it's pretty simple, there's a tact on dev and some really nice yaml file.

A

Where you Cube CTL apply the thing and it installs it for you, so I've already done that I can show you here. I have a namespace called tecton Pipelines, and in here we can see that we have all the services pods deployments, everything that is needed, including our back. All of that is already in there. So it's a one-liner. You get tacked on on kind running, so that's uh basically about what I need in order to to get myself started up.

A

I've uh just pulled the okd operator pipeline repo, where the pipeline code lives right now. Also we got the we'll get the the links uh afterwards and what I'm going to do is simply apply well, first, before applying I'm going to change my name space to okay, the team.

A

uh This is also in the readme. It tells you to use okd team and, as Luigi said uh a bit earlier, we're using customize in order to be able to apply everything so so I'm gonna apply here.

A

uh The uh the customized environment and what I get out of that is everything are back related that makes the tecton pipeline work.

A

I get the pipeline uh deployed as well as three tasks, so uh we've got one task clone to clone the repo from GitHub, and this is basically one of the tasks that exist on the tecton Hub. We we didn't really create that the two that we created are the container all and the bundle so now that we've applied this to our namespace, we can start the pipeline.

A

So, as you can see, there's a lot of parameters here, but it's pretty simple, so I'm telling it to start the pipeline Dev. All this is the name of my Pipeline and then I'm, giving it a set of parameters to tell it here is the operator repository that I want to build.

A

This is the name of the uh the repo where I want to push my um my operator once my all of my Docker images, are built and ready.

A

uh This is the registry, where I want to push the version of the operator and I'm gonna use a volume claim, based on the standard storage class that you get out of the box with kind.

A

So this volume claim is really so that we can share um files between tasks and between steps of the pipeline so that we uh we can, for example, get the advantage of car of caching and not have a long pipeline. So, let's um okay, so let's start our python our pipeline started. It's got this random name and we can even use this line here. To look at the logs see what's happening.

A

So here the pipeline is going to take around five minutes, so this is going to take a little bit of time around uh seven minutes more or less so it's gonna start the first by cloning, the repositories as you see here and then it's gonna go into another um task where it's going to build uh the containers so starting by a big, golang, CI step, which is by the way we've made it not uh to fail the whole pipeline because we know operator to operator the the the linking rules could could change and the the quality standards are not the same so next uh unit tests and then we go into building uh the operator.

A

So while we wait probably for this to kind of finish off, I can show you a little bit uh inside the code. So this is what a tecton task would look like. uh There's the series of parameters, it's basically exactly the same parameters that we passed earlier in the command line. Here we see what we've been doing with the PVCs.

A

If you remember earlier so this is where we load the PVCs into the task so that it uses the cache and the PKG folder from the PVC instead of rebuilding each and every time it's going to start a new step. So uh the yellow step, the goal length, CI, lint step that we saw earlier. Is this verify step? So here you can see that it's taking an image that we've built and it's using that container in order to run a golang CI.

A

So two little things to say here: the go: bundle tools is a container that has basically uh go golang, CI operator framework SDK.

A

um What else Luigi I don't remember, but all of the.

C

uh The tools, customize and supporting tools all.

A

The tools that you might need to get that operator um uh basically built and run and be able to to kind of push it um so going back to here.

A

uh We see that our build step is done, and now what we're doing here is preparing the uh Docker container of the operator to be pushed to uh to a registry we're using canico here, which is very, very practical, because uh we don't need to use a Docker engine. It runs unkind and it can build any container exactly like uh if we were using podman or if we're using a docker.

A

uh It does both the build and the push in the same uh in the same step. uh Basically so here, as you can see, we've pushed the operator you're free to basically add another step in here to build an agent if your operator is using an agent or a cube, rbac proxy, for example, if you're, if you're, using a cube, rbac proxy with your operator- and you need to build that and uh what happens next here is a this second task starts.

A

So the bundle all is where we start: building the olm artifacts or the operator life cycle management artifacts. That are needed so that we can use our operators from, for example, operator, Hub or a community Hub and install them on an openshift cluster on okd cluster and start using them. So first thing we're building a bundle, as you can see, we're using operator SDK things. So it's it's based on operator SDK, and uh so we call bundle validate uh once that is done.

