>>and hello, we're back. I've got my panel and we are doing things real time here. So sorry for the delay a few minutes late. So the way let's talk about things, the reason we're here and we're going around the room and introduce everybody. Got three special guests here. I got my evil or my john and the normal And we're going to talk about get ops I called it future office just because I want to think about what's the next thing for that at the end, we're gonna talk about what our ideas for what's next for getups, right? Um, because we're all starting to just get into get ups now. But of course a lot of us are always thinking about what's next? What's better? How can we make this thing better? So we're going to take your questions. That's the reason we're here, is to take your questions and answer them. Or at least the best we can for the next hour. And all right, so let's go around the room and introduce yourself. My name is Brett. I am streaming from Brett from that. From Brett. From Virginia Beach in Virginia beach, Virginia, United States. Um, and I talk about things on the internet, I sell courses on you, to me that talk about Docker and kubernetes Ive or introduce yourself. >>How's it going? Everyone, I'm a software engineer at axel Springer, currently based in Berlin and I happen to be Brett Brett's teaching assistant. >>All right, that's right. We're in, we're in our courses together almost every day. Mm john >>hey everyone, my name is john Harris, I used to work at Dhaka um, I now work at VM ware is a star field engineer. Um, so yeah, >>and normal >>awesome by the way, you are streaming from Brett Brett, >>I answered from breath to breath. >>Um I'm normal method. I'm a distinguished engineer with booz allen and I'm also a doctor captain and it's good to see either in person and it's good to see you again john it's been a little while. >>It has the pre covid times, right? You're up here in Seattle. >>Yeah. It feels, it feels like an eternity ago. >>Yeah, john shirt looks red and reminds me of the Austin T shirt. So I was like, yeah, so we all, we all have like this old limited edition doctor on E. >>T. That's a, that's a classic. >>Yeah, I scored that one last year. Sometimes with these old conference church, you have to like go into people's closets. I'm not saying I did that. Um, but you know, you have to go steal stuff, you to find ways to get the swag >>post post covid. If you ever come to my place, I'm going to have to lock the closets. That >>that's right, That's right. >>So the second I think it was the second floor of the doctor HQ in SAn Francisco was where they kept all the T shirts, just boxes and boxes and boxes floor to ceiling. So every time I went to HQ you just you just as many as you can fit in your luggage. I think I have about 10 of these. You >>bring an extra piece of luggage just for your your shirt shirt grab. Um All right, so I'm going to start scanning questions uh so that you don't have to you can you help you all are welcome to do that. And I'm going to start us off with the topic. Um So let's just define the parameters. Like we can talk about anything devops and here we can go down and plenty of rabbit holes. But the kind of, the goal here is to talk about get ups and get ups if you haven't heard about it is essentially uh using versioning systems like get like we've all been getting used to as developers to track your infrastructure changes, not just your code changes and then automate that with a bunch of tooling so that the robots take over. And essentially you have get as a central source of truth and then get log as a central source of history and then there's a bunch of magic little bits in the middle and then supposedly everything is wonderful. It's all automatic. The reality is is what it's often quite messy, quite tricky to get everything working. And uh the edges of this are not perfect. Um so it is a relatively new thing. It's probably three, maybe four years old as an official thing from. We've uh so we're gonna get into it and I'll let's go around the room and the same word we did before and um not to push on that, put you on the spot or anything. But what is, what is one of the things you either like or either hate about getups um that you've enjoyed either using it or you know, whatever for me. I really, I really love that I can point people to a repo that basically is hopefully if they look at the log a tracking, simplistic tracking of what might have changed in that part of the world or the environment. I remember many years past where, you know, I've had executive or some mid level manager wants to see what the changes were or someone outside my team went to see what we just changed. It was okay, they need access to this system into that dashboard and that spreadsheet and then this thing and it was always so complicated and now in a world where if we're using get up orbit bucket or whatever where you can just say, hey go look at that repo if there was three commits today, probably three changes happened. That's I love that particular part about it. Of course it's always more complicated than that. But um Ive or I know you've been getting into this stuff recently. So um any thoughts? Yeah, I think >>my favorite part about get ops is >>reproducibility. Um >>you know the ability to just test something and get it up and running >>and then just tear it down. >>Uh not >>being worried that how did I configure it the first time? I think that's my favorite part about >>it. I'm changing your background as we do this. >>I was going to say, did you just do it get ups pushed to like change his >>background, just a dialogue that different for that green screen equals false? Uh Change the background. Yeah, I mean, um and I mean I think last year was really my first year of actually using it on anything significant, like a real project. Um so I'm still, I still feel like I'm very new to john you anything. >>Yeah, it's weird getups is that thing which kind of crystallizes maybe better than anything else, the grizzled veteran life cycle of emotions with the technology because I think it's easy to get super excited about something new. And when I first looked into get up, so I think this is even before it was probably called getups, we were looking at like how to use guest source of truth, like everything sounds great, right? You're like, wait, get everyone knows, get gets the source of truth, There's a load of robust tooling. This just makes a sense. If everything dies, we can just apply the get again, that would be great. Um and then you go through like the trough of despair, right? We're like, oh no, none of this works. The application is super stateless if this doesn't work and what do we do with secrets and how do we do this? Like how do we get people access in the right place and then you realize everything is terrible again and then everything it equalizes and you're kind of, I think, you know, it sounds great on paper and they were absolutely fantastic things about it, but I think just having that measured approach to it, like it's, you know, I think when you put it best in the beginning where you do a and then there's a magic and then you get C. Right, like it's the magic, which is >>the magic is the mystery, >>right? >>Magic can be good and bad and in text so >>very much so yeah, so um concurrence with with john and ever uh in terms of what I like about it is the potential to apply it to moving security to left and getting closer to a more stable infrastructures code with respect to the whole entire environment. Um And uh and that reconciliation loop, it reminds me of what, what is old is new again? Right? Well, quote unquote old um in terms of like chef and puppet and that the reconciliation loop applied in a in a more uh in a cleaner interface and and into the infrastructure that we're kind of used to already, once you start really digging into kubernetes