Skip to content
Terraform Installer for Kubernetes on Oracle Cloud Infrastructure
Branch: master
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
.github Add a github issue template to help with reproducing issues. Nov 3, 2017
docs Fix bug in cloud-init script to get FQDN and update to a mroe recent … Dec 7, 2018
identity Add filestorage to flexdriver policy Feb 20, 2018
instances Fix bug in cloud-init script to get FQDN and update to a mroe recent … Dec 7, 2018
kubernetes Actually use private key (#138) Feb 7, 2018
network Fix bug in cloud-init script to get FQDN and update to a mroe recent … Dec 7, 2018
scripts migrate to kube-flannel Mar 28, 2018
tests Update image shape used by tests Dec 7, 2018
tls Make Load Balancer for Kubernetes Master(s) an optional parameter. Re… Feb 14, 2018
.gitignore Add to git ignore. Jan 30, 2018 initial commit Sep 9, 2017
LICENSE.txt initial commit Sep 9, 2017 Add an example app deployment with LB and volume. Jan 27, 2018 Add prefix to bash source Mar 20, 2018 Update Jan 9, 2018 Etcd was failing to update on AD1 or AD2 instances (#221) Jan 16, 2019 using separate subnets/security lists for NAT instances Feb 15, 2018 Rename all instances of "baremetal" provider to "oci". Oct 11, 2017
terraform.example.tfvars Remove references to BMCS from terraform.example.tfvars Apr 6, 2018 Fix bug in cloud-init script to get FQDN and update to a mroe recent … Dec 7, 2018
wercker.yml Revert changes to wercker.yml Apr 10, 2018

Terraform Kubernetes Installer for Oracle Cloud Infrastructure

wercker status


The Kubernetes Installer for Oracle Cloud Infrastructure provides a Terraform-based Kubernetes installation for Oracle Cloud Infrastructure. It consists of a set of Terraform modules and an example base configuration that is used to provision and configure the resources needed to run a highly available and configurable Kubernetes cluster on Oracle Cloud Infrastructure (OCI).

Cluster Overview

Terraform is used to provision the cloud infrastructure and any required local resources for the Kubernetes cluster including:

OCI Infrastructure

  • Virtual Cloud Network (VCN) with dedicated subnets for etcd, masters, and workers in each availability domain
  • Dedicated compute instances for etcd, Kubernetes master and worker nodes in each availability domain
  • Public or Private TCP/SSL OCI Load Balancer to distribute traffic to the Kubernetes Master(s)
  • Public or Private TCP/SSL OCI Load Balancer to distribute traffic to the node(s) in the etcd cluster
  • Optional NAT instance for Internet-bound traffic on any private subnets
  • 2048-bit SSH RSA Key-Pair for compute instances when not overridden by ssh_private_key and ssh_public_key_openssh input variables
  • Self-signed CA and TLS cluster certificates when not overridden by the input variables ca_cert, ca_key, etc.

Cluster Configuration

Terraform uses cloud-init scripts to handle the instance-level configuration for instances in the Control Plane to configure:

  • Highly Available (HA) Kubernetes master configuration
  • Highly Available (HA) etcd cluster configuration
  • Optional GPU support for worker nodes that need to run specific workloads
  • Kubernetes Dashboard and kube-DNS cluster add-ons
  • Kubernetes RBAC (role-based authorization control)
  • Integration with OCI Cloud Controller Manager (CCM)
  • Integration with OCI Flexvolume Driver

The Terraform scripts also accept a number of other input variables to choose instance shapes (including GPU) and how they are placed across the availability domain (ADs), etc. If your requirements extend beyond the base configuration, the modules can be used to form your own customized configuration.


  1. Download and install Terraform (v0.10.3 or later)
  2. Download and install the OCI Terraform Provider (v2.0.0 or later)
  3. Create an Terraform configuration file at ~/.terraformrc that specifies the path to the OCI provider:
providers {
  oci = "<path_to_provider_binary>/terraform-provider-oci"
  1. Ensure you have Kubectl installed if you plan to interact with the cluster locally
Optionally create separate IAM resources for OCI plugins

The OCI Cloud Controller Manager (CCM) and Volume Provisioner (VP) enables Kubernetes to dynamically provision OCI resources such as Load Balancers and Block Volumes as a part of pod and service creation. In order to facilitate this, OCI credentials and OCID information are automatically stored in the cluster as a Kubernetes Secret.

