GitLab Configure Stage, 11 Mar 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Auto DevOps technical walk-through

Description

https://docs.gitlab.com/ee/topics/autodevops/

GitLab Projects:

- https://gitlab.com/gitlab-org/gitlab
- https://gitlab.com/gitlab-org/cluster-integration/auto-deploy-image/
- https://gitlab.com/gitlab-org/cluster-integration/auto-build-image/
- https://gitlab.com/gitlab-org/cluster-integration/cluster-applications
- https://gitlab.com/gitlab-org/charts/auto-deploy-app

Projects:

- https://github.com/gliderlabs/herokuish
- https://github.com/roboll/helmfile

Development guidelines:

- https://docs.gitlab.com/ee/development/kubernetes.html
- https://docs.gitlab.com/ee/development/auto_devops.html

A

A

Thank you so for configuration this two main strands here and the softer down offs in.

A

I will start with you management because I'm you basically stop.

A

This is very part of it and then so. Basically, we've altered a box explaining its the set of pipelines that allow you to build tests, door, distinct right and the first requirement is cubanelles and then the second requirement, as part of that is, you need to install an ingress specifically in linux, in genesis, ingress controller, so to kind of make that easier for people with providers 1-click installs and that's where q- management comes in so in terms of the one-click install staff.

A

There's two parts to this, so there's where we're gonna get lab the main gate lab or faucet or e. So that's that's the big project, and this is where I canti the cuban a risk management lives. So if we go into a fine foul, a lot of our orchestration classes are is under the fastest namespace. So if you search for cluster, it's kind of there and what we really want is fastest platform.

A

No, no, this is the e-edition.

A

Lost us applications, yeah classifications. We want that one and more ingress.

A

So three installs- and there is a one-click thing where you can. One click install so I'll just quickly show you how it looks like.

A

If I prefer to set up here, it's already got a Cuban in this cluster attached and then we got operations. Cuba, Dennis yeah.

A

Yeah, so we click install uninstall first year, so I installed home to learn and you can store Perez s. So that's this is kind of one click n, it's very limited. You can only configure, maybe one or zero things about it or there's two things, but because of the limited customization and, for example, get a trainer.

A

You can actually can customize a lot of things about Gila rather say it's concurrency.

A

What else you can configure whether it can also scale or not, especially in a Q&A setting? But as you can see here, there is no such thing for you to install to customize. So you know it's very hard. So one of the things that we're doing we have a project right now, which is cluster vacations.

A

This is my book that's by the way, but uh we should document this at some point like all these different projects, so this discussed applications is basically a docker image that uses how file, if you have yourself up, I, have used helm before yes, okay, so how far is it's basically a one level, a project that uses the that shells out the helm to you know install all the different health applications declaratively. So a home, a home fowl looks a bit like.

A

Just a bunch of yellow with you can you can put values to your home file, which is nice and can control that so the whole purpose of this project discussed applications project is to enable you to because this this one-click currently is installing our applications using Ruby and when you install when you say, click install on this. If you get an error, it's really hard to show you anything because of SS RF, and things like that. So you have to.

A

We have to ask people to manually go check the port box, which is really hard to see, and you.

B

C

A

It as well so and I, don't know, what's happening here as well, so the idea is that, instead of doing that, what we do is we use that cost applications project which is a docker image, and then we have it as.

A

Template so we'll get a CI, you can have templates. It basically just saying use this image and then run this binary to install these applications and then what we've done is that you can define.

A

Its own config, so you can say I, want ingress install. So this is just some custom iam, whatever II created for that, so we can't say, install true or false. You can say whether you want a default cluster issue and so that's something base that or eventually runs as a pipeline. So I.

B

Show you an output.

A

So this gives you like the complete out what Y gets installed system that's been installed and because it's running CLABSI I, you can run anything because it's all isolated, so that's kind of the advantage as well I'm missing something here. Oh yes, so it also allows you to customize things, so you can give it by convention. Yes, you can sort of have a values, llamo to say no r1, some custom values or I want. The default is at the cocktail one and then I want more replicas.

