cloudmeesh-kubernetes

see cloudmesh.cmd5

https://github.com/cloudmesh/cloudmesh.cmd5

TEST COMMAND: k3 deploy --host=HOSTS

Deploy a kubernetes cluster by hand

We augmented each steo with

All: to indicate steps to be executed on all PIs
Master: to indicate steps to be executed on the master only
Workers: to indicate steps to be executed on the workers only

Do the steps 1-4 on EACH Pi (Master & Workers)

All: Update and upgrade and Disable swap (Interfers with Kubernetes)

sudo apt-get update && sudo apt-get upgrade
sudo dphys-swapfile swapoff
sudo dphys-swapfile uninstall
sudo update-rc.d dphys-swapfile remove
sudo swapon --summary

Double Check That Swap Was Disabled: If there is no output, then this has been done correctly

All: Edit /boot/cmdline.txt. Add this to the end of the line (put a SPACE after whatever exists and then add this)
```
sudo echo "cgroup_enable=cpuset cgroup_memory=1 cgroup_enable=memory" >> /boot/cmdline.txt
```
All: Reboot the Pi
```
sudo reboot
```
Master: Next we set up kubernetes on the Master

SSH into the master Pi and run this command and get a kubernetes token
```
curl -sfL https://get.k3s.io | sh -
sudo cat /var/lib/rancher/k3s/server/node-token
```
Master: Register the mater address by running this command and replacing MASTER_IP_ADDRESS with the IP address of the master Pi
```
export KUBERNETES_MASTER=http://MASTER_IP_ADDRESS:8080
```
Q: can we use localhost?

Something like this may help
```
ifconfig | fgrep inet | fgrep . | fgrep -v 127 | cut -d ' ' -f 2
```
ifconfig may be if a on newe osses, please check
Workers: On each worker register the master. SSH into each of the worker Pi's and run this command, replacing the MASTER_IP_ADDRESS with the IP address of the master Pi and NODE_TOKEN_HERE with the node token from above
```
curl -sfL https://get.k3s.io | K3S_URL=https://MASTER_IP_ADDRESS:6443 K3S_TOKEN=NODE_TOKEN_HERE sh -
export KUBERNETES_MASTER=http://MASTER_IP_ADDRESS:8080
```
Master: Confirm Setup in the master
```
sudo kubectl get nodes
```
This will list out all the worker Pi's connected to the master Pi. If you have setup everything correctly, you will see the hostname of the master and each worker Pi listed with Ready status. Note, that it will take some time for the nodes to show up, so you need to repeat that command for some time. To automatically watch it every 5 seconds you can use
```
watch -n 5 sudo kubectl get nodes
```

cloudmesh/cloudmesh-kubernetes

cloudmeesh-kubernetes

Deploy a kubernetes cluster by hand