Skip to content

Latest commit

 

History

History
44 lines (44 loc) · 2.24 KB

k8s.md

File metadata and controls

44 lines (44 loc) · 2.24 KB
Raw
  • Monolithic -> microservices: for load-balancing etc.
  • Use microservices over containers instead of real machines to exploit benefits of containers.
  • K8s provides automatic management of containers: it handles CRUD of containers and clustering them.
  • Problems with scaling up containers that k8s helps with:
    • Containers cannot communicate with each other
    • Autoscaling and load balancing is not possible
    • Containers had to be manually managed, introducing scope for human errors
  • Features of k8s:
    • Horizontal and vertical scaling
    • Auto healing
    • Load balancing
    • Platform (cloud/virtual/physical)
    • Fault tolerance
  • Manifest: initial configuration.
  • Cluster -> Node -> Pod -> Container -> Application/microservice.
  • Pod: abstraction over containers to provide a uniform communication interface irrespective of underlying container (may be docker or something else).
  • Node: physical server/VM on which pods are running.
  • Each pod is assigned an IP address, not each container in the pod.
  • Although each pod may theoretically have more than one container, usually only one container is run per pod to avoid tight coupling of containers.
  • Control plane = etcd + kube-API server + kube-scheduler + control manager
  • 1 control plane / master
  • kube-API server: interface to interact with master.
  • etcd: highly available key-value store for shared configuration, service discovery, and scheduler coordination.
  • kube-scheduler: handles pod creation and management, match/assign any node to create and run pods.
  • Controller/manager: responsible for node detection, setting up network routes, setting up load balancers, volumne management
  • Components of a node:
    • Kubelet: agent running on each node that communicates with master via API server. Runs on port 10255.
    • Container engine/runtime: docker etc. Exposing containers on port specified in manifest.
    • Kube-proxy: assigns an IP address to each pod.
  • Deployment: manages a set of Pods to run an application workload, usually one that doesn't maintain state. A Deployment provides declarative updates for Pods and ReplicaSets. You describe a desired state in a Deployment, and the Deployment Controller changes the actual state to the desired state at a controlled rate