A

For example, don't make sense, yeah feel free to jump. You have any questions, so that's one project, probably I, think I already tagged down in a team meeting, so we can copy that for all the list of projects.

A

So that's Cuban areas, management, basically and I- think you might have seen. Maybe some issues this milestone or the next milestone is creating the templates, the Hong Kong templates for it, and then we can. The idea is that we want to keep this interface, that in the backend swap out the Ruby, with CI based home, for instance,.

A

So that's the that's the part and the part that relates out to DevOps. Instead, because we want ingress, we require ingress nginx. Actually, so we give you an easier one: click install you don't have to use out of that box. If you do this, you can just click, click click and you know just for your convenience or some people. They want monitoring for their for their clusters, so they they click, install post per meter since they don't. You really have italia, unrelated to other levels that make sense. Okay and then we come to our.

A

This is a user level documentation for auto, which is kind of user to try out at some point like you already had the demo, so I won't go to that implicit there's another document which it's more useful.

A

It's the ultimate development guide and so Alto Alto DevOps is CR. A Gil FC iPhone have used gillip CI before no.

A

But you use sort of maybe Jenkins or other things. Yes, okay, yeah Jenkins file or Travis Travis, yeah, Travis, Travis or CI yeah. It's very similar to that right. So you can have all those options it's facing options. So what autotrophs is is a template on get up CI that gets provided with each release and.

A

The important distinction there's many templates by the way, so there's a lot of different templates and they're just there for examples, but the difference we are too DevOps is that there is piece of code inside gillip that says that if you don't specify a gillip CMO, this one will be implicitly used. Instead.

A

Now it's not fully anybody on killer comp, but for self-manage installs it is, and it's enabled and I'm working to progressively any will add for Gil accom I think we're 20% of projects I know so um yeah. So saying this is the script that allows you to. It has many stages and you notice it has many kind of sub templates that we do and in the first looking at templates templates.

A

It says my internet is unstable, probably wires getting slow.

D

I, have it open here, jumps.

A

So the two key ones are Butte and deploy this. This explains it really well as well is the same. So if we start from build it used to be that it was one giant screw and then we started splitting it up, and then we started shifting some of the functionality into the doctor image as well. So that's why we had different projects so for for the abuse stage. What what's happening here is it builds our doctor, image, phones, Faust's and how it achieves that.

A

Is it leverages a open source project called Heroku ish you use Heroku before right, so I think.

E

Well, have you build something like so yeah.

E

A

So in Heroku is always open source boot packs, so we just detect it, and then we just use the boot back. That's suitable for that language.

A

In that framework- and you know this is the project that does it basically so and basically what it does is that it has a very simple script to detect whether you have a doc LaFave, ER doc, a file you static, levar, you don't have to record bhagavad and use Heroku wish to do that and then finally, we kind of publish a container image to use in this this template, and then this is just some scripts to say this is the this is where generated the built. Docker image will be for that application right.

A

So, let's build and then now that we have something of you, we can then test and think that as well, just there's lots of different sub templates, but that's we can explore there later and the more interesting bit is deploying things. Oh, this different thing is conditional on that you have a cuba nearest cluster attached to your project.

A

If you don't have it, then none of these jobs will run so this this kind of limits out a DevOps to only a few analysts, but we may extend it to far get or whatnot that's kind of future plans. So what does this thing does is also again very similar pattern. There's a separate project which builds this image that the CI job runs on and it really good.

A

It's so currently it's it's, this giant best script, and this is all totally image by the way kind. Is this giant vest group that all this movement needs to be in the collapse and what I did is basically I, just extracted it and maybe tied it up, some of the best functions and then I added unit tests, so to speak for each function, best function.

A

And it's great because rephrase they didn't have any unit tests which made the whole development cycle very tough, because we every time you change something you had to run a whole pipeline which could take about 20 minutes depending on what you're building most of the time is spent uploading and downloading big container images, because you're given a discussed areas, far away or something like that yeah.

A

So there's a lot of stuff. Oh yes, there's! So! Basically this one uses home to deploy sir and.

