Personal notes on infrastructure and tooling
Spent the weekend wiring up Prometheus and Grafana on my home lab cluster. The bottleneck was always disk space on the metrics retention — solved by moving the TSDB to a separate volume.
Quick recipe: install via apt, set retention to 30 days, and add node_exporter on each host. Dashboards from grafana.com cover 80% of what I need out of the box.
Spent some time digging into how Docker handles bridge networks. The default bridge0 works fine for most cases, but for inter-container DNS you really want a user-defined network.
Also worth noting: --network=host bypasses the bridge entirely. Useful for performance-sensitive apps but breaks isolation.
After years of bare git repo for dotfiles, I migrated to chezmoi. The templating support alone makes it worth it — different configs for work and home machines from a single source.