Skip to content

Kubernetes on AWS with Auto-Applying

Overview

This workflow demonstrates the creation of a multinode Kubernetes cluster on AWS which utilises Flight Solo to automatically configure nodes as workers in the Kubernetes cluster upon boot.

Prepare Network

  1. Create a VPC with
    1. Name: kubecluster1
  2. In "VPC and More"
    1. IPv4 Subnet Block: 10.10.0.0/16
    2. Number of Availability Zones: 1
    3. Public Subnets: 1
    4. Private Subnets: 0
    5. VPC Endpoints: None
  3. Create a Security Group with
    1. Name: kubecluster1-sg
    2. Description: "Security group for kubecluster1 cluster"
    3. VPC: kubecluster1
  4. Inbound Rules
    1. "SSH" from "Anywhere-IPv4"
    2. "HTTP" from "Anywhere-IPv4"
    3. "HTTPS" from "Anywhere-IPv4"
    4. "All Traffic" from 10.10.0.0/16

Build Login Node

Launch the Flight Solo image in AWS marketplace

  1. Select "Launch from EC2"
  2. Instance Type: t3.2xlarge
  3. Click "Edit" on Network Settings and
  4. VPC: kubecluster1
  5. Auto-assign public IP: Enable
  6. Select existing security group: kubecluster1-sg
  7. Set root volume size to at least 20GB
  8. Under Advanced Details -> User Data, add this: 
    #cloud-config
write_files:
  - content: |
      SHAREPUBKEY="true"
      AUTOPARSEMATCH=".*"
      AUTH_KEY=kubecluster1
    path: /opt/flight/cloudinit.in
    permissions: '0600'
    owner: root:root
  - content: |
      /opt/flight/bin/flight profile prepare openflight-kubernetes-multinode
    path: /var/lib/firstrun/scripts/00-prepare-profile.bash
    permissions: '0600'
    owner: root:root

Note

The above data will enable sharing the public key to clients, automatically add any nodes that connect to hunter with a correct auth key, and secure node hunting with the authorisation key of kubecluster1. For more information see the user data documentation

Configure Login Node

  1. Parse & Label Login Node
    1. Run the command flight hunter parse.
    2. Select the node with Space
    3. Press Down to erase the field.
    4. Type in login1 as the label and press Enter.
  2. Run flight profile configure and Select OpenFlight Kubernetes Multinode
    1. Cluster Name: kubecluster1
    2. Default user: flight
    3. Set user password to: Some secure password
    4. NFS server (hostname or flight-hunter label): login1
    5. IP or FQDN for Web Access: Public IPv4 DNS from AWS EC2 Console for Node
    6. IP Range of Compute Nodes: 10.10.0.0/16
    7. IP Range of Kubernetes Pods (must not overlap with Compute Node IP Range): 192.168.0.0/16
  3. Apply the master profile to login1 with flight profile apply login1 master
  4. Wait for flight profile list to show the status of login1 as completed (flight profile view login1 --watch)

  5. Setup Automatic Application of Hunter Nodes

    1. Add the following lines to /opt/flight/opt/hunter/etc/config.yml to automatically apply the worker profile to hunter nodes with labels containing node. 

      auto_apply:
  node: worker

      Note

      You will need to use sudo to have permissions to edit this config file.

    2. Restart hunter service 

      flight service restart hunter

Launch a Compute Node

Launch the Flight Solo image in AWS marketplace

  1. Select "Launch from EC2"
    1. Instance Type: t3.2xlarge
    2. Click "Edit" on Network Settings and select:
      1. VPC: kubecluster1
      2. Auto-assign public IP: Disable
      3. Select existing security group: kubecluster1-sg
    3. Set root volume size to at least 20GB
    4. Under Advanced Details -> User Data (replace LOGIN_SERVER_IPV4_PRIVATE_ADDRESS with the IP of the login node, this can be found with ip addr on that system) 
      #cloud-config
write_files:
  - content: |
      SERVER=<LOGIN_SERVER_IPV4_PRIVATE_ADDRESS>
      LABEL=node01
      AUTH_KEY=kubecluster1
    path: /opt/flight/cloudinit.in
    permissions: '0600'
    owner: root:root
  - content: |
      /opt/flight/bin/flight profile prepare openflight-kubernetes-multinode
    path: /var/lib/firstrun/scripts/00-prepare-profile.bash
    permissions: '0600'
    owner: root:root
  2. Repeat the above to create more nodes, changing the LABEL= field in the cloud-init data to be a unique label.

Checking it Works

Congratulations! You now have an automatically expanding Kubernetes cluster!

Once the compute node has come up it will be automatically added to the accepted list of hosts for the cluster (see flight hunter list ) and will have the Kubernetes worker profile applied to it automatically (see flight profile list ).

What you can do next is:

  • Seamlessly access the cluster via the Flight Web Suite in a web browser by visiting the Public IPv4 DNS from AWS EC2 Console for the login node
  • Run some kubernetes pods
  • Remove unused nodes from the cluster before deleting them from AWS