Show notes
This episode is sponsored by Fiberplane. Your platform for collaborative debugging notebooks!Episode Resources:Try Fiberplane hereFiberplane websiteFiberplane DocsNP-hard VenturesAbout Micha Hernandez van LeuffenMicha Hernandez van Leuffen is the founder and CEO of Fiberplane. He previously founded Wercker, a container-native CI/CD platform that was acquired by Oracle. Micha has dedicated his career to improving the workflows of developers.Read the whole episode (Transcript)[If you want, you can help make the transcript better, and improve the podcast’s accessibility via Github. I’m happy to lend a hand to help you get started with pull requests, and open source work.][00:00:00] Michaela: Hello and welcome to the Software Engineering Unlocked Podcast. I'm host, Dr. McKayla, and today I have the pleasure to talk to Micha Hernandez van Leuffen. He is the founder and CEO of Fiberplane. He previously was the founder of Wercker, a container native CI/CD platform that was acquired by Oracle. Micha has dedicated his career to improving the workflow of developers, so he and I have a lot to talk about today.I'm really, really happy that he's here today and he's also sponsoring today's episode. Welcome to the show. I'm happy that you're here, Micha.[00:00:36] Micha: Thank you for having me. Excited to be on the show.[00:00:38] Michaela: Yeah, I'm really, really excited. So, Micha, I wanted to start really from the beginning. So you are the CEO of Fiberplane and you are the founder of Wercker, which you already sold.So, can you tell me a little bit about how you actually started to this entrepreneur journey of yours and what brought you to the developer experience area.[00:01:03] Micha: Yeah, sure thing. So I have a background in computer science and I did my so, I'm originally from Amsterdam, but I did my thesis at USF.And the topic was autonomous resource provision using software containers. This was all before Docker was a thing, you know, the container format that we now know and love. And I sort of got excited by that field of, of so containers and decided to start a company around it. That company was Worker, so container native CI/CD platform.So we helped developers build tests and deploy their applications to the cloud. We went, I would say, so we went through various iterations of the platform. You know, eventually, you know, we started off with Lxc as a container format and then eventually ended up, you know, having to, to platform on Docker.And Kubernetes. But, you know, it was quite a, quite a journey. So that company eventually got acquired by Oracle to bolster their cloud native strategy. And then, you know, spent a couple years in a Bay area as a VP of software development focusing on their cloud native efforts.Tried to do a little bit of open source there as well, and then, you know, move back to Europe. And so sort of started thinking about what's. Did some angel investing. We're still doing some angel investing as well actually in the sort of same arena. So developer tools, infrastructure building blocks for tomorrow.So I run a, a small precede seat fund with to other friends of mine. But then also started, you know, thinking about what to build next. And you know, we can get into that, but sort of from our experience at running work or this sort of large distributed. Sort of fiber plane was, was born.[00:02:26] Michaela: Cool. Yeah. And so how, how was the acquisition for you? I, from the time I'm, you said you were studying at the university, but then did you write out of university, you know, start worker or maybe already while[00:02:40] Micha: you were Yeah. More or less studying? Yeah. Yeah, more or less just out of university. So it was around 20, 20 12, 20 13.And then, you know, expanded the team. Of course we got an office in San Francisco and, and London. And then 2017 we got acquired by Whirlpool. Oh,[00:02:56] Michaela: very cool. Wow. Cool. So, and you were the, you were the founder of that and also probably cto, CEO. At, at the beginning you were one person shop, or was this, or have this idea and I get some funding and I already, you know, have a team when I'm starting out, or was it more bootstrapped way?How, how was that?[00:03:16] Micha: Yeah, yeah. We both gates, both fiber plane and, and, and worker. We got some funding early on. Then eventually got a CTO. For worker was one of the co-founders of, of OpenStack. So also, you know, very early in the, in that sort of, mm-hmm. container and, and cloud infrastructure journey.And then if for fiber plane, Yeah. There, there's no cto. I'm. I'm both CEO and cto, I guess[00:03:38] Michaela: at the same time. Yeah. Cool, cool. Can you tell me a little bit about fiber plane? What is fiber plane? You know, what does, what does it has to do with containers and with developer experience? What, what kind of of a product is it?