what I don't like and just this is in concurrence with the other Panelist is it's relatively new. It has um, so it has a learning curve and it's still being, you know, it's a very active um environment and community and that means that things are changing and constantly and there's like new ways and new patterns as people are exploring how to use it. And I think that trough of despair is typically figuring out incrementally what it actually is doing for you and what it's not going to solve for you, right, john, so like that's that trough of despair for a bit and then you realize, okay, this is where it fits potentially in my architecture and like anything, you have to make that trade off and you have to make that decision and accept the trade offs for that. But I think it has a lot of promise for, for compliance and security and all that good stuff. >>Yeah. It's like it's like the potentials, there's still a lot more potential than there is uh reality right now. I think it's like I feel like we're very early days and the idea of especially when you start getting into tooling that doesn't appreciate getups like you're using to get up to and use something else and that tool has no awareness of the concept so it doesn't flow well with all of the things you're trying to do and get um uh things that aren't state based and all that. So this is going to lead me to our first question from Camden asking dumb questions by the way. No dumb questions here. Um How is get apps? Not just another name for C. D. Anybody want to take that as an answer as a question. How is get up is not just another name for C. D. I have things but we can talk about it. I >>feel like we need victor foster kids. Yeah, sure you would have opinions. Yeah, >>I think it's a very yeah. One person replied said it's a very specific it's an opinionated version of cd. That's a great that's a great answer like that. Yeah. >>It's like an implement. Its it's an implementation of deployment if you want it if you want to use it for that. All right. I realize now it's kind of hard in terms of a physical panel and a virtual panel to figure out who on the panel is gonna, you know, ready to jump in to answer a question. But I'll take it. So um I'll um I'll do my best inner victor and say, you know, it's it's an implementation of C. D. And it's it's a choice right? It's one can just still do docker build and darker pushes and doctor pulls and that's fine. Or use other technologies to deploy containers and pods and change your, your kubernetes infrastructure. But get apps is a different implementation, a different method of doing that same thing at the end of the day. Yeah, >>I like it. I like >>it and I think that goes back to your point about, you know, it's kind of early days still, I think to me what I like about getups in that respect is it's nice to see kubernetes become a platform where people are experimenting with different ways of doing things, right? And so I think that encourages like lots of different patterns and overall that's going to be a good thing for the community because then more, you know, and not everything needs to settle in terms of only one way of doing things, but a lot of different ways of doing things helps people fit, you know, the tooling to their needs, or helps fit kubernetes to their needs, etcetera. Yeah, >>um I agree with that, the, so I'm gonna, since we're getting a load of good questions, so um one of the, one of the, one of the, I want to add to that real quick that one of the uh from the, we've people themselves, because I've had some on the show and one of things that I look at it is distinguishing is with continuous deployment tools, I sort of think that it's almost like previous generation and uh continuous deployment tools can be anything like we would consider Jenkins cd, right, if you if you had an association to a server and do a doctor pull and you know, dr up or dr composed up rather, or if it did a cube control apply uh from you know inside an ssh tunnel or something like that was considered considered C. D. Well get ops is much more rigid I think in terms of um you you need to apply, you have a specific repo that's all about your deployments and because of what tool you're using and that one your commit to a specific repo or in a specific branch that repo depends on how you're setting it up. That is what kicks off a workflow. And then secondly there's an understanding of state. So a lot of these tools now I have uh reconciliation where they they look at the cluster and if things are changing they will actually go back and to get and the robots will take over and will commit that. Hey this thing has changed um and you maybe you human didn't change it, something else might have changed it. So I think that's where getups is approaching it, is that ah we we need to we need to consider more than just a couple of commands that be runnin in a script. Like there needs to be more than that for a getups repo to happen anyway, that's just kind of the the take back to take away I took from a previous conversation with some people um >>we've I don't think that lost, its the last piece is really important, right? I think like for me, C d like Ci cd, they're more philosophical ideas, write a set of principles, right? Like getting an idea or a code change to environments promoting it. It's very kind of pipeline driven um and it's very imperative driven, right? Like our existing CD tools are a lot of the ways that people think about Cd, it would be triggered by an event, maybe a code push and then these other things are happening in sequence until they either fail or pass, right? And then we're done. Getups is very much sitting on the, you know, the reconciliation side, it's changing to a pull based model of reconciliation, right? Like it's very declarative, it's just looking at the state and it's automatically pulling changes when they happen, rather than this imperative trigger driven model. That's not to say that there aren't city tools which we're doing pull based or you can do pull based or get ups is doing anything creatively revolutionary here, but I think that's one of the main things that the ideas that are being introduced into those, like existing C kind of tools and pipelines, um certainly the pull based model and the reconciliation model, which, you know, has a lot in common with kubernetes and how those kind of controllers work, but I think that's the key idea. Yeah. >>Um This is a pretty specific one Tory asks, does anyone have opinions about get ops in a mono repo this is like this is getting into religion a little bit. How many repos are too many repose? How um any thoughts on that? Anyone before I rant, >>go >>for it, go for it? >>Yeah. How I'm using it right now in a monitor repo uh So I'm using GIT hub. Right, so you have what? The workflow and then inside a workflow? Yeah, mo file, I'll >>track the >>actual changes to the workflow itself, as well as a folder, which is basically some sort of service in Amman Arepa, so if any of those things changes, it'll trigger the actual pipeline to run. So that's like the simplest thing that I could figure out how to, you know, get it set up using um get hubs, uh workflow path future. Yeah. And it's worked for me for writing, you know? That's Yeah. >>Yeah, the a lot of these things too, like the mono repo discussion will, it's very tool specific. Each tool has various levels of support for branch branching and different repos and subdirectories are are looking at the defense and to see if there's changes in that specific directory. Yeah. Sorry, um john you're going to say something, >>I was just going to say, I've never really done it, but I imagine the same kind of downsides of mono repo to multiple report would exist there. I mean, you've got the blast radius issues, you've got, you know, how big is the mono repo? Do we have to pull does the tool have to pull