Tagged: ansible

If you execute the default bare metal installation of an Nginx Ingress server you'll find out that it's actually not listening on the "normal" 80/443 ports, but rather on some client ports. What's worse is that there's a rather small limit for the proxying, so if you're exporting a binary registry, and you want to publish binaries from outside the cluster, you're in trouble. Here's how to fix this.

If you're trying to install a Kubernetes on bare metal, it's useful to document this experience in an Ansible playbook. This makes the installation of new clusters trivial. But after finishing the installation, on a reboot of the node, you might find out your cluster is not coming up. One possible reason is having the swap still enabled.

At its core, Ansible is a declarative system. You describe the state in which you expect your parts of the system to be in, and Ansible, via its modules, tries to get the system there. That's the reason why modules are idempotent. But how are we to approach simple shell executions?

Ansible is an agent-less configuration management system. While agent-less it's true, it still needs two things for it to operate: SSH and python. But what if on the brand new instance we only have SSH access, but no python? Is there a way to install python with Ansible?