[00:03:51] Micha: Yeah, sure thing. So, so guess coming back to the worker days, right? So we, we, you know, we're running this distributed system cic cd, so we were also running users arbitrary code. You know, any, any sort of job could happen on the platform on top of Kubernetes, inside of containers. So one of the things that, you know, stuck with me was it was very hard to always sort of debug the system, like figure out what's really going on when we had some kind of issue.You know, we've going back and forth between metrics, logs, traces, trying to figure out what is the root cause of an issue. So sort of that, that was sort of one thing. So we're thinking a lot about, you know, surely there must be a better way to, to, to help you on this, on this journey. . The other thing that I started thinking about a lot was sort of just challenge the assumption of the dashboard, mm-hmm.So if you think about it, like a lot of the monitoring observability tools are modeled after the dashboard, like sort of cockpit like view of your infrastructure. But I'd say that those are great for the known knowns. So dashboard is great. You set it up in advance, you know exactly what's gonna go wrong.These are the things to monitor. These are the things, you know, to keep tabs on. But then reality hits and you know, the thing that you're looking at, at the dashboard is not necessarily a thing that's. Going wrong. Right? So started thinking a lot about you know, what, what is a better form factor to support that sort of more investigative explorative debugging of your infrastructure.And not to say that dashboards don't have their place, right? It's like still that sort of cockpit view of your infrastructure. I think that's a, a good thing to have. But for debugging, you might wanna sort of more explorative a form factor that also gives you actionable intelligence. I think the other thing that you see a lot with dashboards, like everybody's monitoring everything and now you get a lot of signal and a lot of inputs, but not necessarily the actionable intelligence to figure out what's going on.So that's sort of the other piece where it, then the other, like, the third like I would say is collaboration sort of thing that stuck with. Was also like we've come to enjoy tools like Notion, you know, Google Docs obviously. You know, in the design space we got Figma where collaboration is built in from the get go and it is found that it was kind of odd how in the developer tools and then sort of specifically DevOps.We don't really have sort of these collab collaboration not really built in. Right. If you think about it you know, the status quo of, of you and I debugging an issue is we get on, you know, we get on a. You share your screen you open some dashboard and we started talking over it or something.Right. And so it's, and it's, you know, I guess sort of covid accelerated his thinking a bit, but you know, of everybody going remote you know, how can you make that experience more collaborative?[00:06:22] Michaela: Mm-hmm. . So it's in the incident space, it's in the monitoring space, and you want to bring more collaboration.So how does it work? Yeah,[00:06:32] Micha: yeah, yeah, exactly. So what's your solution now? Yeah. Now I've explained sort of the in inception. Yeah. But yeah, but what is it? What is it? Right. So it's, it's it's a notebook form factor. So very much inspired by data science, right? Like rc, like Jupiter. Yeah, we can Jupyter Notebooks.Yeah. Think of, think of that form factor. Mm-hmm. . We don't use Jupyter or anything like that. We've written everything from scratch. But it's a sort of, yeah, a notebook form factor and you know, built in with collaboration. So you can add, mention people like you would on Slack. You can leave, you know, comments or discussions and all and all that.But where it gets interesting, we've got these things called providers, which are effectively plugins. So they're web assembly bundles, which we can sort of dive into into that as well. But they're providers that connect to your infrastructure, right? So we have, for instance, a provider for Elastic Search for your logs.We have a provider for Prometheus for your. And it allows you to connect to these observability systems and kind of pull 'em together into one form, factor the notebook, and then, you know, start collaborating around that. Mm-hmm. . So, you know, imagine if Notion and Datadog would have a baby . Yeah.That's kind what you get. Yeah.[00:07:41] Michaela: That's cool. So I can imagine that. Let's. I'm on call and hopefully I'm not alone. A call. You are also on call, right? Yeah, and so we would open a fiber plane notebook.[00:07:52] Micha: Hopefully we're in the same time zone and we don't need to like wake up in the middle of that. Yeah.[00:07:57] Michaela: Hopefully. Yes. And then we want to understand. How the system is behaving. And so we are pulling in observers. These are data sources. Yeah. More or less. Right. And then we can do some transformation with those data. Data sources or[00:08:12] Micha: Yeah, yeah. That, yeah, exactly. That, that might be the case. The other thing that we integrate with is, for instance, PagerDuty.So an alert goes off indeed we are on call, but an alert goes off and we have this PagerDuty integration. And subsequently a notebook is created for us already. Mm-hmm. . Okay. Maybe, maybe even with, you know, some, some charts and logs that are already related to the service that might be down.Okay. So depend, So depending on the alert, obviously you're, as you know, you're as good as how you've instrumented your alerts. But say we've written some good alerts, we now have a notebook ready to go. Based off a template. So that's another thing that we, that we have as well, which is this template mechanism.And now, you know, we're ready to, to, to go in, get in into things and start debugging. So we might have a checklist, you know you look at the metrics, I'll look at the logs, sort of this action plan. We pull in that data we start a discussion around it. Mm-hmm. , hopefully we come, we come to the, to the, you know, the root cause of, of our issue.