Kubernetes

`Note: the k8s docs are your bible to learn anything about k8s!

rule of thumb 👍: to learn about any command always use the cli with --help subcommand if it is present (most linux cli tools have it!) only if it doesn't explain very well go to the docs 🙂

what is k8s (my POV)?

• after containerization came into picture, people started containerizing apps inside VMs, and to scale it multiple VM copies were created.
• Bottleneck?
  • what if I need to create several 100 of these replicas?
    • lot of manual work (ansible playbook)
    • manual deletion and scaling of VMs
    • if containers crash it had to be respawned manually
    • a container was unaware about its own replicas in other VMs
• enter k8s! (the savior 🦸)
  • a container orchestration tool
  • an OS for the VMs (now the cloud)
  • manages the VMs using control plane (which is a VM in itself), the VMs are called worker nodes (they can also be physical machines)
  • the control plane takes the desired state of the containers from the user and achieves it through its worker nodes
• benefits ➕
  • no manual work
  • just state your desired state of containers in a YAML file and k8s achieves it!
  • k8s also has the ability to automatically scale up or down based on metrics like cpu usage, number of requests, mem usage, time of the day, or any other metric! - no manual scaling
  • if a container crashes, it respawns another one, takes care of all the networking and traffic (basically it maintains the desired state of the application without hampering business) - amazing piece of tech right? 😍

A wonderful picture to explain how k8s works in a local setup using rancher desktop:

• kubectl - a terminal tool to manage the k8s cluster
• control plane - manages the worker nodes and takes instructions from api server which exposes k8s apis, it also has a scheduler to schedule pods and ETCD to preserve the state of the app so that no data gets lost!

in a typical prod setup, a k8s cluster consists of few worker nodes (or VMs) handling the workload and a control plane node to which the kubectl commands interact to declare the desired state of the app!

Links:

kubectl rancher-desktop pod deployment k8s networking

202508202239