elsonrodriguez

windows:
 - name: kube-rolling
   root: ~/
   layout: main-horizontal
   panes:
     - commands:
       - export update_command="kubectl rolling-update <rc-name> --update-period=1s --image=<image_url>"
       - clear
       - read -p "$update_command"
       - $update_command

elsonrodriguez

#!/bin/bash
REGISTRY_URL=domain.com/name/project
RC_NAME=replication-controller
GITVER=`git rev-parse -q HEAD | cut -c1-8`
DOCKER_TAG=$REGISTRY_URL:$GITVER

docker build -t $DOCKER_TAG .
docker push $DOCKER_TAG

kubectl rolling-update $RC_NAME --image=$DOCKER_TAG

elsonrodriguez

#get all nodes running pod with specified label
kubectl get pods -l “key=value" -o template --template='{{range .items  }}{{.spec.nodeName}}{{"\n" | printf "%v"}}{{end}}’

elsonrodriguez

#KERNEL=="js?", ATTRS{name}=="HJZ Mayflash Wiimote PC Adapter", NAME="input/js100"
#SUBSYSTEM=="input", KERNEL=="js[0-9]*", ATTRS{name}=="HJZ Mayflash Wiimote PC Adapter", ENV{ID_INPUT_JOYSTICK}=="?*", MODE="0000", ENV{ID_INPUT_JOYSTICK}="", RUN+="/bin/rm %E{DEVNAME}"
SUBSYSTEM=="input", ATTRS{name}=="HJZ Mayflash Wiimote PC Adapter", ENV{ID_INPUT_JOYSTICK}=="?*", RUN+="/bin/rm %E{DEVNAME}", ENV{ID_INPUT_JOYSTICK}=""
SUBSYSTEM=="input", ATTRS{name}=="HJZ Mayflash Wiimote PC Adapter", KERNEL=="js[0-9]*", RUN+="/bin/rm %E{DEVNAME}", ENV{ID_INPUT_JOYSTICK}="" 

elsonrodriguez

Quick and Dirty Node Selection

So first, let's create our pod.

$ kubectl create -f https://gist.githubusercontent.com/elsonrodriguez/60e53e2479dc3146447b/raw/3b4d703855500c829ae58d70e386b4e41e4f7996/pod.yaml
pod "gputest" created

Now let's see it go:

elsonrodriguez

scp ~/Downloads/*.deb user@host:
scp ~/Downloads/etcd*.tar.gz user@host:
scp ~/oss/kubernetes/_output/dockerized/bin/linux/amd64/k* user@host:

elsonrodriguez

Intro

This is the formula for terminating k8s nodes that have a bad kubelet. This effectively fences nodes that are unable to unmount/detach their own storage.

Caveats

Early testing shows that this may not work if a node is rebooted in response to a kubelet/system issue. Therefore the action taken in response to a problem on a kubernetes node MUST BE to terminate it.

Overview

Make LB like this:

 "HealthCheck": {

elsonrodriguez

Summarizing 16 Failures:

[Fail] [k8s.io] EmptyDir volumes [It] should support (non-root,0777,default) [Conformance] 
/Users/eorodrig/oss/kubernetes-elsonrodriguez/_output/local/go/src/k8s.io/kubernetes/test/e2e/framework/util.go:1617

[Fail] [k8s.io] EmptyDir volumes [It] should support (root,0644,tmpfs) [Conformance] 
/Users/eorodrig/oss/kubernetes-elsonrodriguez/_output/local/go/src/k8s.io/kubernetes/test/e2e/framework/util.go:1617

[Fail] [k8s.io] EmptyDir volumes [It] volume on default medium should have the correct mode [Conformance] 
/Users/eorodrig/oss/kubernetes-elsonrodriguez/_output/local/go/src/k8s.io/kubernetes/test/e2e/framework/util.go:1617

elsonrodriguez

Summarizing 20 Failures:

[Fail] [k8s.io] ConfigMap [It] should be consumable from pods in volume as non-root [Conformance] 
/Users/elsonrodriguez/oss/kubernetes-elsonrodriguez/_output/local/go/src/k8s.io/kubernetes/test/e2e/framework/util.go:1620

[Fail] [k8s.io] EmptyDir volumes [It] should support (non-root,0777,default) [Conformance] 
/Users/elsonrodriguez/oss/kubernetes-elsonrodriguez/_output/local/go/src/k8s.io/kubernetes/test/e2e/framework/util.go:1620

[Fail] [k8s.io] EmptyDir volumes [It] should support (root,0644,tmpfs) [Conformance] 
/Users/elsonrodriguez/oss/kubernetes-elsonrodriguez/_output/local/go/src/k8s.io/kubernetes/test/e2e/framework/util.go:1620

elsonrodriguez

Course Prerequisites

• Local (No VM) terminal with working bash and curl (for Windows users, gitbash or cygwin can accomplish this).
• Windows, Linux, OSX
• Slack

Installing kubectl, the Kubernetes client

First create directories for this class: