GitLab Configure Stage, 20 Apr 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Configure Design Sprint - Homework Assignment 2 - Kenny Johnston

Description

https://gitlab.com/gitlab-org/configure/general/-/issues/50#pre-workshop-tasks

Lightning Walkthrough video

Each participant is asked to prepare a short video (5 minutes max) to talk about Kubernetes from their own perspective (this will be a video where you record your screen and audio to be presented back to the team async). These will act as inspiration for the team as the Sprint progresses through the design concepting phases. Participants will use them as a resource, checking back with them as necessary… inspiration! Some topics that you can cover in the video are (deadline explained in Part 3):

- (Required) Spend 1 minute talking about the answers you provided to the expert interview questions (below).
- Problems that you see in the current Kubernetes GitLab integration vision.
- The state of the technology and what is happening in the field and how you envision the future.
- Companies or implementations that inspire you (Kubernetes specific).
- Demo walkthrough that goes through a competitor experience and functionality you find interesting or spot on.

A

Hey my name is Kenny Johnston I am doing the configure design sprint homework assignment number two. Let me go through the questions one by one, so the first one is spend a minute. Talking about your answers to the expert interview, questions I pulled this up. First question was: can you describe in a few words what the product is and I said? It's kind of a beginner entry point for kubernetes and utilizing communities for your application? Was a product trying to solve I? Think it's trying to solve for communities is complicated and I.

A

Don't really know how to use it and I just kind of want to get started with my getting my app to production and quickly so kind of like moving quickly through the stages of a DevOps loop, all in one application, but not giving me kind of what I would want if I were doing a production type of deploy who's using the product currently and who would want to use it I think yeah, as I mentioned people who are just getting started, people who are trying to understand people who might want to be doing kind of demos or really small apps using honored about DevOps, primarily the having clusters at different levels seems to attract a different type of user, and so having group level clusters means that there might be some kind of overarching platform user using it.

A

But I don't think that our capability of configuring, your cluster, well suits that more sophisticated use case leaves problems. I think are that being able to configure your infrastructure configure that cluster in a more complex way and still utilize it? It seems to be one and four with no. What would the product look like people have talked to me about this before know that I have a specific tool that I always come back to you I'm blanking on it is.

A

I'm gonna forget about it.

A

It's called cantenna lens yeah, which is kind of like an ide for managing the configuration of your kubernetes cluster seems pretty useful. I think that actually, our work in IAC is probably more in line with a git lab' way of doing something similar.

A

Obviously, things I want to get out of the workshop I'd love to just kind of get a shared model to learn a lot I, don't not a community's expert by any stretch of the imagination, I'm norm, I some one of the profile who would be regularly interacting last in configuring, a kubernetes cluster so like learn like to get some shared understanding, I'd like to have some concrete steps of what we can do to start iteratively approaching some new kind of direction that we agree on okay, so that was part one of the homework assignment, which was one minute overview took more than that was three minutes problems you see in the current give lab integration vision, yeah I think it is that the integration vision and maybe the structure around get lab, managed apps is very confining and so to the extent that that becomes more flexible and that you can treat maybe kubernetes like another set of infrastructure as code the cluster itself, but manage that and are really a robust way.

A

I think would be pretty cool still. The aim I think we, if we focus on people who have yet to adopt kubernetes and still don't know much about it, easing still considering that journey of adopting kubernetes, easily I think is going to be critical state of the technology and was having a feel and how you envision. It I'm, not sure if I know the answer to that. I know that there are a lot of complexities to kubernetes and it's an evolving piece of open source technology.

A

I think that things that are cool to me are things that help make kubernetes more accessible and more manageable, and so you know I know, there's also a bunch of work in making communities more scalable and applicable for different types of applications. So there's a couple of different vectors there, companies or implementations that inspire you I, really like our CI cd-based Korea's, managed app in config process.

A

I, think that's really cool and really smart and a great way to utilize bunch of gitlab strengths, as I mentioned I like content all ends just in terms of the like UI experience that you get for understanding and managing a cluster.

A

That's something I forgot to mention: I I think that the one of the big points where complexity comes in to managing a cluster is when something went wrong and you're trying to figure out what went wrong and so I think some of our current capabilities, obscure more what's happening inside of the cluster, as opposed to being transparent and I can make understanding like.

A

Why did my application not deploy or what's going on with this error, or how do I get my application back online more difficult because of the the obscurity we've done in service of trying to make it easier to use demo walk through? That goes to a competitor experience and functionally I. Are you find interesting? I? Don't I have a demo walkthrough set up I yeah no theme, maybe I could go through our current experience. I was just looking at it a second ago for a different purpose.

A

I'll just show my screen, and maybe I can talk through a little bit about.

A

Okay, this is an attached cluster or a group level. Cluster I love how this has evolved from having the separate environments tab to now having health and applications. I think that's really cool in a cool separation.

A

I think the applications thing starting to worry me because is starting to get long and not giving a lot of description to our users about what and why they might use these applications and instead of as opposed to just giving them the list, and maybe some of these applications need to go to where you would actually use them like when you're doing CI ACD settings, you install install the runner there rather than just showing the install button process here, and this can be more about what applications are running as opposed to adding new ones.

A

I love, as I mentioned, the cluster management project process. I think that's really ingenious in the way that we connect projects that are meant to be either the configuration or the definition of infrastructure and then the configuration of whether the definition of this, the application that are running and kind of like connect. The two I think, if you play on that pattern, it can get pretty robust and useful for a bunch of different use cases so think of like I know this.

A

This application lands on this infrastructure for this environment, but those are connected by projects or this application. I know is landing on this cluster and I'm connected to Q, and so there's a lot of things that give up can like natively, do once you've identified that connection for you as a developer. So you don't have to always manage that complexity yourself between projects. So what else I I have recently had this realization that I think, if you look at most of these managed cluster providers, as obviously Google is farther ahead than others. I think.

A

But Google has a really interesting setting for clusters that I noticed recently, which was the ability to stay. Keep your cluster upgraded. There's been a lot of like oh well, who's going to manage the state of my cluster, but now, when you create a new cluster in gke.

A

It gives you these release channels, so it's always going to keep you upgraded and if you choose a release channel.

A

And I get back to you the more complicated.

A

Oh, maybe something okay, yeah pools.

A

There's a place where I could choose when I wanted it to update.

A

Maybe it's automation, oh yeah, like I, could say: I want maintenance, windows and maintenance exclusions and here's when you're, it's okay for you to or not to upgrade my cluster.

A

This is all happening at the management of cloud service provider, which today is a bunch of complexity that either infrastructure teams or their tooling have to handle so I think it's pretty evidence to me that this is going to be handled by more and more, by kind of, like ongoing management of the health of the cluster is managed by the infrastructure provider, and so that developers and platform teams are more focused on the health of the application running on that cluster, not necessarily the helps with the cluster itself.

A

So that's an interesting pattern that I've noticed I'm Way over time, I'm almost at ten minutes. Hopefully this answers the questions. How I was intended to answer things aren't looking forward to the design of Sprint, take care.