Kubernetes

Kubernetes is like an operating system for computer clusters. It frees developers from implementation of infrastructure mechanisms like:

• service discovery
• horizontal scaling
• load balancing
• self-healing
• leader election

Kubernetes clusters are divided in Control Plane and Workload Plane or Data Plane

Benefits of using kubernetes

• self service deployment of applications
• reducing costs via better infrastructure utilization (tightly applications together)
• automatically adjusting to changing Load: in cloud you can even provide more nodes through the cloud providers api
• keeping application running smoothly
• simplifying application development

Control Plane Components

• scheduler: what node should run instance of an application
• controllers: each controller have different tasks
• api server: components that run in worker node and developer/operators communicate with this api
• etcd: distributed datastore, the server itself is stateless

Worker node components

• kubelet: node agent manages applications running on this node and talks with the api server
• container runtime: typically docker
• kubeproxy: creates a virtually load balancer for each application that needs it

Modern Application Architecture

• Cloud Native
• Modular
• Microservice Base
• Scalability
• Reliability
• Resilience

Requirements for containers

• Packaging
• Versioning
• Isolation
• Fast startup
• Low overhead

Requirements for Orchestration

• Clustering
• Discovery
• Configuration: allowing to reconfigure without a rebuild and redeploy
• Access control
• Load balancing
• Monitoring: identify failed Microservice instaces
• Resilience: Automatically recover from failures

Pods

Is the smallest building block in kubernetes. A pod is made up of one or more container that share network and storage resources. Pods are incorporated in a higher level abstraction like ReplicaSet.

ReplicaSet

Are used to keep fixed number of identical pods. You need it if you want to create a orchestration behavior.

Deployments

Is a resource that manage pods and ReplicaSets. The main job is to maintain the state that is configured in its manifest

StatefulSets

Resource to manage stateful applications like MongoDB, Postgres and etcd. Like deployments it can manage the state of a pod in a manifest.

Services

Used to expose application in a pod withing the kubernetes or the internet. There are the following services types:

• ClusterIP
• Headless
• NodePort: Expose the service in the node ip address and port
• LoadBalacer: Expose the service externally
• ExternalName

Volume

Basically a directory or file that all container in a pod can access. If the entire POD is killed the Volume and its data will be removed. If needed to persist data use Persistent Volume. A PV is a resource in a cluster just like a node. If running in AWS you can use Amazon Elastic Block Storage (Amazon EBS)

Secrets

ConfigMaps

Namespaces

It provides a logical separation of resources. If you dont specify while creating a resource it will be named default

Control Plane

• kublet
• kube-proxy
• container runtime
• etcd a key value distributed database

API Server

• API Server Aggregation (Service Catalog, etc)
• Custom Resource Definition

Open Service Broker

• Connect applications to hosted service and partner services using Kubernetes API

Controller Manager

Controller/Operator pattern

Examples:

• Prometheus Operator

Scheduler

Kubelet