latest posts

  • Code Rot

    As I was writing my last post a couple of months ago, the age of this site's decade old toolchain became evident. Maybe it was my new Apple Silicon Mac, or perhaps it was the fact that I hadn't updated any of the gems in over a decade. Whatever the reason, nothing worked! My site had finally succumbed to the dreaded code rot. I was left with a choice: either modernize the site a bit or start over from scratch. I chose to modernize.

  • Fractal Design Terra Jade Workstation

    I originally set out to make a headless Linux server build running Proxmox -- basically a homelab that I'd keep in the basement -- but as I built out my parts list I realized I was making more of a desktop workstation. And as I explored the case options of 2024 I quickly fell in love with the Fractal Design Terra case in jade. Once I went down the mITX form factor path, I knew this thing wouldn't be suitable as a homelab. But that's OK though, I'm really happy with it. 😁

  • Producing a Software Bill of Materials the Easy Way

    A Software Bill of Materials, aka an SBOM, is a key component in software supply chain security. SBOMs are really just a fancy term for a nested list of all of the dependencies that make up a piece of software. There are a number of different utilities that can generate SBOMs for container images after the fact, but I like to get them for free without doing any additional work. That's where Cloud Native Buildpacks come in.

  • Installing Folding at Home on a Raspberry Pi

    Back when I was a kid I used to run Folding@Home on our old family computer -- a silver Gateway tower with a 1.8 GHz Pentium 4. I'm not sure how I found out about it, but I remember being inordinately excited by the prospect of contributing CPU cycles to simulate protein folding and help discover cures for cancer and other diseases. Today, you can buy a quad core Raspberry Pi that is more powerful than that Gateway for $35. Let's have some fun and see how well it can fold proteins!

  • Capturing Network Traffic from a Kubernetes Pod with Ephemeral Debug Containers

    The other day I had a situation where I needed to debug network traffic between an app and its sidecar proxy. The app was using a minimal distroless container image and didn't have easy access to tcpdump. Fortunately, newer versions of Kubernetes have alpha support for ephemeral debug containers which allow us to spin up temporary containers (with debug tools) inside a running pod! In this post we'll see how we can use ephemeral containers and tcpdump to capture network traffic from a running pod.

  • Public Routing to a Private Kubernetes Cluster

    For personal projects and hacking around on Kubernetes, I'm a fan of inexpensive managed clusters. For about ten bucks a month, I can spin up a small single-node cluster and hit the ground running. Unfortunately, with these small clusters you only get about a gig of usable RAM and a single vCPU at best. If you want more, prices quickly climb. At home, I've got a pretty decent Linux box with 12 vCPU and 48G of RAM that I've lovingly named `tomorrowlan`. A similar configuration on Digital Ocean would run about $240 a month! When I'm messing around on personal projects, I really don't care about resiliency or production worthiness. It would be awesome to use my existing hardware while retaining the ability to share what I've built with others on the wider internet. Thanks to a handful of open-source tools, I can make this dream a reality. With Inlets, Contour, KinD, and external-dns, I can easily spin up new clusters and get them externally routable within seconds.

    , updated
  • The Great Raspberry Pi Cooling Bake-Off

    Why is my Raspberry Pi 4 running so hot? You may know you need something to cool it down, but what? In this post we compare the performance of various Raspberry Pi coolers. All the way from the humble heatsink to a massive cooling tower complete with RGB fans.

    , updated
  • Save Money and Skip the Kubernetes Load Balancer

    LoadBalancer Services are super convenient. Create one and your cloud provider will provision a new cloud load balancer, external IP address, and firewall rules to make your workload reachable to the world. They also come with a cost. In this post I'll show how you can save money by skipping the load balancer for your development clusters. We'll see how with a few minor tweaks we can route traffic with an Ingress Controller and plain old fashioned DNS.

  • Using kbld to Rapidly Iterate on Kubernetes Deployed Apps

    When creating applications that extend or interact with Kubernetes, there are times when it's necessary to deploy and develop against a real K8s cluster. While Kubernetes makes it trivial to apply and roll out new changes, the building and pushing new dev Docker images for your application can be a rigamarole. On top of that, you also have to remember to configure the imagePullPolicyfor your containers to Always. Fortunately, there is a tool that can help solve all of these problems: kbld. The kbld CLI (pronounced "k build") assists with all things around image building and pushing for Kubernetes.

