GitLab GDK and Gitpod walkthroughs, 20 Jun 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Setting up GKE with GDK

Description

Tristan walks through the process for setting Google Kubernetes Engine (GKE) to work with the Gitlab Developer Kit (GDK)

https://docs.gitlab.com/ee/user/project/clusters/#add-existing-kubernetes-cluster

A

Oh welcome to a recording of the setup of kubernetes for monitor development, so the goal of this is to is to be able to, at the end of the dislike, be able to see the cluster, see that page, the kubernetes cluster house charts and also the monitoring dashboard in all of the data on there there's a few prerequisites so having the GDK up and running. There's a pre rick, sir I'm not going to cover that today. It's just assumed that you've got that going also having docker installed and having cubed control install so yeah.

A

This assumes use GDK, set up the docket install and cube control or cube CTL, as it was sometimes pronounced, I'm also going to cover some stuff with the get by Brunner. So it also assumes that the runner has good labor anna has at least been downloaded and installed yeah. So, to start with we're going to create a new project in running GK.

A

So projects new project, so this is the project that is going to be accessing the kubernetes cluster. You can also do group level clusters, but I'm just going to show project level today, yeah, so there's one or that does some order, DevOps stuff.

A

So this one here, I'm gonna call it could be Nene's dinner.

A

Let's give her this, like I play, haven't used that already and create the project. So there's this Auto DevOps deploy project just has like some CIA CD stuff that will, that is, gonna, be over like deploy to a pasta and so.

A

Yeah, so the next step is to take a look at the kubernetes pane for this project. Now, as you can see, there's no cluster attached so yeah. What we want to do is we're going to add an existing cluster, but this cluster doesn't exist yet. So that's what I'm going to show next, how to set up a cluster on gke.

A

By the way, if anyone has any questions, just like feel free get Python or check them in the chat.

A

Yes, so alright kubernetes sitting up for kubernetes cluster in gke, so the monitor team has this. Has this namespace so at least for monitor team members- and this is the one to use- there's also get lab internal which any get lab employees can use, but it's it's limited by, like the I can umber of IPs and a few other things.

A

So if you're, if you do create a cluster beer and just make sure to remove it, if you're not using it anymore, so for now, I'm going to going to go with monitoring development and and create a new cluster.

A

I'm gonna choose to see your first cluster because it has some really like low requirements. Actually now, let's go to standard class that just maybe to show so like have a higher number of nodes and stuff. So.

A

Yeah I can give it a name. So in this case, ideally, you probably want to be identifiable back to you. So I've put my first initial and my last name.

A

You need to see that demo. It doesn't really matter what the name is. It doesn't get referred to by anything standard, CPU couple of nodes for development. You probably can just be like one node, if you like, depends while you're using it for bath. Everyday usage, doesn't need more than one.

A

So there's now being set up.

A

So yeah this will take a couple of minutes.

A

Yeah, so the next steps I'm going to go through I can show peak, will be there documented here and then get my documentation connecting get lay up with a kubernetes cluster.

A

So going down to this, adding an existing kubernetes cluster is what we're going to be following and.

A

Yeah so I've actually already gone to this page it he saw earlier to add an existing cluster, so we'll give it a name.

A

Good news demo, the URL test will come once the clusters created.

A

Uh-Huh looks like I: have something has a your endpoint at least.

A

So sip it as the URL.

A

There's there's also other ways of getting this information, so here there are a bunch of shell commands that will do it.

A

Any questions, while this is the play.

A

Does the default Google cloud platform use HTTP for this, or is it HTTP only that's a good question: hey it's only HTTP! You cannot use HTTP right. I can look at what I've done for another cluster.

A

Looks like HTTP, so that should be HTTP.

A

This will also be given to you. Oh nice, that's finished by these shell commands. So if I want to run these I first need to sit my like active cluster, so that keep control knows which one to talk to you. So you can do this with this connect command.

A

A

If I run this and then I run this command.

A

The IP address should come up yep, oh so this is what gets copied into this just feel.

A

Certificates, so that's here or similarly that there's a command for their service token. So as these Doc's explain, we need to set up a couple of accounts on the cluster so that the GDK can interact with it and then because we'll be interacting using the tokens to authorize itself as these accounts, so the instruction says create a demo file with these contents.

