AJ's Blog

Terminal AI Coding Tools

I mostly use AI tools for code related tasks and over the past year I have found myself preferring terminal based tools over web apps or chat interfaces. That probably makes sense because when I am doing work on a computer, I am already heavily using a terminal. I am searching through files, checking git diffs, running tests, reading logs, and jumping between directories. A terminal based AI tool fits into that workflow more naturally than opening a separate browser tab or app and pasting snippets back and forth.

Manage Helm apps with ArgoCD app-of-apps

I recently started cleaning up how I manage applications in Kubernetes with ArgoCD (Check out those links for an intro k8s and ArgoCD). For a while I was applying individual Application manifests with kubectl manually, which works fine when there are only a few apps. Over time that started to feel messy because there are so many apps.

I wanted something more predictable. The pattern I landed on is a root ArgoCD application that discovers child applications from Git. Each app gets its own directory with an app.yaml, a values.yaml, and optionally a templates/ directory for extra manifests. Once that structure is in place, managing Helm charts becomes much more repeatable whether the chart comes from a public repository or from one of my own apps.

k3s kubernetes cluster

Kubernetes helps manage container based workloads across many machines. If you are not familiar with k8s, check out a previous post for an intro. I am going to switch to k3s as a more lightweight version of Kubernetes for my homelab. Previously I have used kubespray to set up a cluster. I am going to try out k3s which advertises itself as a more lightweight platform than regular k8s.

My goal is to run a Kubernetes cluster on my homelab network for experimenting with container-based workloads and running my own software. Instead of using one or two large, power-hungry servers, I prefer several small machines like Raspberry Pi or mini PCs. K8s (Kubernetes) lets me treat these nodes as a single pool of capacity, so I can deploy and manage applications centrally without configuring each server by hand.

apcupsd_exporter forks and app setup

This is a follow-up to my previous apcupsd post. Check that post out for an overview of this project but essentially the goal is to collect metrics from a UPS system manufactured by APC and present them in Prometheus format to scrape with Prometheus.

I forked the GitHub repositories needed for apcupsd_exporter and kept the MIT license in place. That means anyone can still use, modify, and redistribute the code under the same open source terms.

Store Metrics with InfluxDB v3

Modern time-series data demands modern solutions. Whether you’re monitoring server metrics, tracking IoT sensor data, or analyzing application performance, InfluxDB v3 Core combined with Telegraf provides a foundation for an Observability platform. In this post, we’ll walk through setting up a complete monitoring stack using Docker Compose, querying data via HTTP API, and understanding the migration path from older InfluxDB versions.

I have not used InfluxDB a lot in my homelab but I have used it at my job to store custom business application metrics. Once you have millions of metrics to keep track of, a single database and disk will become a bottleneck. If you are not familiar with Telegraf and InfluxDB, check out a previous post for an introduction to these tools and InfluxDB version 1 (open-source version).

Set up a fully featured editor with LazyVim

I have used vim and vi for a long time after I needed to learn a command-line only text editor for my first job deploying and managing server racks. Eventually I started using neovim as it was easier for me to customize than learning vimscript. Since neovim uses Lua for configuration, it was easier for me to learn the syntax. I only customized a few features though because I do not like spending a lot of time reviewing config files just to get beyond the “80 percent” of use cases I need in my editor.