that or cashier every time it needs to determine def so what is the support for being able to just look at directories versus you know, I think we can get way down into a deeper conversation. Maybe we'll save it for later on in the conversation about what we're doing. Get up, how do we structure our get reposed? We have super granular repo per environment, Perper out reaper, per cluster repo per whatever or do we have directories per environment or branches per environment? How how is everything organized? I think it's you know, it's going to be one of those, there's never one size fits all. I'll give the class of consultant like it depends answer. Right? >>Yeah, for sure. It's very similar to the code struggle because it depends. >>Right? >>Uh Yeah, it's similar to the to the code problem of teams trying to figure out how many repose for their code. Should they micro service, should they? Semi micro service, macro service. Like I mean, you know because too many repose means you're doing a bunch of repo management, a bunch of changes on your local system, you're constantly get pulling all these different things and uh but if you have one big repo then it's it's a it's a huge monolithic thing that you usually have to deal with. Path based issues of tools that only need to look at a specific directory and um yeah, it's a it's a culture, I feel like yeah, like I keep going back to this, it's a culture thing. Does your what is your team prefer? What do you like? What um what's painful for everyone and who's what's the loudest pain that you need to deal with? Is it is it repo management? That's the pain um or is it uh you know, is that that everyone's in one place and it's really hard to keep too many cooks out of the kitchen, which is a mono repo problem, you know? Um How do we handle security? So this is a great one from Tory again. Another great question back to back. And that's the first time we've done that um security as it pertains to get up to anyone who can commit can change the infrastructure. Yes. >>Yes. So the tooling that you have for your GIT repo and the authentication, authorization and permissions that you apply to the GIT repo using a get server like GIT hub or get lab or whatever your flavor of the day is is going to be how security is handled with respect to changes in your get ups configuration repository. So um that is completely specific to your implementation of that or ones implementation of of how they're handling that. Get repositories that the get ups tooling is looking at. To reconcile changes with respect to the permissions of the for lack of better term robot itself. Right? They get up tooling like flux or Argosy. D Um one kid would would create a user or a service account or uh other kind of authentication measures to limit the permissions for that service account that the Gaddafi's tooling needs to be able to read the repose and and send commits etcetera. So that is well within the realm of what you have already for your for your get your get um repo. Yeah. >>Yeah. A related question is from a g what they like about get apps if done nicely for a newbie it's you can get stuff done easily if you what they dislike about it is when you have too many get repose it becomes just too complicated and I agree. Um was making a joke with a team the other week that you know the developer used to just make one commit and they would pass pass it on to a QA team that would then eventually emerging in the master. But they made the commits to these feature branches or whatever. But now they make a commit, they make a pR there for their code then they go make a PR in the helm chart to update the thing to do that and then they go make a PR in the get ups repeal for Argo. And so we talked about that they're probably like four or five P. R. Is just to get their code in the production. But we were talking about the negative of that but the reality was It's just five or 4 or five prs like it wasn't five different systems that had five different methodologies and tooling and that. So I looked at it I was like well yeah that's kind of a pain in the get sense but you're also dealing with one type. It's a repetitive action but it's it's the one thing I don't have to go to five different systems with five different ways of doing it. And once in the web and one's on the client wants a command line that I don't remember. Um Yeah so it's got pros and cons I think when you >>I think when you get to the scale where those kind of issues are a problem then you're probably at the scale where you can afford to invest some time into automation into that. Right? Like what I've when I've seen this in larger customers or larger organizations if there ever at that stage where okay apps are coming up all the time. You know, there's a 10 X 100 X developer to operations folks who may be creating get repose setting up permissions then that stuff gets automated, right? Like, you know, maybe ticket based systems or whatever. Developers say I need a new app. It templates things or more often using the same model, right of reconciliation and operators and the horrific abuse of cogs that we're seeing in the communities community right now. Um You know, developers can create a crd which just says, hey, I'm creating a new app is called app A and then a controller will pick up that app a definition. It will go create a get a repo Programmatically it will add the right definitely will look up and held up the developers and the permissions that need to be able to get to that repo it will create and template automatically some name space and the clusters that it needs in the environments that it needs, depending on, you know, some metadata it might read. So I think, you know, those are definite problems and they're definitely like a teething, growing pain thing. But once you get to that scale, you kind of need to step back and say, well look, we just need to invest in time into the operational aspect of this and automating this pain away, I think. Yeah, >>yeah. And that ultimately ends in Yeah. Custom tooling, which it's hard to avoid it at scale. I mean, there's there's two, there's almost two conversations here, right. There is what I call the Solo admin Solo devops, I bought that domain Solo devops dot com because, you know, whenever I'm talking to dr khan in the real world, it's like I asked people to raise hands, I don't know how we can raise hands here, but I would ask people to raise hands and see how many of you here are. The sole person responsible for deploying the app that your team makes and like a quarter of the room would raise their hand. So I call that solo devops like those, that person can't make all the custom tooling in the world. So they really need dr like solutions where it's opinionated, the workflow is sort of built in and they don't have to wrangle things together with a bunch of glue, you know, in other words bash. Um and so this kind of comes to a conversation uh starting this question from lee he's asking how do you combine get ops with ci cd, especially the continuous bit. How do you avoid having a human uh sort of the complaint the team I was working with has, how do you avoid a human editing and get committing for every single deploy? They've settled on customized templates and a script for routine updates. So as a seed for this conference, this question I'm gonna ask you all uh instead of that specific question cause it's a little open ended. Um Tell me whether you agree with this. I I kind of look at the image, the image artifact because the doctor image or container image in general is an artifact that I I view it that way and that thing going into the registry with the right label or right part of the label. Um That tag rather not the label but the tag that to me is like one of the great demarche points of, we're kind of done with Ci and we're now into the deployment phase and it doesn't necessarily