surfd4wg/aws-terraform-kubernetes

Kubernetes Setup using AWS and Terraform

This repository is copied from opencredo/k8s-terraform-ansible-sample, but it didn't consider kubespray.

https://github.com/opencredo/k8s-terraform-ansible-sample

Also, This repository will create 3 workers, 3 master, and 3 etcd Kubernetes cluster by default. You can adjust the number of each node by changing below Variables.

Step 1. Install Terraform

All steps will be conducted under Docker container for beginners.

# docker run -it --name terraform-aws-kube -h terraform-aws-kube ubuntu:16.04

Install required packages.

$ apt update && apt install git python python-pip unzip wget vim -y && \
    git clone https://github.com/alicek106/aws-terraform-kubernetes.git && \
    cd aws-terraform-kubernetes/terraform

Download terraform binary.

$ wget https://releases.hashicorp.com/terraform/1.0.11/terraform_1.0.11_linux_amd64.zip && \
    unzip terraform_1.0.11_linux_amd64.zip && \
    rm terraform_1.0.11_linux_amd64.zip && \
    mv terraform /usr/bin && chmod +x /usr/bin/terraform

Export your own AWS Access / Secret keys

$ export AWS_ACCESS_KEY_ID=<Your Access Key in AWS>
$ export AWS_SECRET_ACCESS_KEY=<Your Access Key in Secret>

Initialize terraform and generate your SSH key pair for aws_key_pair
```
$ terraform init && ssh-keygen -t rsa -N "" -f ../keys/tf-kube
```
Adjust the number of etcd, worker, and master nodes using Step 2 as shown below.
Create all objects in AWS. It will trigger to create VPC, Subnet, EC2 instances, etc.
```
$ terraform apply
```

Step 2. Set Variables in variables.tf

You can change configuration of file variables.tf, such as the number of each node.

number_of_controller : The number of master nodes that act only as a master role.
number_of_etcd : The number of etcd nodes that act only as a etcd role.
number_of_controller_etcd : The number of nodes that run etcd and master at the same time.
number_of_worker : The number of workers.

It is recommended that the number of [etcd + controller_etcd], [controller + controller_etcd] to be odd. For example, below setting is desirable and can be converted into inventory as shown below. Note that below inventory and setting is just example, not really written configuration.

number_of_controller = 2

number_of_etcd = 2

number_of_controller_etcd = 1

.. is same to

[kube_control_plane:children]
Instnace-A # (1st master)
Instnace-B # (2nd master)
Instnace-C # (3rd master)  (1st etcd)

[etcd:children]
Instnace-C # (3rd master)  (1st etcd)
Instnace-D #               (2nd etcd)
Instnace-E #               (3rd etcd)

[etcd:children]
...

Above example is just example. In groups, instances are represented as group, such as _controller. These groups are filtered by inventory_aws_ec2.yml using filters. If you want to change filters to choose instance correctly (e.g. because of change of Owner tag, or change of region, default to ap-northeast-2), edit inventory_aws_ec2.yml file.

[Optional] if you want to change ClusterID, set cluster_id_tag to another value, not alice.

Step 3. Ansible and Kubespray

First of all, edit inventory_aws_ec2.yml file to match your own configurations, e.g. region and other tags.

In ansible directory, install all dependencies package.
```
$ cd ../ansible && pip3 install -r requirements.txt
```
Install python related modules using raw ansible module to all EC2 instances.
```
$ ansible-playbook --private-key ../keys/tf-kube infra.yaml
```
To check whether it works, use below ansible ping module
```
$ ansible --private-key ../keys/tf-kube -m ping all
```

Download kubespray. You can adjust proper version, but I used v2.18.1 kubespray :D

$ wget https://github.com/kubernetes-sigs/kubespray/archive/v2.18.1.zip && \
    unzip v2.18.1.zip && rm v2.18.1.zip

Warning! Variables of Kubespray (ansible/hosts/group_vars/) is copied from v2.18.1. If you want to use another version of kubespray, you have to remove ansible/hosts/group_vars directory and copy sample variables directory from specific kubespray version. It is usally located in kubespray-x.x.x/inventory/sample/group_vars.

Install Kubernetes. Thats all.

$ ansible-playbook -b --private-key \
  ../keys/tf-kube kubespray-2.18.1/cluster.yml

Test

SSH to your master instance, and get nodes.

root@aws-kube:/aws-terraform-kubernetes/ansible# ssh -i ../keys/tf-kube ubuntu@<Master IP>

...
Last login: Mon Mar 25 10:03:32 2019 from 13.124.49.60
ubuntu@ip-10-43-0-40:~$ sudo su
root@ip-10-43-0-40:/home/ubuntu# kubectl get nodes
NAME                                                      STATUS   ROLES                  AGE     VERSION
ec2-13-125-117-199.ap-northeast-2.compute.amazonaws.com   Ready    <none>                 4m13s   v1.22.8
ec2-13-125-54-209.ap-northeast-2.compute.amazonaws.com    Ready    control-plane,master   5m34s   v1.22.8
ec2-13-209-20-227.ap-northeast-2.compute.amazonaws.com    Ready    <none>                 4m12s   v1.22.8
ec2-3-34-94-130.ap-northeast-2.compute.amazonaws.com      Ready    control-plane,master   6m1s    v1.22.8
ec2-3-38-165-142.ap-northeast-2.compute.amazonaws.com     Ready    control-plane,master   5m22s   v1.22.8
ec2-52-79-249-245.ap-northeast-2.compute.amazonaws.com    Ready    <none>                 4m13s   v1.22.8

Cleanup

In terraform directory, use below command. It will destroy all objects, including EC2 Instances

$ terraform destroy

Limitations

It assumes that master acts as an etcd node. It should be modified to separate etcd and master role.(solved)
Health check of master node is impossible using https:6443 in ELB. (It is recommended to use another proxy such as nginx in Master Node for healthcheck. Health check proxy should be deployed by yourself :D)
Node IP range is limited beacuse node IP is allocated between VPC CIDR + 10, 20, 30... etc. It should be changed if you want to use in production environment.

surfd4wg / aws-terraform-kubernetes