GitLab Configure Stage, 7 Jul 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: GitLab Kubernetes Agent CI tunnel demo

Description

The CI/CD Tunnel enables users to access Kubernetes clusters from GitLab CI/CD jobs even if there is no network connectivity between GitLab Runner and a cluster. GitLab Runner does not have to be running in the same cluster.

In the current iteration, only CI/CD jobs in the Configuration project are able to access one of the configured agents.

https://docs.gitlab.com/ee/user/clusters/agent/ci_cd_tunnel.html

A

uh Hello, okay from the configure team here, uh I'd like to demo the site tunnel functionality. So this is my personal project for the demo purposes, and here we have an agent with basically an empty configuration. Just extra login is enabled and.

A

You have the following: ci configuration file which so the agent id is 1 142. Then we create the configuration file for cubesat.

A

It will be talking to gitlab gitlab's cast installation on this address using the ci job token for authentication and agent id this number, and then it should just get the version of cuba, cto and the server and then print all the ports. You know many spaces, so I have just run this.

A

I've just turned the side job which does that, and it just fails. This is to show that it's not working uh when there is no agent. So this is not working because I don't have an agent connected, and this is quick ctrl version with extra debugging and that extra debugging video shows that it tries to talk to this address and it couldn't like request timed out 32 seconds now and keep ctrl just prints this factories.

A

So I did that on purpose. This is what happens if there is no agent, but let's deploy the agent. So in the same project I have this flap agent directory. This is a tpt package.

A

The official our official kpt package to install the agent, but I cloned it, configured it to work with gitlab.comcast and my secret and so on, and I checked in here for convenience.

A

So that's what we will be using to deploy uh the agent into my local cluster, which is running in docker. So I usually kind events and docker and.

A

A

This is what I have.

A

Okay, so to deploy this package we will be using command. It will run customize to build the cluster directory.

A

This directory cluster- um this is the directory, I'm sure, and it will pipe that to publicity apply and it will apply to my local cluster and that will be creating the namespace x2.

A

To which it will deploy the agent, let's see if it's done, that already yep agent is there it's running.

A

um That here actually.

A

This is my agent one and it's last used 30 seconds ago. This is my talking, so it's working, okay, let's leave it here with block tailing.

A

Now I don't like that job name deploy that's because it's not deploying anything to for the demo purposes and to trigger the job. Let's edit the yaml file and rename the report demo.

A

That and this should trigger this hijab.

A

Like this yep called demo.

A

It's taking us a bit of time to download the.

A

Image, okay, it's like logs are behind because kubernetes agent is already printing that it's handled a connection and we are also getting resource exhausted. um I think that's because we are.

A

We have configured the rate limiting on the server side quite conservatively, so it turned out that uh those two commands I showed youtube ctrl version and get ports in online spaces. These two commands make 125 api calls, so obviously keep still makes them as fast as it can. So we are throttling it. That's why uh logs are taking a while to appear as if it's just working and it's been throttled, so we need to tune the throttle little bit uh to integrate limits in a little bit to get less protein anyway. It worked.

A

Fine um here is the version.

A

So we get a lot of details here, because I specified your boss 10 and we got the reply in 500 milliseconds from the casket.com, and the server version is what I have got away: 121 0.1 and these are the ports and the extreme space where I just deployed my um what's edge.

A

That's probably thank you.