Skip to content

Latest commit

 

History

History
127 lines (101 loc) · 4.04 KB

README.md

File metadata and controls

127 lines (101 loc) · 4.04 KB
Raw

Docker Automated build

Base image for k8s workshops. The main idea is to provide an environment where participants can kubectl right away, "no-installation-needed"

  • each user has its own namespace
  • each user works inside a prepared pod
  • preinstalled tools:
    • kubectl,helm
    • curl,bash,bash-completion,unzip,git,dig,net-tools
    • tmux/vim/micro
  • sample git repo pre-pulled into $HOME
  • KUBECONFIG set up to own namespaces with token and server ca.pem set

For instant kubectl access a browser based connectivity is provided. Under the hood bash is shared via a websocket, based on: yudai/gotty.

See the last section for explanation about different access methods, and why the browser based is preferred. Please note, that while browser based is poreferable for first-time users, additional ssh access can be easily provided. See sshfront

Roadmap

  • Add browser based editor. Possible zed
  • Record sessions (presenter/participant): ttyrec/ascicast
  • Persistent storage on VPC

Usage

First source all the functions:

. workshop-functions.sh

A ClusterRole and a binding is needed initially

init

Create a devenv for a participant:

$ dev user5
namespace/user5 created
serviceaccount/sa-user5 created
role.rbac.authorization.k8s.io/role-user5 created
rolebinding.rbac.authorization.k8s.io/rb-user5 created
clusterrolebinding.rbac.authorization.k8s.io/crb-user5 created
secret/user5 created
deployment.apps/user5 created
service/user5 created

Share the url for browser connection:

$ get-url user5
open http://35.199.111.25:31234/1g7j7u0l/

Related work

Presenter

The presenter can share its terminal session in as read-only via browser. Participants can easily copy-paste commands.

Share the presenter url with participats:

$ presenter-url

Start the presenter session (user0):

$ presenter

tl;dr

There are various ways for participants to work with the k8s cluster:

Method-1: kubectl from laptop

  • pros:
    • own/custom editor
    • own/custom key-bindings
    • no "vi freak-out"
  • cons:
    • win/osx/linux differences
    • potential network/proxy/firewall issues
    • time consuming for first-time users (upto 30 minutes)

Method-2: ssh into k8s nodes:

  • pros:
    • no-install-needed
    • simple concept
  • cons:
    • potential "vi freak-out"
    • turnkey solutions (gke/aks/eks) can use esoteric/readonly os
    • ssh password/key setup
    • potential network/proxy/firewall issues
    • hard to scale/assign (5 nodes / 20 participants)

Method-3: ssh into non-k8s VMs:

  • pros:
    • no-install-needed
    • easier to scale
    • can use different node-type cpu/mem
    • preemptible/spot instances to save $
  • cons:
    • potential "vi freak-out"
    • CluserIp unreachable
    • potential network/proxy/firewall issues
    • lots of ssh accounts to manage
    • ansible or similar tool needed

Method-4: browser connection to a container:

  • pros:
    • absolutely no install needed
    • easy to scale
    • instant access
    • no ssh password/key sharing needed
  • cons:
    • key-bindings can be confusing

Please note, that while browser based is poreferable for first-time users, adiitional ssh access can be easily provided. See sshfront