A

We can push that bundle to to um to a registry, and next we need to create an index and a catalog. So these might look like a fancy fancy words, but they're they're really related to the way olm handles uh the bundles. So uh what olm in the operator Hub shows is operator. Catalogs and catalogs contain either one operator like what we're we're creating here, we're creating a catalog containing one index which contains one operator, but if you take, for example, the uh uh the red hat uh catalog it contains for a release.

A

For example, it contains all of the operators that this release has so here to make things simple and to allow you to use your operator really easily. We've created the catalog image for you and the index image for you. Based on that, you can use, for example, a subscription um yaml on your okd cluster, and you can very easily deploy the your operator to the cluster from that catalog that you've built.

A

So that's about it as.

C

We said it's pretty opinionated.

A

C

A

C

Finished, would you mind just doing a tkn describe um the pipeline bill that you've just made? Yes,.

A

So uh here is our uh last run of the of the pipeline, so you can see basically that this time it even took a minute less than last time, which.

B

A

Is a little bit thanks to the camera, the.

C

Caching is working nicely, so if you don't mind just TK in um PR describe the name and it'll give you a nice fancy little.

A

A

That's how it looks it lists all of the tasks and the time that each task took.

A

So should we look at the query I o to to see uh what bundles got, what images sorry got uh uh got pushed Maybe. uh So here is well done. I'm just gonna share this here.

A

And sorted by last modified, so here are all of the images that we've built. We've built the operator we've built the bundle that basically describes the operator describes all the related images there is it describes which channel it. It belongs to with version which are back it needs, which deployments it needs. All of that is in the bundle.

A

It's really something that the operator framework SDK builds for you, then we have an index which contains uh the operator bundle and finally, the catalog, which contains one or more indexes. According to your, um not your use case.

A

So that's about it, we're hoping that soon, we're gonna be able to run this on operate. First right.

C

Yeah yeah we had um we had uh previously, it ran on operate first, um but we had some credential issues, so we couldn't unfortunately show it today, but uh we will, we will be together with the s-cos build pipeline will show both of them um in the in the near future in the next couple of days. Really so I'm going to I'll share the last bit just to show the regards the links and.

C

Yeah, can you see that.

B

Yes looks great so.

C

Great, so um the the GitHub is okd project and um specifically okd operator pipeline, so please download Fork it play with it um as Shireen showed you, it's really easy to build locally and I think this is the power of what we wanted is that you can build these pipelines locally. You don't need a heavy um kubernetes infrastructure or openshift infrastructure to build it. But having said that, it really makes it makes it really works well on a heavy openshift, okd streams, a type of uh of cluster and will work well. So PRS are welcome.

C

As I've mentioned these the SDK operator framework.io, there's some really good documentation um have a look at operatefirst.cloud and then obviously okd.io for information on our okd and okd streams. So that's it um we'll hand over to Diane.

B

Well, thank you very much for that. um It's great to see the progress, we're making I, know, operators and getting operators to work with okd um and creating the catalogs has been one of the the initiatives within the okd working group. That's been going on for a long time, so this is really amazing to see it in a pipeline.

B

um The other thing I think you mentioned slightly, but we um we are going to put these up on tecton hub for people to use so they're right now, they're in the okd project, um repo, so go there make comments, tweak them out.

B

We know you know, I know these guys sound like and are now experts in this pipeline, but we know there are tecton experts out there in gurus and we'd love um to have your feedback on this and see if we can't make them more efficient um and effective and, as always, um probably better documented than you know, we do on our first passes. So if you have um PR's or comments or any issues, you try running this and you you have some feedback. Something didn't work quite right.

B

Please let us know these things should run on any kubernetes cluster, um so um they're not totally specific to running on openshift or okd, but if we put anything in there, that is um we want to hear about it. So um again, thank you, Shireen and Luigi for taking the time today and I'll. Add this to the playlist of okd streams, um videos that are up on our YouTube channel and we will be following this up with more in the not too distant future.

B

So, if you're interested in this stuff, um please go to okd dot. Io um come to one of the working group meetings, um they happen, bi-weekly, um there's a Community Development one on the other week and there's lots of things. People can do to get involved and we just would love to hear from you.