CKAD theory pt.1 - Core Concepts

I thought that I could publish my CKA/CKAD preparation notes, maybe it will be useful for someone. Mostly it’s a compilation of already existing articles, but somewhere I’ve added my own notes. So enjoy pt 1 - core components!.

Core Components

Master Components

• kube-apiserver - exposes Kubernetes API on the master node
• etcd - HA key-value store, used to store all Kubernetes data
• kube-scheduler - schedules pods to nodes
• kube-controller-manager - runs controllers on master
  • node controller - noticing and responding when node go down
  • replication controller - keeping corrent number of pods for every replica
  • endpoint controller - updates services with correct pod IP addresses
  • service account & token controllers - create default account and API for new namespaces
• cloud-controller-manager - interact with underlying cloud providers(AWS, Azure, …) K8s 1.6+

Node components

• kubelet - agent on each node in the cluster, makes sure that containers are running
• kube-proxy - network proxy that runs on each node in the cluster, maintains network rules on nodes
• container-runtime - underlying software that runs containers(Docker, cri-o, containerd)

Kubernetes objects

Object in Kubernetes is a persistent entity which represent state of API primitives(objects). Every object must be described using following API conventions:

• Required fields
  • apiVersion - which version of k8s API to use to create this object
  • kind - what kind of object you’d like to create
  • metadata -data that helps uniquely identify the object, including
    • namespace - if not defined, used default namespace
    • name - string that uniquely identifies that object in current namespace
    • uid - unique value, generated identifier used to distinguish between objects
    • optional or generated
      • resourceVersion, generation, creationStamp, deletionStamp
      • labels - map of string key:value that used to organize and categorize objects
      • annotations - map of string key:value that can be used by external tools
• Spec - description of desired state of an object
• Status - description of an actual state of an object, updated by k8s

Pods

Pod is a container or a set of containers in the same namespace. All containers inside a single Pod share storate/network, and spec how to run the containers.Each Pod is assigned with a unique IP address. Containers inside a Pod can communicate with one another using localhost. Usually Pods are not used by themselves and manager by Controllers. In this case Controllers manager scheduling, replication, rollout, self-healing and other capabilities.

Namespaces

Namespaces is a logically separated space which keeps containers, services, network, storage and permissions separated. Namespaces can be thought of as Virtual Clusters. Each namespace can be assigned with a set of resources. Object names need to be unique for each object in the same namespace.

Default namespace can be changed by:

kubectl config set-context --current --namespace=xxx

Notes

• ContainerPort defines which port to expose on the pod’s IP address(Basically to which port create NAT rules)

Links

Kubernetes API conventions