mean the tooling is a clear cut there, but that artifact being shipped in a specific way or promoted as we sometimes say. Um what do you think? Does anyone have opinions on that? I don't even know if that's the right opinion to have so mhm. >>So um I think what you're, what you're getting at is that get ups, models can trigger off of different events um to trigger the reconciliation loop. And one way to do that is if the image, if it notices a image change in the registry, the other is if there's a commit event on a specific rebo and branch and it's up to, you are up to the person that's implementing their get ups model, what event to trigger there, that reconciliation loop off of, You can do both, you can do one or the other. It also depends on the Templeton engine that you're using on top of um on top of kubernetes, such as helm or um you know, the other ones that are out there or if you're not even doing that, then, you know straight. Yeah, mo um so it kind of just depends, but those are the typically the two options one has and a combination of of those to trigger that event. You can also just trigger it manually, right? You can go into the command line and force a a, you know, a really like a scan or a new reconciliation loop to occur. So it kind of just, I don't want to say this, but it depends on what you're trying to do and what makes sense in your pipeline. Right? So if you're if you're set up where you are tag, if you're doing it based off of image tags, then you probably want to use get ups in a way that you're using the image tags. Right. And the pattern that you've established there, if you're not really doing that and you're more around, like, different branches are mapped to different environments, then triggered off of the correct branch. And that's where the permissions also come into play. Where if you don't want someone to touch production and you've got your getups for your production cluster based off of like uh you know, a main branch, then whoever can push a change to that main branch has the authority to push that change to production. Right? So that's your authentication and permissions um system same for the registry itself. Right. So >>Yeah. Yeah. Sorry, anyone else have any thoughts on that? I was about to go to the next topic, >>I was going to say. I think certain tools dictate the approach, like, if you're using Argosy d it's I think I'm correct me if I'm wrong, but I think the only way to use it right now is just through image modification. Like, the manifest changes, it looks at a specific directory and anything changes then it will do its thing. And uh Synchronize the cost there with whatever's and get >>Yeah, flux has both. Yeah, and flux has both. So it it kind of depends. I think you can make our go do that too, but uh this is back to what we were saying in the beginning, uh you know, these things are changing, right? So that might be what it is right now in terms of triggering the reconciliation loops and get ups, tooling, but there might be other events in the future that might trigger it, and it's not completely stand alone because you still need you're tooling to do any kind of testing or whatever you have in terms of like the specific pipeline. So oftentimes you're bolting in getups into some other part of broader Cfd solution. That makes sense. Yeah, >>we've got a lot of questions about secrets or people that are asking about secrets. >>So my my tongue and cheek answered the secrets question was, what's the best practices for kubernetes? Secrets? That's the same thing for secrets with good apps? Uh getups is not last time I checked and last time I was running this stuff get ups is not has nothing to do with secrets in that sense. It's just there to get your stuff running on communities. So, um there's probably a really good session on secrets at dr concept. I >>would agree with you, I agree with you. Yeah, I mean, get off stools, I mean every every project of mine handles secrets differently. Uh huh. And I think I'm not sure if it was even when I was talking to but talking to someone recently that I'm very bullish on get up actions, I love get up actions, it's not great for deployments yet, but we do have this new thing and get hub environments, I think it's called. So it allows me at least the store secrets per environment, which it didn't have the concept of that before, which you know, if you if any of you running kubernetes out there, you typically end up when you start running kubernetes, you end up with more than one kubernetes, like you're going to end up with a lot of clusters at some point, at least many multiple, more than two. Um and so if you're trying to store secret somewhere, you do have and there's a discussion happening in chat right now where people are talking about um sealed secrets which if you haven't heard of that, go look that up and just be versed on what sealed secrets is because it's a it's a fantastic concept for how to store secrets in the public. Um I love it because I'm a big P. K. I nerd but um it's not the only way and it doesn't fit all models. So I have clients that use A W. S. Secrets because they're in A W. S. And then they just have to use the kubernetes external secret. But again like like like normal sand, you know, it's that doesn't really affect get ops, get ops is just applying whatever helm charts or jahmal or images that you're, you're you're deploying, get off. It was more about the approach of when the changes happen and whether it's a push or pull model like we're talking about and you know, >>I would say there's a bunch of prerequisites to get ups secrets being one of them because the risk of you putting a secret into your git repo if you haven't figured out your community secrets architecture and start diving into getups is high and removing secrets from get repose is you know, could be its own industry, right. It's >>a thing, >>how do >>I hide this? How do I obscure this commit that's already now on a dozen machines. >>So there are some prerequisites in terms of when you're ready to adopt get up. So I think is the right way of saying the answer to that secrets being one of them. >>I think the secrets was the thing that made me, you know, like two or three years ago made me kind of see the ah ha moment when it came to get ups which, which was that the premier thing that everyone used to say about get up about why it was great. Was its the single source of truth. There's no state anywhere else. You just need to look at git. Um and then secrets may be realized along with a bunch of other things down the line that is not true and will never be true. So as soon as you can lose the dogmatism about everything is going to be and get it's fantastic. As long as you've understood everything is not going to get. There are things which will absolutely never be and get some tools just don't deal with that. They need to earn their own state, especially in communities, some controls on their own state. You know, cuz sealed secrets and and other projects like SOps and I think there are two or three others. That's a great way of dealing with secrets if you want to keep them in get. But you know, projects like vault more kind of like what I would say, production grade secret strategies. Right? And if you're in AWS or a cloud, you're more likely to be using their secrets. Your secret policy is maybe not dictated by you in large organizations might be dictated by CSO or security or Great. Like I think once if you, if you're trying to adopt getups or you're thinking about it, get the dogmatism of get as a single point of truth out of your mind and think about getups more as a philosophy and a set of best practice principles, then you will be in much better stead, >>right? Yeah. >>People are asking more questions in chat like infrastructure as code plus C d essentially get ups or C I rather, um, these are all great questions