Kubernetes

Kubernetes is a container orchestrator letting you manage hundreds of containers at a time. While managing multiple containers with engines such as Docker and Podman, it is not really feasible if you want hundreds of them.

Kubernetes is composed of multiple components and can be intermingled with other related tools. For this, pre-bundled solutions (also known as distributions) such as K3s exists with some related tools embedded to make installation easier.

Most installations feature the kubectl that controls the Kubernetes cluster. This is the program that you'll most likely going to pay attention to.

• composed of multiple smaller components:

  • kubelet runs on all nodes, managing nodes — i.e., notifying the API server where the nodes belongs to

  • kubeconfig is the term for a Kubernetes cluster configuration; mainly written in YAML file

• Kubernetes uses the building blocks structure (or resources) that composes of simple blocks that can be combined to create complex operations

• it also has extensibility mainly through exposing the API and letting you create custom resource definitions to fully extend Kubernetes with your app

• Kubernetes manages the cluster along with some resources

  • the nodes which contain multiple pods

    • nodes are categorized either as a master or a worker node

    • a master node is where the control plane is located

    • a worker node is where the data plane is located

  • pods which is a collection of containers; usually, they are managed indirectly with deployments or resource sets; while pods can run multiple containers, it often contains one container as a recommended practice

  • deployments create the specification of the application state

  • replica sets ensure that pods are running at any given time even while replacing the pods

  • configmaps manages secrets in the cluster; this includes passing configurations between different pods; secrets are either set imperatively (in the command line) or declaratively (from a file)

  • namespaces separates application from each other, similarly to namespacing found in programming languages — e.g., Rust with modules, C++ namespaces, etc.

• while you can manage resources in the command line, you can create manifests that Kubernetes can simply apply

  • think of it as declarative way of cluster management (i.e., with files) instead of imperative (i.e., invoking in the command line)