A

So I have already done this.

A

As you can see here, there is this file and then I think it's a nice format yeah. So if I apply it to my currently assigned cluster, this will create a couple of accounts. So, as we can see here, it's created the accounts now I need to retrieve the token for the account.

A

This is my token, and this is used in this field here all right, so that's everything that we need to add an existing cluster to a project.

A

So here we go we're now connected to a cluster if I go back to the previous page now, instead of just the Add button, there's also this the list of the connected clusters.

A

For development purposes enough, and just like just with the default settings and and whatnot so I'll go ahead and store helm till the helm. Teller is there to install some of these other services and in particular, we want to install prometheus because Prometheus is going to provide us with metrics data about the cluster itself.

A

Yes, so they'll take a minute, or so once that's done once this is done, we'll be uninstalled Prometheus, it's done, I mean then once prometheus is installed, we'll be able to see these see the cluster health charts. So this is kind of the first objective that that you want to get to when you're like sitting everything up as you want to.

A

You want to see the cluster health charts, and then you know that at least the cluster is running and for some development tasks like this will be enough if you're, just if all you're doing, is multiplying their cluster health charts.

A

Once yeah once this is installed, it also will take a few minutes before it gets data I think by default, their polls once every minute, so I will show a cluster that I've prepared earlier there we go so yeah. This is what you'll see I'll just refresh the page, so you can see it. Loading is off.

A

This is what you'll see once the pasta has been connected for a while, and it's got some some data I believe that it is also a way to preload it would data, but for basic development like getting your environment set up for the first time. It's not required yeah. So you can see these charts there yeah, they just show kind of the memory usage and the CPU usage of of the class, the entire cluster.

A

If we want to look at like the memory usage or CPU of the specific node, that's when we get to the metrics dashboard, so that'll come next.

A

Let's have a look at how our original costs are doing.

A

Looks like it cool there. We go so it's loaded, but there's no data yet so looks like we have a single data point cool. So the next thing is now that we have cluster. We want to deploy something to it.

A

So for that we need get labor runner so get low Burana. Well, it was basically like a separate application that can deploy things and it gets talked to by the CI CD scripts.

A

So for this, if we look at pipelines, you can see there's no pipelines or no environments, but if we were to run a pipeline it would not complete unless there was a runner available so run it runners execute pipelines. So if this is your first time sitting up the environment, you probably won't have a runner available.

A

It is possible to install as we go back to my cluster, to install a runner on the cluster, and this is this is generally like the preferable approach, except for one thing, so the runner needs to be able to access your computer or it needs to be able to access your your GDK. Basically, the get lab instance.

A

The problem with that. If you're running the GDK locally is it would need to be able to access your local computer from the kubernetes cluster. Since this this install button, installs it inside become an english cluster, so gke would need to be able to see your computer so you're. Essentially, your computer would need to be available to the internet.

A

There is a way around us, so there's a there's, a proxy that get lab employees can exist. I think you need to open that the instructions are in a document somewhere this one, how to run it.

A

This isn't a GDP see it up.

A

No, actually, it's not the instructions and the auto develops how to.

A

Well, let's see if.

A

They may find these I'm using a not the browser that I normally use.

A

How to order devups.

A

So this this kind of details to the proxy.

A

So if you, if you want to test, if you have a reason to be testing the runner and it needs to be in kubernetes- and you have a local DBK, you can use this- you can go through this to to get them to talk to each other, but for today and for like initial developments, it up. It's also something that you don't necessarily need. So I'm gonna show how to get it running with a local runner.

A

So heaven get labour honor running on your local machine. It solves the problem because it doesn't have to access something over the internet just accesses the GDK from the same within the same network.

A

So the get labor on our docks. Explain this I'm not going to go through this whole thing, I'm just going to show how to register a runner. So if we go to our so assuming you have this assumes you have get my runner installed.

A

The install instructions are in the docks and they like the platform to pin it so like whatever platform you need. The other prerequisite for this step is that you have docker running because we're going to do the docker executors install.

A

Yes, so one thing you need to do is to because the runner talks to your instance needs to know what IP it is, so we're going to go and get the IP address of the computer.

