Tag Archives: open-shift

JavaOne – Docker, Kubernetes & Open Shift

Posted on by
Reply

“Project Jigsaw – Integration with Tools”

Speaker: Marek Jelen & Diogenes Rettori

Sponsored session by RedHat [I didn’t know they had sponsored sessions; I asked if they applied to JavaOne or Oracle Open World and they said neither; they paid. I was curious why they were JavaOne and yet had laptops provided by Oracle. That explains why three sessions of this class were offered.]

For more blog posts from JavaOne, see the table of contents

Can do self paced tutorials in cloud install at https://learn.openshift.com

[Nice assortment of shipping container pictures on the screen as he reviews what Docker/Kubernetes are for and why this matters. A bit of preaching to the choir here…]

“Very diverse community of contributors” The top six contributors are all from Google and RedHat. (except one who move to Microsoft) [sounds more like a strong partnership than a diverse community]

Open Shift and Kubernetes

  • OpenShift is an enterprise version of Kubernetes with extra stuff.
  • OpenShift 3.6 in lab
  • Can use kubectl with Open Shift cluster
  • OpenShift sometimes uses own version of Kubernetes objects and wrap others
  • RHEL (red hat enterprise linux) has master and nodes.
  • Nodes used to be called minions.
  • Works with cloud and non-cloud deployments
  • OpenShift adds Developer CLI,, Admin CLI, web console, REST APIs, creating an image from source, authentication and users/groups from LDAP, Also wrap namespace in a project and more

Workshop

bit.ly/javaoneworkshop

http://content-workshop.apps.javaone3.gce.pixy.io/index.html#/workshop/training/module/install

I’ve read about and played with Kubernetes. I didn’t realize OpenShift was Kubernetes with different commands.

Concepts from lab

  • pod – smallest deployable unit
  • service – group of pods
  • replication controllers – ensure correct number of pods exist
  • deployment configuration – how should be configured
  • horizonal pod auto scaler – monitors CPU and scales. See here
  • route – so external clients can access
  • service account – non-user account, get one by default

Commands in lab

  • oc login url – to login
  • oc get x – basic info
  • oc get x -o yaml – more detail in yaml
  • oc describe service x – even more detail
  • oc scale –replicas=x y – or can click up arrow in UI
  • oc delete pod x – to delete
  • oc expose service x – to create route
  • oc logs x – view logs
  • oc policy add-role-to-user view -z default – grant access to view for default user
  • oc policy add-role-to-user view user1 – grant access to view for user1

In UI

  • Add project > docker
  • Change number of replicas with up/down arrows once expand overview
  • View logs by clicking “…” on overview
  • Add project > catalog > python – use source to image so can go straight from github source code

Real world

  • 18% adults have anxiety problems
  • “not all anxiety is bad” depends on the task; issue on more complex/unfamiliar/difficult tasks
  • 20% of downtime is caused by tech failure or natural disasters. Rest caused by people
  • batch size – amount of features in a release

A/B Deployment demo

  • installed Node.JS app directly from github
  • then created a separate app from github. (so now have two apps in open shift)
  • edit route in open shift console to divide traffic between the two services. can choose percentages/weights
  • changed RAM; did rolling deployment so old one stays up until new one is ready

My take: We immediately had a problem because the laptop was setup differently than the lab; odd. not a big deal to type ./oc instead of oc. But that seems easy for them to have fixed in the VM. then the project we were supposed to switch to we were already on. then it says to login with your email vs the name we were given. I’ll stop commenting on the differences now, but this clearly wasn’t written for or tested for JavaOne. It even refers to “in the morning session” for how you find out what a pod is. I asked and they said this was written to be a full day of training. So glad that I’ve already read about Kubernetes and was able to follow it anyway. I like when they showed how to do the same task in both the UI and at the command line

[I did 9/12 of the labs. It wasn’t that they were bad or long. I was having trouble focusing. It’s 100% self paced and I wasn’t asking any questions. Or learning much. A small percentage of the lab is the UI and the rest I’ve done already with raw Kubernetes. So decided to use the remaining time to check twitter posts about other parts of the conference. I didn’t leave because the last 20 minutes are a demo of something I haven’t seen so wanted to see that.]