By default, the credentials of the user creating the cluster is used. However, in some cases, it makes sense to use a more restricted set of credentials whose policies are limited to a particular set of resources within the compartment.

To Terraform separate IAM users, groups, and policy resources, run the terraform plan and terraform apply commands from the identity directory and set the appropriate input variables relating to your custom users, fingerprints, and key paths.

Quick start

Customize the configuration

Create a terraform.tfvars file in the project root that specifies your configuration.

# start from the included example
$ cp terraform.example.tfvars terraform.tfvars
  • Set mandatory OCI input variables relating to your tenancy, user, and compartment.
  • Override optional input variables to customize the default configuration.

Deploy the cluster

Initialize Terraform:

$ terraform init

View what Terraform plans do before actually doing it:

$ terraform plan

Use Terraform to Provision resources and stand-up k8s cluster on OCI:

$ terraform apply

Access the cluster

The Kubernetes cluster will be running after the configuration is applied successfully and the cloud-init scripts have been given time to finish asynchronously. Typically, this takes around 5 minutes after terraform apply and will vary depending on the overall configuration, instance counts, and shapes.

A working kubeconfig can be found in the ./generated folder or generated on the fly using the kubeconfig Terraform output variable.

Your network access settings determine whether your cluster is accessible from the outside. See Accessing the Cluster for more details.

Verify the cluster:

If you've chosen to configure a public cluster, you can do a quick and automated verification of your cluster from your local machine by running the located in the scripts directory. Note that this script requires your KUBECONFIG environment variable to be set (above), and SSH and HTTPs access to be open to etcd and worker nodes.

To temporarily open access SSH and HTTPs access for, add the following to your terraform.tfvars file:

# warning: is wide open. remember to undo this.
etcd_ssh_ingress = ""
master_ssh_ingress = ""
worker_ssh_ingress = ""
master_https_ingress = ""
worker_nodeport_ingress = ""
$ scripts/
[] Running some basic checks on Kubernetes cluster....
[]   Checking ssh connectivity to each node...
[]   Checking whether instance bootstrap has completed on each node...
[]   Checking Flannel's etcd key from each node...
[]   Checking whether expected system services are running on each node...
[]   Checking status of /healthz endpoint at each k8s master node...
[]   Checking status of /healthz endpoint at the LB...
[]   Running 'kubectl get nodes' a number of times through the master LB...

The Kubernetes cluster is up and appears to be healthy.
Kubernetes master is running at
KubeDNS is running at
kubernetes-dashboard is running at

Deploy a simple load-balanced application with shared volumes

Check out the example application deployment for a walk through of deploying a simple application that leverages both the Cloud Controller Manager and Flexvolume Driver plugins.

Scale, upgrade, or delete the cluster

Check out the example cluster operations for details on how to use Terraform to scale, upgrade, replace, or delete your cluster.

Known issues and limitations

  • The OCI Load Balancers that gets created and attached to the VCN when a service of type --type=LoadBalancer is an out-of-band change to Terraform. As a result, the cluster's VCN will not be able to be destroyed until all services of type LoadBalancer have been deleted using kubectl or the OCI Console.
  • The OCI Block Volumes that gets created and attached to the workers when persistent volumes are create is also an out-of-band change to Terraform. As a result, the instances will not be able to be destroyed until all persistent volumes have been deleted using kubectl or the OCI Console.
  • Scaling or replacing etcd members in or out after the initial deployment is currently unsupported
  • Failover or HA configuration for NAT instance(s) is currently unsupported
  • Resizing the iSCSI volume will delete and recreate the volume
  • GPU Bare Metal instance shapes are currently only available in the Ashburn region and may be limited to specific availability domains
  • Provisioning a mix of GPU-enabled and non-GPU-enabled worker node instance shapes is currently unsupported


Tests run automatically on every commit to the main branch. Additionally, the tests should be run against any pull-request before it is merged.

See Testing for details.


This project is open source. Oracle appreciates any contributions that are made by the open source community.

See Contributing for details.

You can’t perform that action at this time.