and a part of the debate, I'm actually just going to throw up on screen. I'm gonna put this in chat, but this is, this is to me the source, Right? So we worked with when they coined the term. We, a lot of us have been trying to get, if we talk about the history for a minute and then tell me if I'm getting this right. Um, a lot of us were trying to automate all these different parts of the puzzle, but a lot of them, they, some things might have been infrastructure as code. Some things weren't, some things were sort of like settings is coded, like you're going to Jenkins and type in secrets and settings or type in a certain thing in the settings of Jenkins and then that it wasn't really in get and so what we was trying to go for was a way to have almost like eventually a two way state understanding where get might change your infrastructure but then your infrastructure might also change and needs to be reflected in the get if the get is trying to be the single source of truth. Um and like you're saying the reality is that you're never gonna have one repo that has all of your infrastructure in it, like you would have to have, you have to have all your terra form, anything else you're spinning up. Right. Um but anyway, I'm gonna put this link in chat. So this guide actually, uh one of things they talk about is what it's not, so it's, it's kind of great to read through the different requirements and like what I was saying well ago um mhm. Having having ci having infrastructure as code and then trying a little bit of continuous deployment out, it's probably a prerequisite. Forget ops so it's hard to just jump into that when you don't already have infrastructure as code because a machine doing stuff on your behalf, it means that you have to have things documented and somewhere and get repo but let me put this in the in the >>chitty chat, I would like to know if the other panelists agree, but I think get apps is a okay. I would say it's a moderate level, it's not a beginner level communities thing, it's like a moderate level advanced, a little bit more advanced level. Um One can start off using it but you definitely have to have some pre recs in place or some understanding of like a pattern in place. Um So what do the other folks think about that opinion? >>I think if you're if you're trying to use get out before, you know what problem you have, you're probably gonna be in trouble. Right. It's like having a solution to it probably don't have yet. Mhm. Right. I mean if if you're just evil or and you're just typing, keep control apply, you're one person right, Get off. It doesn't seem like a big a big jump, like, I mean it doesn't like why would I do that? I'm just, I'm just gonna inside, it's the type of get commit right, I'm typing Q control apply. But I think one of the rules from we've is none of your developers and none of your admins can have cute control access to the cluster because if you can't, if you do have access and you can just apply something, then that's just infrastructure as code. That's just continuous deployment, that's, that's not really get ops um, getups implies that the only way things get into the cluster is through the get up, get automation that you're using with, you know, flux Argo, we haven't talked about, what's the other one that Victor Farsi talks about, by the way people are asking about victor, because victor would love to talk about this stuff, but he's in my next life, so come back in an hour and a half or whatever and victor is going to be talking about sys, admin list with me. Um >>you gotta ask him nothing but get up questions in the next, >>confuse them, confuse them. But anyway, that, that, that's um, it's hard, it's hard to understand and without having tried it, I think conceptually it's a little challenging >>one thing with getups, especially based off the we've works blog post that you just put up on there. It's an opinionated way of doing something. Uh you know, it's an opinionated way of of delivering changes to an environment to your kubernetes environment. So it's opinionated were often not used to seeing things that are very opinionated in this sense, in the in the ecosystem, but get apps is a opinionated thing. It's it's one way of doing it. Um there are ways to change it and like there are options um like what we were talking about in terms of the events that trigger, but the way that it's structured is an opinion opinionated way both from like a tooling perspective, like using get etcetera, but also from a devops cultural perspective, right? Like you were talking about not having anyone access cube control and changing the cluster directly. That's a philosophical opinion that get ups forces you to adopt otherwise. It kind of breaks the model and um I just I want everyone to just understand that. That is very opinion, anything in that sense. Yeah, >>polygamy is another thing. Infrastructure as code. Um someone's mentioning plummy and chat, I just had actually my life show self plug bread that live go there. I'm on Youtube every week. I did the same thing. These these are my friends um and had palami on two weeks ago uh last week, remember uh and it was in the last couple of weeks and we talked about their infrastructure as code solution. Were actually writing code instead of um oh that's an interesting take on uh developer team sort of owning coding the infrastructure through code rather than Yamil as a data language. I don't really have an opinion on it yet because I haven't used it in production or anything in the real real world, but um, I'm not sure how much they are applying trying to go towards the get up stuff. I will do a plug for Solomon hikes. Who has a, the beginning of the day, it's already happened so you can go back and watch it. It's a, it's a, what's it called? Q. Rethinking application delivery with Q. And build kit. So go look this up. This is the found co founder of Dr and former CTO Solomon hikes at the beginning of the day. He has a tool called dagger. I'm not sure why the title of the talk is delivering with Q. And built it, but the tool is showing off in there for an hour is called dagger. And it's, it's an interesting idea on how to apply a lot of this opinionated automated stuff to uh, to deployment and it's get off space and you use Q language. It's a graph language. I watched most of it and it was a really interesting take. I'm excited to see if that takes off and if they try that because it's another way that you can get a little bit more advanced with your you're get deployments and without having to just stick everything in Yemen, which is kind of what we're in today with helm charts and what not. All right. More questions about secrets, I think. I think we're not going to have a whole lot of more, a lot more about secrets basically. Uh put secrets in your cluster to start with and kubernetes in encrypted, you know, thing. And then, you know, as it gets harder, then you have to find another solution when you have five clusters, you don't wanna have to do it five times. That's when you have to go for Walton A W. S secrets and all >>that. Right? I'm gonna post it note. Yeah. Crm into the cluster. Just kidding. >>Yes, there are recordings of this. Yes, they will be later. Uh, because we're that these are all gonna be on youtube later. Um, yeah, detects secrets cushion saying detect secrets or get Guardian are absolute requirements. I think it's in reference to your secrets comment earlier. Um, Camels asking about Cuban is dropping support for Docker that this is not the place to ask for that, but it, it is uh, basically it's a Nonevent Marantz has actually just created that same plug in available in a different repos. So if you want to keep using Docker and kubernetes, you know, you can do it like it's no big deal. Most of us aren't using doctor in our communities anyway, so we're using like container D or whatever is provided to