Note (11/21/2014): This mostly works, but doesn't currently register minions correctly.
The example below creates an elastic Kubernetes cluster with 3 worker nodes and a master.
- Cluster bootstrapping using cloud-config
- Cross container networking with flannel
- Auto worker registration with kube-register
- Kubernetes v0.4.2 official binaries
- kubecfg CLI
- aws CLI
- CoreOS 490.0.0+
The cloudformation-template.json can be used to bootstrap a Kubernetes cluster with a single command.
aws cloudformation create-stack --stack-name kubernetes --region us-west-2 \
--template-body file://aws/cloudformation-template.json \
--parameters ParameterKey=KeyPair,ParameterValue=<keypair>
It will take a few minutes for the entire stack to come up. You can monitor the stack progress with the following command:
aws cloudformation describe-stack-events --stack-name kubernetes
Record the Kubernetes Master IP address
aws cloudformation describe-stacks --stack-name kubernetes
Skip to kubecfg client configuration
The following commands use the CoreOS 490.0.0 alpha AMI ami-e18dc5d1
from the us-west-2
region. For a list of different regions and corresponding AMI IDs see the CoreOS EC2 cloud provider documentation.
aws ec2 create-security-group --group-name kubernetes --description "Kubernetes Security Group"
aws ec2 authorize-security-group-ingress --group-name kubernetes --protocol tcp --port 22 --cidr 0.0.0.0/0
aws ec2 authorize-security-group-ingress --group-name kubernetes --protocol tcp --port 80 --cidr 0.0.0.0/0
aws ec2 authorize-security-group-ingress --group-name kubernetes --source-security-group-name kubernetes
aws ec2 run-instances --image-id ami-e18dc5d1 --key-name <keypair> \
--region us-west-2 --security-groups kubernetes --instance-type m3.medium \
--user-data file://master.yaml
Record the
InstanceId
for the master.
Gather the public and private IPs for the master node:
aws ec2 describe-instances --instance-id <instance-id>
{
"Reservations": [
{
"Instances": [
{
"PublicDnsName": "ec2-54-68-97-117.us-west-2.compute.amazonaws.com",
"RootDeviceType": "ebs",
"State": {
"Code": 16,
"Name": "running"
},
"PublicIpAddress": "54.68.97.117",
"PrivateIpAddress": "172.31.9.9",
...
Edit node.yaml
and replace all instances of <master-private-ip>
with the private IP address of the master node.
aws ec2 run-instances --count 3 --image-id ami-e18dc5d1 --key-name <keypair> \
--region us-west-2 --security-groups kubernetes --instance-type m3.medium \
--user-data file://node.yaml
aws ec2 run-instances --count 1 --image-id ami-e18dc5d1 --key-name <keypair> \
--region us-west-2 --security-groups kubernetes --instance-type m3.medium \
--user-data file://node.yaml
This command enables secure communication between the kubecfg client and the Kubernetes API.
ssh -f -nNT -L 8080:127.0.0.1:8080 core@<master-public-ip>
Once the worker instances have fully booted, they will be automatically registered with the Kubernetes API server by the kube-register service running on the master node. It may take a few mins.
kubecfg list minions
Create a pod manifest: pod.json
{
"id": "hello",
"kind": "Pod",
"apiVersion": "v1beta1",
"desiredState": {
"manifest": {
"version": "v1beta1",
"id": "hello",
"containers": [{
"name": "hello",
"image": "quay.io/kelseyhightower/hello",
"ports": [{
"containerPort": 80,
"hostPort": 80
}]
}]
}
},
"labels": {
"name": "hello",
"environment": "testing"
}
}
kubecfg -c pod.json create pods
kubecfg list pods
Record the Host of the pod, which should be the private IP address.
Gather the public IP address for the worker node.
aws ec2 describe-instances --filters 'Name=private-ip-address,Values=<host>'
{
"Reservations": [
{
"Instances": [
{
"PublicDnsName": "ec2-54-68-97-117.us-west-2.compute.amazonaws.com",
"RootDeviceType": "ebs",
"State": {
"Code": 16,
"Name": "running"
},
"PublicIpAddress": "54.68.97.117",
...
Visit the public IP address in your browser to view the running pod.
kubecfg delete pods/hello