  • Leaving Breadcrumbs

    This post is going to be a bit meta. I'm going to write a bit about why I take notes publically and, to a lesser extent, blog in general. I started posting to this blog six years ago for kind of a lame reason. I had purchased the `` domain name and had no clue what to do with it! My earliest content is reflective of that lack of intentionality. Nowadays, I primarily write for three reasons: to cement my own understanding of a subject, to create content where there is none, and to leave digital breadcrumbs for myself and others to find in the future.

  • Desired State Versus Actual State in Kubernetes

    It was once acceptable — and even expected — for web services to go down for maintenance or when under heavy load. Today, however, services are measured in the number of nines of availability they provide. A single server no longer cuts it. One way to achieve higher availability is by running countless copies, or replicas, of our services across geographies. Now, though, we've got ourselves a distributed system to wrangle! In this post we touch on the CAP theorem and the concept of Desired vs Actual State in distributed systems.

  • 2020 Goals

    When it comes to goals and self-improvement, I have a nasty habit of being over-ambitious and over-committing myself -- as evidenced by the piles of technical books and unused yoga equipment cluttering my apartment. In this post I describe what goals, both technical and otherwise, that I will focus on for 2020.

  • Reflections on Kubecon 2019

    A little over a week ago, I had the privilege of attending Kubecon in San Diego. It was an amazing experience, if not just a tad overwhelming in attendance and scale. It was a lot to take in -- and also a lot of fun! Now that I've had a bit of a break, I've had a chance to revisit my notes. In this post, I'll jot down a few of my takeaways from the conference.

  • Simplify Kubernetes App Deployments With Cloud Native Buildpacks and kapp

    In recent years the Kubernetes wave has taken the software world by storm. And for good reason. Kubernetes makes it easy for developers to build robust distributed systems. It provides powerful building blocks for deploying and managing containerized workloads. This makes it an enticing platform for the sprawling microservice "apps" of today. In this post we'll look at simplifying app deployments on Kubernetes with Cloud Native Buildpacks and kapp.

  • Completing Georgia Tech's Online Master of Science in Computer Science

    I applied to Georgia Tech's Online Master of Science in Computer Science (OMSCS) program back in September 2015. One of my friends was applying and encouraged me to apply as well. So I did, not really knowing what I was getting myself in to. I was accepted and took my first class, Software Development Process in Spring 2016 -- right as I was moving out to California. Now, three and a half years later, it's over. I've graduated and completed the Computing Systems specialization while working full time! Now I want to take the time to look back and reflect on my experiences in the program.

  • Finding Max Flow using the Ford-Fulkerson Algorithm and Matthew McConaughey

    The max flow problem is an optimization problem for determining the maximum amount of 'stuff' that can flow at a given point in time through a single source/sink flow network. A flow network is essentially just a directed graph where the edge weights represent the flow capacity of each edge. The 'stuff' that flows through these networks could be literally anything. Maybe it's traffic driving through a city, water flowing through pipes, or bits traveling across the information superhighway. This post walks through how to use the Ford-Fulkerson to determine the max flow of a network.

    , updated
  • The Helman and JáJá List Ranking Algorithm

    In one of our labs in Georgia Tech's Intro to High Performance Computing course, we had to explore the parallel computing problem of list ranking. List ranking is essentially just traversing a linked list and assigning each node a "rank", or distance from the list's head. This post explores two parallel algorithms for list ranking -- Wylies's and Helman-Jájá.

  • Learning to build an Operating System in Rust via CS140e

    I first found out about Stanford's experimental Rust-based Operating Systems course via the GATech OMSCS Slack. I was intrigued by the fact that the course combined two of my favorite topics: the Rust programming language and Raspberry Pis. Now that I've hit a bit of a lull, having just wrapped up grad school for the semester, I've finally had a few cycles to spare and really dig in to the material in Stanford's CS140e.

    , updated
  • Adventures in CS 7641 - Machine Learning

    I just wrapped up Georgia Tech's CS 7641 Machine Learning course last evening and although it was a pretty arduous experience, I got through it. CS 7641 really demystified the field of machine learning for me and took me out of my comfort zone as far as computer science classes go. It was a time-consuming and challenging journey, but ultimately very worth it. In this post, I reflect back on the course and provide resources that I found helpful.

  • Cgroups on the Raspberry Pi Zero

    Cgroups are a feature of the Linux kernel which allow groups of processes to be managed and monitored. For a resource constrained environment such as the five dollar Raspberry Pi Zero, limiting access to scarce resources like memory and CPU can come in handy. In this post we'll take a closer look at the memory, cpu, and freezer cgroups in Raspbian Linux.