us by our provider. Um yeah, thank you so much for all these comments. These are great people helping each other and chat. I feel like we're just here to make sure the chats available so people can help each other. >>I feel like I want to pick up on something when you mentioned pollux me, I think there's a um we're talking about getups but I think in the original like the origination of that I guess was deploying applications to clusters right, picking up deployment manifest. But I think with the gloomy and I obviously terra form and things have been around a long time, folks are starting to apply this I think I found one earlier which was like um kub stack the Terror Forms get ups framework. Um but also with the advent of things like cluster A. P. I. Um in the Cuban at the space where you can declare actively build the infrastructure for your clusters and build the cluster right? We're not just talking about deploying applications, the cluster A. P. I will talk to a W. S. Spin up, VPc spin up machines, you know, we'll do the same kind of things that terra form does and and those other tools do I think applying getups principles to the infrastructure spin up right, the proper infrastructure as code stuff, constantly applying Terror form um you know, plans and whatever, constantly applying cluster Api resources spinning up stuff in those clouds. That's a super interesting. Um you know, extension of this area, I'd be curious to see if what the folks think about that. >>Yeah, that's why I picked this topic is one of my three. Uh I got I got to pick the topics. I was like the three things that there like the most bleeding edge exciting. Most people haven't, we haven't basically we haven't figured all this out yet. We as an industry, so um it's I think we're gonna see more ideas on it. Um what's the one with the popsicle as the as the icon victor talks about all the time? It's not it's another getups like tool, but it's um it's getups for you use this kubernetes limit and then we have to look it up, >>You're talking about cross plane. >>So >>my >>wife is over here with the sound effects and the first sound effect of the day that she chooses to use is one. >>All right, can we pick it? Let's let's find another question bret >>I'm searching >>so many of them. All right, so uh I think one really quick one is getups only for kubernetes, I think the main to tooling to tools that we're talking about, our Argosy D and flux and they're mostly geared toward kubernetes deployments but there's a, it seems like they're organized in a way that there's a clean abstraction in with respect to the agent that's doing the deployment and the tooling that that can interact with. So I would imagine that in the future and this might be true already right now that get ups could be applied to other types of deployments at some point in the future. But right now it's mostly focused and treats kubernetes as a first class citizen or the tooling on top of kubernetes, let's say something like how as a first class citizen? Yeah, to Brett, >>to me the field, back to you bret the thing I was looking for is cross plane. So that's another tool. Um Victor has been uh sharing a lot about it in Youtube cross plane and that is basically runs inside a kubernetes, but it handles your other infrastructure besides your app. It allows you to like get ops, you're a W. S stuff by using the kubernetes state engine as a, as a way to manage that. And I have not used it yet, but he does some really great demos on Youtube. So people are liking this idea of get off, so they're trying to figure out how do we, how do we manage state? How do we uh because the probably terra form is that, well, there's many problems, but it's always a lot of problems, but in the get outs world it's not quite the right fit yet, It might be, but you still, it's still largely as expected for people to, you know, like type the command, um, and it keeps state locally the ss, clouds and all that. And but the other thing is I'm I'm now realizing that when I saw the demo from Solomon, I'm going back to the Solomon hikes thing. He was using the demo and he was showing it apply deploying something on S three buckets, employing internet wifi and deploying it on google other things beyond kubernetes and saying that it's all getups approach. So I think we're just at the very beginning of seeing because it all started with kubernetes and now there's a swarm one, you can look up swarm, get office and there's a swarm, I can't take the name of it. Swarm sink I think is what's called swarm sink on git hub, which allows you to do swarm based getups like things. And now we're seeing these other tools coming out. They're saying we're going to try to do the get ups concepts, but not for kubernetes specifically and that's I think, you know, infrastructure as code started with certain areas of the world and then now then now we all just assume that you're going to have an infrastructure as code way of doing whatever that is and I think get off is going to have that same approach where pretty soon, you know, we'll have get apps for all the clouds stuff and it won't just be flexor Argo. And then that's the weird thing is will flex and Argo support all those things or will it just be focused on kubernetes apps? You know, community stuff? >>There's also, I think this is what you're alluding to. There is a trend of using um kubernetes and see rDS to provision and control things that are outside of communities like the cloud service providers services as if they were first class entities within kubernetes so that you can use the kubernetes um focus tooling for things that are not communities through the kubernetes interface communities. Yeah, >>yeah, even criticism. >>Yeah, yeah, I'm just going to say that sounds like cross plane. >>Yeah, yeah, I mean, I think that's that's uh there were, you know, for the last couple of years, it's been flux and are going back and forth. Um they're like frenemies, you know, and they've been going back and forth with iterating on these ideas of how do we manage this complicated thing? That is many kubernetes clusters? Um because like Argo, I don't know if the flux V two can do this, but Argo can manage multiple clusters now from one cluster, so your, you can manage other clusters, technically external things from a single entity. Um Originally flux couldn't do that, but I'm going to say that V two can, I don't actually >>know. Um I think all that is gonna, I think that's going to consolidate in the future. All right. In terms of like the common feature set, what Iver and john what do you think? >>I mean, I think it's already begun, right, I think haven't, didn't they collaborate on a common engine? I don't know whether it's finished yet, but I think they're working towards a common getups engine and then they're just going to layer on features on top. But I think, I mean, I think that's interesting, right, because where it runs and where it interacts with, if we're talking about a pull based model, it shouldn't, it's decentralized to a certain extent, right? We need get and we need the agent which is pulling if we're saying there's something else which is orchestrating something that we start to like fuzzy the model even right. Like is this state living somewhere else, then I think that's just interesting as well. I thought flux was completely decentralized, but I know you install our go somewhere like the cargo has a server as well, but it's been a while since I've looked in depth at them. But I think the, you know, does that muddy the agent only pull model? >>I'm reading a >>Yeah, I would say that there's like a process of natural selection going on as as the C. N. C. F. Landscape evolves and grows bigger and a lot of divide and conquer right now. But I think as certain things kind of get more prominent >>and popular, I think >>it starts to trend and it inspires other things