[00:09:11] Michaela: Okay. And so this discussion and this pulling in data, this happens all in the notebook. Can you explain me a little bit more, and also our listeners Exactly. When we are on this, you know, on this call now, having a fiber plane notebook in front of our, what do we see, right? How does that, how does the tool look?[00:09:28] Micha: It's, it's very similar to, I would say, like a Notion Page or a Google Doc page. Mm-hmm. . So we've got like different, different headings. The other thing that we have is, so you might have a title for a notebook, right? You know, the billing, the billing API is now. The other thing that we have is sort of this, this time range.So maybe usually when there's an issue, you know, we've seen this behavior over the last three hours, so we can sort of have that time range locked into place. So we only want to see our. For the last three hours. And that means that any chart that we plot or any log that we pull in will adhere to that global timeframe.So that's what we see. Mm-hmm. . We have support for labels, so, you know, obviously big fans of Kubernetes and, and Promeus. So we, you know, labels are. A first class primitive on the platform. So you're able to sort of populate the notebook with the labels that might maybe be related to our service.Right? So it's a US East one, which is our region. It might, you know, say service is the billing. It might be, you know, environment is production. And the status of our incident is, mm-hmm. ongoing, stuff like that. So we have, we've got, go ahead.[00:10:34] Michaela: Cool. Yeah. And, and so is it then from top to bottom we are writing and we are investigating and we are writing out down the questions that we have and the investigation.Yeah, exactly. We do.[00:10:44] Micha: Yeah. Yeah. And so, so[00:10:45] Michaela: we might have, Is it an Yeah. Is it Yes,[00:10:48] Micha: we our work? Yeah. Yeah. It's sort of Exactly. And I think in the most ideal use case, right. And I do it most ideal scenario, you're kind of like writing your postmortem as you go along. That's what[00:10:59] Michaela: I, I was thinking exactly that.Right. And then maybe next time I'm on call again and I get PagerDuty and something is down, it's again, billing. Can I search in the fiber plane notebooks to find, you know, what we did last time and then[00:11:13] Micha: Exactly. So you'll, you'll search, jump to the conclusion . Yeah. Yeah, exactly. Hopefully, hopefully if you, if you experience, you know, the same issue multiple times at some point, we'll, we'll, you know, do a little commit on GitHub and we, we fix our, fix our, Yeah.Do. But yeah, indeed, so you can search Yeah. Cool. On the notebooks and see if you've, you know, ran into similar issues. So that's, you know, it's great for building up this, this system of record, right. This knowledge base of mm-hmm. . Mm-hmm. . Of infrastructure issues and, and incidents. And it's also great for onboarding, right?If a new person joins, like, this is our process. These are some of examples that we've run into you know, have a look. And now you've got a sense for you know, how we, how we handle things and some of the issues that we've investigated. Yeah. Cool. One more thing on the, on the product. So the other, so, you know, sort of explained the, the notebook form factor.We've got these providers, right, that pull in data. From different, different data sources like Elastic Search or, or Prometheus. The other thing that we have is a command line interface which is called fp. Mm-hmm . And apart from, you know, being able to create notebooks from your terminal and you know, even invite people from the terminal, all this sort of usual stuff that you would, you know, expect from interacting with an API, with, with a product like this, there's two other things that we do.So one is a command called FP Run. And it allows you to, if you are typing a command like cube, ctl logs for a specific pod, you can pipe that the output of that command to a notebook. And why that is useful is of course, you know, when we're de debugging this issue, you and I, and you're start typing things in your terminal.I have no idea what you just did. And this is a way sort of to capture that. So you're piping these these, these outputs from your, the stuff that you're typing into your into your. into the notebook. And the cool thing is, you know, in, on your laptop you just, you know, sort of see text, right?Monospace o…
Full show notes at the publisher 
Our TOPPODCAST Picks 
Stay Connected
