Kubernetes

Prerequisites

• Datajoint

• Docker

Introduction

Typically docker works great for small scale applications, such as deploying to your own personal machine, however often times there is a need to massively scale up which creates a problem. Docker by itself is a bit clucky to deploy to mutiple machines, thus at Google they develop a tool call Kubernetes, a container orchsertration solution.

Kubernetes is basically a giant system that handles scheduling, deploying, and managing many containers from many users with different requirements. With this a single user can easily scale up their applications to use all the resources avliable on the cluster with only a .yaml file and a one line command.

Setup Your Account

Before starting, please message Daniel Sitonic on slack to get an K8 account, he will create an account for you on master node of the cluster which is at-kubemaster1.

Upon your account creation, you will be able ssh in and being using K8

K8 Terminology

• Cluster: A group of machines networked together with kubernetes consistent of a least one master node and a number of worker nodes

• Nodes: A machine or virtual machine with compute resources (CPUs, Mem, GPUs, Storage)

• Pods: A K8 environment that encapsulate a container that specifies the requirements of running container such as number of GPUs, memory, volumes to be mounted, etc.

• Containers: (Docker in this case)