Cloud Native Computing Foundation GitOpsCon NA 2022, 15 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: ⚡Lightning Talk: Why Do We Do This? the Heart of GitOps - Leigh Capili, VMware

Description

Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon Europe in Amsterdam, The Netherlands from April 17-21, 2023. Learn more at https://kubecon.io The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

⚡Lightning Talk: Why Do We Do This? the Heart of GitOps - Leigh Capili, VMware

Much toil and work is wasted! in the pursuit of glory, the sacrifice of mission, and the hosting... of web applications. In a computable world of information, there are often many valid solutions to the same problem, and naturally, there are a lot of ways to effectively deliver software, but why is GitOps a good way to do that? Is GitOps really just a trendy name for Continuous Delivery, or is there something more here? GitOps has a heart. There is a full-bodied reason for why this community beats to the rhythms and habits begged for by it's tools. This talk will break that down and speak to the question, "Why do we do this?"

A

uh Thank you so much everyone for coming. This is really exciting. It's honestly, an honor anytime. We have a get Ops specific event because I think I remember when this was not a community. It was a blog post right as all good communities start out as right and um just make sure that I will account for time here. So you know we had a couple, get UPS cons, a couple of get UPS uh days as well, and now we're sitting here.

A

We got two tracks: two tracks of lightning talks, two tracks of awesome presentations and so I'm sure that we've all been sitting around because we're cool practitioners hanging out in the infrastructure doing all the things we got the super glue out. Some of it got stuck on our hands. It's okay! We just try to get some sleep at night until we get paged right right, but uh yeah. You know, my name is Lee capili I'm into some things, snowboards I'm, Filipino, Etc, come come chat with me out there.

A

If you're into any of those things like being Filipino and uh and we're all excited about, get Ops right. We got whole communities forming new, tooling this. This is dope, but uh not everyone is always so optimistic right.

A

uh In fact, some people maybe make some spicy takes on the internet. Saying oh, you know those kid UPS people do they really know what they're doing haven't we been just doing continuous deployment for a long time and we get up this continuous deployment. You know continuously making sure that the cluster is in the way that it should be the desired state. This is not new stuff I mean you know, I, remember having controlled repos with puppet and all that stuff. This is devops all right, continuous deployment.

A

It's just one of the things that we learned from these brilliant people. Who've been around for more than a decade right, and so is this just marketing hype you know: um did we did we write some software uh when we really should have had no code?

A

uh Why do we do this? Why? Why am I still excited about this? I mean it's been years for me, man, I've I, don't know so.

A

um The first thing I want to talk about is that devops comes from paint right. Remember when we had the super glue stuck on our hands, sometimes it feels like that when we're using our tools, sometimes we're just trying to get the job done right. Why do we do what we do?

A

We got our apps deployed on their servers and it's not because we want to have good kpis or when I have that fresh satisfied feeling of logging into a server and, like you see that uptime dude that thing's been up for like two years and then you go to immutable infrastructure and you're like man, my uptime never is over three minutes, you know. So it's it's about doing the work with each other right.

A

The pain comes from the fact that we invite complexity when we try to solve hard problems, and let me tell you when you're, using a bunch of computers- and you know whatever the clouds or whatever co-locations you've got it's kind of frustrating, sometimes because it seems like you're, not the only person who wants to use the computer. You got to share it with your teams and they want to go and do all these things and you've got to go patch. The servers, but they want to go, update the apps.

A

You know they want to deploy all the time or maybe they don't want to deploy all the time because it breaks all the time and then you're calling them and debugging on Thursday, that's not fun, and so, as a community. We got kind of good at this right. Let's say hey: we can do better. Let's build some tools, let's figure out the right patterns.

A

Devops devops days got the devops email list right: puppet Chef ansible. We got some config management going some smart people who invented terraform, now I bunch of have like a bunch of this, like Json, looking DSL in my git repos, it's great. It's tying together infrastructure apis that were never meant to be glued together in a formal way.

A

You know it's true, there's nothing formal about our jobs right and then after we got good at that and stuff sucked a little bit less. Then the orchestration wars started and we're like. Oh I, got a new thing to fight for a hill to die on mesos kubernetes mine was the docker swarm.