A

We need a chart, then, because most projects, because we build some source and most projects, wouldn't some people. A lot of people wouldn't even know what what helmets okay, they know they want to use kubernetes victim. So we've provided an example. A typical chart. That's seems port 5000, and it seems that your web, where you have you, need an ingress and you need a database, basically so how to deploy here.

A

So the helm chart is called auto, deploy yeah. So this page me and you can bookmark this page. It has all the things to order different projects.

A

So this is the auto deploy app, which is the project with a home chart, and it's gone very generic templates for deployment ingress and.

A

It's got a oh, it's got, which is good and it's got a condition. It's got a dependency on Postgres, so it's got a Postgres.

A

It's got a something lamppost guy's helmet go and because it's so generic, we have a lot of options, but we try to not have too many options.

A

What we do instead is that we encouraged, if special needs for your applications or any kind of needs, data there's not covered by this, then, instead of making it more and more generic and more and more options, we try to say: okay, your option or specifying a different chart, and you know trying to try and use your own chart instead and you know alter them.

A

They also deploy, allows you to do that somewhere in here or somewhere in here so yeah, so you specify different chart either by URL or the chart name or even a even file. Yes, so I think those are the different images. Have you got any questions so far.

E

I'm just trying to understand the overall structure, so let me try to kind of explain what you just said yep to see if I understood it correctly. So there is the image that has the detection logic of which language something is written. Yes, then eat.

E

Then it uses the home chart, that is the generic home chart, and this is home to deploy that chart using with the doctor image that hasn't been built using.

C

E

Beaut backs from karach, oh.

E

A

That's it basically yeah it's not baby I mean it's got four or five steps. Yeah. Let's try and write it down. Shall we.

A

So the legend is best ways to write it, but the overflow is.

A

We have the sauce right of the project of the application.

A

A

If the sauce has a donkey, let s see I, don't yamo.

A

Then then, you know it's it's very normal dengue, let's see I. Will we use that right? So then, then then there becomes killer said. Then there's nothing to do. Well, it's not configure it's more. The Rhino team right and but if not and then after DevOps gets use instead out of the pups, yeah I call it a CI template. You know what I mean.

A

A

What happens is that we tried we.

D

D

B

Right, I skip and step you. If has a good file.

D

How to Bob's we'll use that.

A

D

B

So that's one so.

A

This is the main get that project.

A

And I really should studio out that pop star and then what we do.

D

Push the Butte.

A

So Gila has a continued interest rate might as well, because most of the time you'll be building in angular, and it's probably the closest container registry to push to redundancy docker happy something else so so far, I mean, in my view, possible to be the option into outer devotes to do that. But uh there hasn't been that demand for you early and everything in other other something run tests.

A

License King good quality. It's like there's a lot of different things, but I mean that's not just very simple and, let's see I think that's.

A

And deploy so. The first thing is: if.

D

The project and.

D

D

A

A

Which is all Bosch and also we have to use the chart right so.

A

Some point is that.

D

D

D

A

Is the project.

A

Yes, so I know once is deployed, and that's it really. So that's that's. How out to do box works on the really high-level flow that makes sense? Oh yeah, yes, and then we can talk about. Do you want to talk about a little bit about how to ingress works? Sure I'm only getting to watch d2 today.

A

Yeah, so just around things off quickly, and the only reason why we need nginx controller is because we have saved and.

B

Nginx here it's in some ways what we did.

A

A

Or is it service somewhere lost my mind, I thought we did.

A

I'm mistaken, then there.

D

Was something that said.

A

Mr. class mr. notations Oh.

A

Various and values.

A

Yeah, it's yeah.

A

Yes, yeah. We were specifically hard-code that only Ingenix controllers can can use this ingress.

A

You know something of anything else. Really anything can say anything.

B

A

Any other projects.

A

Yeah, not really I, think we mention all of it. Do you have any other questions.

C

What no I don't at the moment, mm-hm.

E

C

E

C

A

Fun all right, thanks, um I stopped my screen share.