Sat down to practice etcdctl snapshot save, one of the first "must-know" CKA skills. Instead, I spent the evening tuning my Azure sandbox lab automation first.
What I actually did:
- Fixed a permission bug in my resource-group handling
- Split node provisioning so control-plane and workers only install what they need
- Added missing kubeadm prerequisites I'd overlooked (conntrack, socat, ebtables)
- Built a setup-ssh.sh script to handle key permissions and load them into the SSH agent automatically
- Added a connect.sh script so I can SSH into either node with one command
Side effect: provisioning crept from 5:45 to 6:02, because of the extra packages. Worth it for nodes that are actually complete instead of missing half the tools you need for real troubleshooting.
And in the end: I finally got to run my first real etcd exercise.
Sometimes you learn more from a "detour" than from the main task 🤩
At least I now have an easier way to start the provisioning and a much smoother way to connect to the nodes!
My whole lab startup now looks like this:
- git clone the repo
- Upload SSH keys
- source setup-ssh.sh
- bash sandbox-up-calico.sh
- bash connect.sh control