and then it starts to aggregate and you know, kind of get back into like a unified kind of like core. Maybe like for instance, cross plane, I feel like it shouldn't even really exist. It should be, it like it's a communities add on, but it should be built in, it should be built into kubernetes, like why doesn't this exist already >>for like controlling a cloud? >>Yeah, like just, you know, having this interface with the cloud provider and be able to Yeah, >>exactly. Yeah, and it kinda, you're right. That kinda happens because you do, I mean when you start talking about storage providers and networking providers was very specific implementations of operators or just individual controllers that do operate and control other resources in the cloud, but certainly not universally right. Not every feature of AWS is available to kubernetes out of the box. Um and you know, it, one of the challenges across plane is you gotta have kubernetes before you can deploy kubernetes. Like there's a chicken and egg issue there where if you're going to use, if you're going to use our cross plane for your other infrastructure, but it's gotta, but it has to run on kubernetes who creates that first kubernetes in order for you to put that on there. And victor talks about one of his videos, the same problem with flux and Argo where like Argo, you can't deploy Argo itself with getups. There has to be that initial, I did a thing with, I'm a human and I typed in some commands on a server and things happened but they don't really have an easy deployment method for getting our go up and running using simply nothing but a get push to an existing system. There's something like that. So it's a it's an interesting problem of day one infrastructure which is again only day one, I think data is way more interesting and hard, but um how can we spend these things up if they're all depending on each other and who is the first one to get started? >>I mean it's true of everything though, I mean at the end of that you need some kind of big bang kind of function too, you know, I started running start everything I >>think without going over that, sorry, without going off on a tangent. I was, I was gonna say there's a, if folks have heard of kind which is kubernetes and Docker, which is a mini kubernetes cluster, you can run in a Docker container or each container will run as a as a node. Um you know, that's been a really good way to spin up things like clusters. KPI because they boot strap a local kind, install the manifests, it will go and spin up a fully sized cluster, it will transfer its resources over there and then it will die itself. Right? So that, that's kind of bootstrapping itself. And I think a couple of folks in the community, Jason to Tiberius, I think he works for Quinyx metal um has, has experimented with like an even more minimal just Api server, so we're really just leveraging the kubernetes ideas of like a reconciliation loop and a controller. We just need something to bootstrap with those C R D s and get something going and then go away again. So I think that's gonna be a pattern that comes up kind of more and more >>Yeah, for sure. Um, and uh, the next, next quick answer to the question, Angel asked what your thoughts on getups being a niche to get or versus others vcs tools? Well, if I knew anyone who is using anything other than get, I would say no, you know, get ops is a horrible name. It should just be CVS office, but that doesn't or vcs ops or whatever like that, but that doesn't roll off the tongue. So someone had to come up with the get ups phrase. Um but absolutely, it's all about version control solutions used for infrastructure, not code. Um might get doctor asks a great question, we're not gonna have time for it, but maybe people can reply and chat with what they think but about infrastructure and code, the lines being blurred and that do develop, how much of infrastructure does developer do developers need to know? Essentially, they're having to know all the things. Um so unfortunately we've had way more questions like every panel here today with all the great community, we've got way more questions we can handle in this time. So we're gonna have to wrap it up and say goodbye. Go to the next live panel. I believe the next one is um on developer, developer specific setups that's gonna be peter running that panel. Something about development in containers and I'm sure it's gonna be great. Just like this one. So let's go around the room where can people find you on the internet? I'm at Brett fisher on twitter. That's where you can usually find me most days you are? >>Yeah, I'm on twitter to um, I'll put it in the chat. It's kind of confusing because the TSR seven. >>Okay. Yeah, that's right. You can't just say it. You can also look at the blow of the video and like our faces are there and if you click on them, it tells you our twitter in Arlington and stuff, john >>John Harris 85, pretty much everywhere. Get hub Twitter slack, etc. >>Yeah >>and normal, normal faults or just, you know, living on Youtube live with Brett. >>Yeah, we're all on the twitter so go check us out there and thank you so much for joining. Uh thank you so much to you all for being here. I really appreciate you taking time in your busy schedule to join me for a little chit chat. Um Yes, all the, all the cheers, yes. >>And I think this kid apps loop has been declarative lee reconciled. >>Yeah, there we go. And with that ladies and gentlemen, uh bid you would do, we will see you in the next, next round coming up next with Peter >>bye.

>>Live from Barcelona, Spain, execute covering CubeCon cloud native con Europe, 2019 onto you by red hat, the cloud native computing foundation and ecosystem partners. >>Welcome back. We're here in Barcelona, Spain where 7,700 attendees are here for Q con cloud native con. I'm Stu Miniman and this is the cubes live two day coverage having to have on the program to returning guests to talk about five years of Kubernetes. To my right is Tim Hawkin wearing the Barna contributors shirt. Uh, and uh, sitting to his right is gay Bon Roy. So, uh, I didn't introduce their titles and companies, but you know, so Tim's and Google gives it Microsoft, uh, but you know, heavily involvement in uh, you know, Coobernetti's since the very early days. I mean, you know, Tim, you're, you're on the Wikipedia page game, you know, I think we have to do some re editing to make sure we get the community expanded in some of the major contributors and get you on there. But gentlemen, thanks so much for joining us. Thanks for having us. >>Alright. Uh, so, you know, Tim just spoke to Joe Beda and we talked about, you know, the, the, the idea of, you know, Craig and Brendan and him sitting in the room and, you know, open source and, you know, really bringing this out there to community. But let's start with you. Cause he, you know, uh, I remember back many times in my career like, Oh, I read this phenomenal paper about Google. You know, we're going to spend the next decade, you know, figuring out the ripple effect of this technology. Um, you know, Coobernetti's has in five years had a major impact on, on what we're doing. Uh, it gives a little bit of your insight is to, you know, what you've seen from those early days, you know. >>Yeah. You know, um, in the early days we had the same conversations we produced. These papers are, you know, seminal in the industry. Um, and then we sort of don't follow up on them sometimes as Google. Um, we didn't want this to be that, right. We wanted this to be alive living thing with a real community. Uh, that took root in a different way than MapReduce, Hadoop sort of situation. Um, so that was very much front of mind as we work through what are we going to build, how are we going to build and how are we going to manage it? How are we going to build a community? How, how do you get people involved? How do you find