A

So within my local network, this is my IP and then two to configure this, where you have to go to the get lady ml, which is in the root directory of the get lip, then go down to get lab settings, get lab host and set assess the current IP address, and once we do this and then restart the GDK, it will now.

A

It will now run what's what like, with the knowledge of its IP address, rather than like localhost. For some reason, localhost doesn't work with the runner I'm, actually not sure why I.

A

Don't know if it explained it's the reason here, but it it just says you need. You need to have your IP address so, like you can see, I'm accessing I was accessing my gbk from the IP. So if that works, then the IP is configured correctly and I've got a couple of run-ins already but I'm going to show how to add another one.

A

Yeah I just wait for the gbk to start.

A

Cool looks like it's working, so we're gonna go, get live rather so. You'll have access to this once you've installed, get laid runner and then register to register a new runner I'm going to use this URL for the corner. Url. This token.

A

Whoa cool doc around the demo.

A

If you add tags that will mean it will only run for jobs with those tags, not something that we're we're gonna use at the moment and then I'm gonna specify docker and.

A

Just pick a pick, a docker image.

A

A

So we'll see this sharpen the list now it is any of a contacted. Our GDK.

A

It will do so at some point, but just to kind of faucet I'm gonna run to get a runner verify, which gives all of the runners to try and contact to attempt to contact their coordinator.

A

So I'll remove these other ones so that we can make sure that we run with our fresh, fresh runner.

A

I have a bunch of old runners that point to random IP addresses at Antony's anymore, so this does cause a bit of noise.

A

I haven't found a good way to clear out the the registration list, but if you find one, let me know.

A

So what we're looking for is the a particular runner the TA TMP did show up.

A

Seems to do this one at a time as well, which is annoying because it means all of these failures are going to slow it down. So.

A

I can show what we're gonna do next anyway, so we're not waiting for that.

A

So I've got my my demo of my project and my question: oh yeah looks like a run that is alive, so we have this project this site, although develops fork so now we're going to try and run a pipeline.

A

A

We can see that this is in contact now, so that's good, so that that is going to run in the meantime, I'm going to show what happens once it has run in my sample project, so the environments page for this right now the project that we've just created this MD no deployments yet so this is one that has had a successful pipeline run. It's the same project, yeah just a clone, so we can look it once it's deployed.

A

We can now look at the monitoring page and this will contact the kubernetes cluster that we hooked up earlier. That has production deployed onto it now and it will gather some metrics so by default, you'll see these system metrics. So this is what you'll get with like the default basics unit. There are also some response, metrics, which are enabled by.

A

By switching on ingress.

A

So if you install ingress, you get a few extra tarts by default. Oh yeah, this is so. This is the matrix page, and this is more or less where you need to get to to be able to start monitor development, some things you can do with this page. You can create alerts, so you should be able to test that out now and you should be able to add your own metrics using pop QL.

A

Yeah, let's see if that's environment perhaps being deployed no still hasn't deployed but uh yeah. So let's start to finish how to go from a GD k to having matrix chats in equations.

B

For creating the like cluster rule, binding files and like setting up your cluster I know that requires your Google account to have. Oh, yes,.

A

B

Kubernetes engine admin, permissions, yeah.

A

But that's an important point: yeah thanks for mentioning it yeah. So that's that's noted, and these install Docs. You need a specific permission.

A

So this way to do to do that is via an access request.

A

I'm, just thinking if there's something I can link a little so after this demo I'm going to transfer what I said basically to like a Google Doc, not a Google Doc like a a markdown readme somewhere so in there I'll make sure to link a sample access request so that you can get the same access.

A

Also, we should like monitor. Team members should, in theory, be given access to this cluster I, don't know exactly what permissions we are given, but I know for myself. I didn't have to request access to this, so at least for monitor team members. It should be by default that you have access at least for this cluster, but yeah thanks for me.

A

That is something that that's important and would be a pretty frustrating roadblock. He didn't know I.

B

Think we're running up on time here go.

A

So thanks very much. There's awesome, no worries yeah. Hopefully that was a good good explanation and yeah, we'll upload it and all I had at the companion, some companion Docs, so that there's a text format for this as well: cool thanks hero.