Container Basics On Linux (Namespaces/Cgroups)

Give your weird service its own clean process tree

🧩 The Challenge

Ever tried to test a startup script or a small daemon that *insists* on being PID 1 or needs its own pristine process environment to function correctly? Running it straight on your host just isn’t an option; it’ll clash with your real `init` system, and its signals will get all kinds of weird. You need a truly clean process slate without firing up a whole KVM guest or building an OCI image just for a five-minute debug session. Nobody tells you how annoying this can be when you’re just trying to debug one thing, and I’ve wasted hours trying to fake it before.

💡 The Fix

Create a temporary, isolated PID namespace right on your current system. Your command will launch as PID 1 within its own little world, completely invisible and safe from your host’s processes, letting you test those tricky signal handlers or init-style scripts.

sudo unshare --pid --fork --mount-proc bash -c 'echo "Hello from new PID ns, my PID is $$"; ps axf; sleep 10'

⚙️ Why It Works

The `–pid` flag creates a new PID namespace, giving your command its own independent process ID tree. And `–fork` ensures the *first* process in this new namespace is the one you specified (in this case, `bash`), making it PID 1. Then, `–mount-proc` mounts a fresh `/proc` filesystem inside that namespace, so `ps` and other tools see only the processes from *that* isolated world.

🚀 Pro-Tip: Pair `unshare` with `–uts` for a separate hostname, or `–net` for a completely isolated network stack if you’re feeling adventurous.

Linux Tips & Tricks | © ngelinux.com | 10/31/2026