A

Hill I I actually did die on that Hill, um so it was kind of a sad thing, because that one was a really nice one to use, but the darling of this ecosystem right after we had the problem of application sprawl and then we wanted to try and Wrangle it and have good dashboards and tools to keep everything all together.

A

We got kubernetes right. This whole conference is basically about kubernetes, but we call it Cloud native cup. um Kubernetes is a very special machine right. It I want to talk about the machine of kubernetes, because it's special it's different from what's come before it.

A

Kubernetes is a fast computer and the thing that it computes and the way that you program it is a little bit different than how we used to program things like puppet, which was kind of this hacked and Bluetooth thing together with lots of machines. Kubernetes is a distributed computer and the way that it gets reconciled. Is we broke up the reconciler instead of having one reconciler on each machine that does everything we're like? Oh hey?

A

Maybe we should have a couple of different control loops that can run anywhere they'll each get a piece of what needs to be done and we'll use a little bit of Promise Theory. These controllers will collaborate and they'll all try to take the piece of the puzzle and tell you if the promise is being kept, and then we wrote a little bit of really really sweet.

A

Grpc Magic called the Informer, which is both a pull Loop, but also a subscription that is reactive and so when stuff changes in the computer stuff that has no business being reconciled as fast as it does I mean. This has never happened in my puppet data center of 2000 nodes right, I, just update the config map, and now it's all over on 200 machines and all the pods are rolling and they're like what is happening. This is like this is only one puppet run, but kubernetes does not like that.

A

It's reactive, it's fast and is hooked up to a consistent state store, which is the thing that everybody wanted right. They were like deploying console you got like Vault over here. Maybe, let's like store a distributed lock, you don't got to do all that complicated stuff anymore. It's in the computer, one computer, and so this thing is pretty fancy mechanistically. It's already like I mean you guys remember when smartphones had like plastic keyboards. This thing is like way different than that: it's a way better computer. This is a better computer. Now, what did we do?

A

What do we do with it? Well, we used apps and infrastructure on it, and then we extend it except the landscape. Doesn't look like that anymore. It looks like this, so you can do all the things with the computer, which means you have all the people who want to do all the things that's pretty dope.

A

So you're, like man I, want to get me. Some kubernetes there's a fast computer with a new programming model about promises, and it can do all these things. It looks like this now I got a bunch of clusters. I got a bunch of people to satisfy and these things are not good at collaboration, because that's not what it's for you see. This thing is the Ferrari. It's the thing that gets your job done, but what about the race team?

A

What about all of the engineering that makes the Ferrari go around the track? Well, it's not a forum, it's not a collab tool. You can't Version Control inside a kubernetes, but you can do that outside and we can take our programming model to our place of collaboration, and so this is why we do this right. We can answer these hard questions. We can say what does our app need? How much of it who does it depend on you got people asking what app? What app are you talking about?

A

I got 10 of those in a folder over there who owns that? What should they have access to? How can we enable them? How can we onboard other teams like them? Do they even know what they're doing, because it looks pretty sketchy in their folder? How do we keep them safe? How do we know? What do we do?

A

Did we release a manager is wondering that somewhere, some CIO what changed last week, how come we're, having problems? What do we do last year, you see when we ask these questions.

A

We can use our tools to describe ourselves specifically and our tools, the way that we work with each other shape our habits. Our habits are individual and they're. Also organizational. We have communal habits, shaped by the way that we work with each other and sufficiently advanced tools when everybody can use them, and everybody has a reason to use them when you have a computer that can do so many things and it's so good at it stuff actually fundamentally changes. This is innovative.

A

This is changing the way that our industry works, because now we talk to each other, we build community, we have ways of working that translate across the industry and so when you're thinking about when your software becomes Legacy, maybe you could ask yourself what sorts of habits do we need to do today so that in five years Legacy is not a bad word, or maybe it's a good word right, and so we got kubernetes is this computer that can make promises for us, but if we're talking about get UPS get Ops is about making promises to each other.

A