folks like Gaiman and Deus and get them to say we're in, we want to be a part of this. >>All right, so Gabe, it was actually Joe corrected me when I said, well, Google started it and they pulled in some other like-minded vendors. Like he said, no, no stew. We didn't pull vendors in. We pulled in people and people that believed in the project and the vision, you were one of those people that got pulled in early. He were, you know, so help give us a little context in your, your viewpoint. I did. And, and, and you know, at the time I was working for a company, uh, called, uh, that I had started and we were out there trying to make developers more productive in industry using modern technology like containers. And you know, it was through the process of trying to solve problems for customers, sort of the lens that I was bringing, uh, to this where, um, I was introduced to some really novel technology approaches first through Docker. >>Uh, and you know, I was close with Solomon hikes, the, the founder over there. Uh, and then, you know, started to work closely with folks at Google, uh, namely Brendon burns, who I now work with at Microsoft. Um, you know, part of the, the founding Kubernetes team. Uh, and I, I agree with that statement that it is really about people. It's really about individual connections at the end of the day. Um, I think we do these things that at these coupons, uh, events called the contributor summits. And it's very interesting because when folks land at one of these summits, it's not about who you work for, what Jersey you're wearing, that sort of thing. It's people talking to people, trying to solve technical problems, trying to solve organizational challenges. Uh, and I think, you know, the, the phenomenon that's happened there and the scale with which that's happened is part of the reason why there's 8,000 people here in Barcelona today. >>Yeah. It's interesting to him cause you know, I used to be involved in some standards work and I've been, you know, working with the open source community for about 20 years. It used to be ah, you know, it was the side project that people did at nights and everything like that. Today a lot of the people that are contributing, well they do have a full time job and their job will either let them or asking them to do that. So I do talk to people here that when they're involved in the working groups, when they're doing these things, yes. You think about who their paycheck comes for, but that's secondary to what they're doing as part of the community. And it is, you know, some of the people what, what >>absolutely. It's part of the ethos of the project that the project comes first and if company comes second or maybe even third. Uh, and for the most part, this has been wildly successful. Uh, there's this huge base of trust among, uh, among the leadership and among the contributors. Um, and you know, it's, it's a big enough project now that I don't know every one of the contributors, but we have this web of trust. And, you know, I, I have this, this army of people that I know and I trust very well and they know people and they know people and it works out that the project has been wildly successful and we've never yet had a major conflict or strife that centered on company this or company that. >>Yeah. And I don't, I'd also add that it's an important development has happened in the wake of Kubernetes where, you know, for example, in my teams at Microsoft, I actually have dedicated PM and engineering staff where their only job is to focus on community engagements, right? Running the release team for communities one 15 or working on IPV six support or windows container support. Uh, and, and that work, that upstream work, uh, puts folks in contact with people from all different companies, Google, uh, uh, you know, Microsoft working closely together on countless initiatives. Uh, and the same is true really for the entire community. So I think it's really great to see that you can get not just sort of the interpersonal interactions. We can also get sort of corporate sponsorship of that model. Cause I do think at the end of the day people need to get their paychecks. Uh, and oftentimes that's going to come from a big company. Uh, and, and seeing that level of investment is, I think, uh, pretty encouraging. Okay. Well, you know, luckily five years in we've solved all the problems and everything works perfectly. Um, if that's not maybe the case, where do we need people involved? What things should we be looking at? Kind of the, the, the next year or two in this space, you know, a project >>of this size, a community of this size, a system of this scope has infinite work to do, right? The, the, the barrel is never going to be empty. Um, and in some cases it's filling faster than it's draining. Um, every special interest group, every SIG, it has a backlog of issues of things that they would like to see fixed of features that they have some user pounding the table saying, I need this thing to work. Uh, IPV six is a great example, right? And, and we have people now stepping up to take on these big issues because they have customers who need it or they see it as important foundational work for building future stuff. Um, so, you know, there's, there's no shortage of work to do. That's not just engineering work though, right? It's not just product definition or API. We have a, what we call a contributor experience. People who work with our community to entre online, uh, new contributors and um, and, and streamline how to get them in and involved in documentation and testing and release engineering. And there's so much sort of non-core work. Uh, I could go on on this for. >>Yeah, you're just reminding me of the session this morning is I don't manage clusters. I manage fleets. And you have the same challenge with the people. Yeah. And I also had another dimension to this about just the breadth of contribution. We were just talking before the show that, um, you know, outside at the logo there is this, uh, you know, characters, book characters, and such. And really that came from a children's book that was created to demonstrate core concepts, uh, to developers who were new to Kubernetes. And it ended up taking off and it was eventually donated to the CNCF. Um, but things like that, you can't underestimate the importance and impact that that can have on making sure that Kubernetes is accessible to a really broad audience. Okay. Uh, yeah, look, I want to give you both a, just the, the, the final word as to w what you shout out, you one for the community and uh, yeah. And any special things that have surprised you or exciting you? Uh, you know, here in 2019, >>uh, you know, exciting is being here. If you rewind five years and tell me I'm going to in Barcelona with with 7,500 of my best friends, uh, I would think you are crazy or are from Mars. Um, this is amazing. And uh, I thank everybody who's here, who's made this thing possible. We have a ton of work to do. Uh, and if you feel like you can't figure out what you need to work on, come talk to me and we'll, we'll figure it out. >>Yeah. And for me, I just want to give a big thank you to all the maintainers folks like Tim, but also, you know, some other folks who, you know, may, you may not know their name but they're the ones slogging it out and to get hub PRQ you know, trying to just make the project work and function day to day and were it not for their ongoing efforts, we wouldn't have any of this. So thank you to that. Well and look, thank you. Of course, to the community and thank you both for sharing with our community. We're always happy to be a small piece of a, you know, helping to spread the word and uh, give some voice to everything that's going on here. Thank you so much. All right, so we will be back with more coverage here from coupon cloud native con 2019 on Stu Miniman and thank you for watching the cube.