This repository contains a Terraform module for creating an EKS cluster and all the necessary infrastructure to install Jenkins X via jx boot.

The module makes use of the Terraform EKS cluster Module.

• Jenkins X EKS Module
  • What is a Terraform module
  • How do you use this module
    • Prerequisites
    • Cluster provisioning
      • AWS_REGION
    • Cluster Autoscaling
    • Long Term Storage
    • Secrets Management
    • NGINX
    • ExternalDNS
    • cert-manager
    • Customer's CA certificates
    • Velero Backups
      • Enabling backups on pre-existing clusters
    • Production cluster considerations
    • Configuring a Terraform backend
    • Using Spot Instances
    • Worker Group Launch Templates
      • Enabling Worker Group Launch Templates
      • Transitioning from Worker Groups to Worker Groups Launch Templates
    • EKS node groups
      • Custom EKS node groups
    • AWS Auth
      • map_users
      • map_roles
      • map_accounts
    • Using SSH Key Pair
    • Using different EBS Volume type and size
      • Resizing a disk on existing nodes
    • Support for JX3
    • Existing VPC
    • Existing EKS cluster
    • Examples
    • Module configuration
      • Providers
      • Modules
      • Requirements
      • Inputs
      • Outputs
  • FAQ: Frequently Asked Questions
    • IAM Roles for Service Accounts
  • Development
    • Releasing
  • How can I contribute

A Terraform module refers to a self-contained package of Terraform configurations that are managed as a group. For more information about modules refer to the Terraform documentation.

This Terraform module allows you to create an EKS cluster ready for the installation of Jenkins X. You need the following binaries locally installed and configured on your PATH:

• terraform (=> 0.12.17, < 2.0.0)
• kubectl (>=1.10)
• aws-cli
• aws-iam-authenticator
• wget

A default Jenkins X ready cluster can be provisioned by creating a main.tf file in an empty directory with the following content:

module "eks-jx" {
  source = "jenkins-x/eks-jx/aws"
}

output "jx_requirements" {
  value = module.eks-jx.jx_requirements
}

output "vault_user_id" {
  value       = module.eks-jx.vault_user_id
  description = "The Vault IAM user id"
}

output "vault_user_secret" {
  value       = module.eks-jx.vault_user_secret
  description = "The Vault IAM user secret"
}

All s3 buckets created by the module use Server-Side Encryption with Amazon S3-Managed Encryption Keys (SSE-S3) by default. You can set the value of use_kms_s3 to true to use server-side encryption with AWS KMS (SSE-KMS). If you don't specify the value of s3_kms_arn, then the default aws managed cmk is used (aws/s3)

⚠️ Note: Using AWS KMS with customer managed keys has cost considerations.

Due to the Vault issue 7450, this Terraform module needs for now to create a new IAM user for installing Vault. It also creates an IAM access key whose id and secret are defined in the output above. You need the id and secret for running jx boot.

The jx_requirements output is a helper for creating the initial input for jx boot.

If you do not want Terraform to create a new IAM user or you do not have permissions to create one, you need to provide the name of an existing IAM user.

module "eks-jx" {
  source     = "jenkins-x/eks-jx/aws"
  vault_user = "<your_vault_iam_username>"
}

You should have your AWS CLI configured correctly.

In addition, you should make sure to specify the region via the AWS_REGION environment variable. e.g. export AWS_REGION=us-east-1 and the region variable (make sure the region variable matches the environment variable)

The IAM user does not need any permissions attached to it. For more information, refer to Configuring Vault for EKS in the Jenkins X documentation.

Once you have your initial configuration, you can apply it by running:

terraform init
terraform apply

This creates an EKS cluster with all possible configuration options defaulted.

You then need to export the environment variables VAULT_AWS_ACCESS_KEY_ID and VAULT_AWS_SECRET_ACCESS_KEY.

export VAULT_AWS_ACCESS_KEY_ID=$(terraform output vault_user_id)
export VAULT_AWS_SECRET_ACCESS_KEY=$(terraform output vault_user_secret)

If you specified vault_user you need to provide the access key id and secret for the specified user.

⚠️ Note: This example is for getting up and running quickly. It is not intended for a production cluster. Refer to Production cluster considerations for things to consider when creating a production cluster.

This does not automatically install cluster-autoscaler, it installs all of the prerequisite policies and roles required to install autoscaler. The actual autoscaler installation varies depending on what version of kubernetes you are using.

To install cluster autoscaler, first you will need the ARN of the cluster-autoscaler role.

You can create the following output along side your module definition to find this:

output "cluster_autoscaler_iam_role_arn" {
  value = module.eks-jx.cluster_autoscaler_iam_role.this_iam_role_arn
}

With the ARN, you may now install the cluster autoscaler using Helm.

Create the file cluster-autoscaler-values.yaml with the following content:

awsRegion: us-east-1

rbac:
  serviceAccount:
    name: cluster-autoscaler
    annotations:
      eks.amazonaws.com/role-arn: "arn:aws:iam::12345678910:role/tf-your-cluster-name-cluster-autoscaler"

autoDiscovery:
  clusterName: your-cluster-name

image:
  repository: us.gcr.io/k8s-artifacts-prod/autoscaling/cluster-autoscaler
  tag: v1.19.1

Notice the image tag is v1.19.1 - this tag goes with clusters running Kubernetes 1.19. If you are running 1.20, 1.21, etc, you will need to find the image tag that matches your cluster version. To see available tags, visit this GCR registry

Next, you'll need to fetch the chart, apply your values using helm template and then apply the resulting Kubernetes object to your cluster.

helm fetch stable/cluster-autoscaler --untar

And then

helm template --name cluster-autoscaler --namespace kube-system ./cluster-autoscaler -f ./cluster-autoscaler-values.yaml | kubectl apply -n kube-system -f -

You can choose to create S3 buckets for long term storage of Jenkins X build artefacts with enable_logs_storage, enable_reports_storage and enable_repository_storage.

During terraform apply the enabledS3 buckets are created, and the jx_requirements output will contain the following section:

storage:
  logs:
    enabled: ${enable_logs_storage}
    url: s3://${logs_storage_bucket}
  reports:
    enabled: ${enable_reports_storage}
    url: s3://${reports_storage_bucket}
  repository:
    enabled: ${enable_repository_storage}
    url: s3://${repository_storage_bucket}

If you just want to experiment with Jenkins X, you can set the variable force_destroy to true. This allows you to remove all generated buckets when running terraform destroy.

⚠️ Note: If you set force_destroy to false, and run a terraform destroy, it will fail. In that case empty the s3 buckets from the aws s3 console, and re run terraform destroy.

⚠️ Note: A notice from Amazon: Amazon S3 will automatically enable S3 Block Public Access and disable access control lists for all new buckets starting in April 2023. To accomodate this acl setting was removed for buckets and the enable_acl variable was introduced and set to false (default). If the requirement is to provide ACL with bucket ownership conrols for the bucket, then set the enable_acl variable to true.

Vault is the default tool used by Jenkins X for managing secrets. Part of this module's responsibilities is the creation of all resources required to run the Vault Operator. These resources are An S3 Bucket, a DynamoDB Table and a KMS Key.

You can also configure an existing Vault instance for use with Jenkins X. In this case

• provide the Vault URL via the vault_url input variable
• set the boot_secrets in main.tf to this value:

boot_secrets = [
    {
      name  = "jxBootJobEnvVarSecrets.EXTERNAL_VAULT"
      value = "true"
      type  = "string"
    },
    {
      name  = "jxBootJobEnvVarSecrets.VAULT_ADDR"
      value = "https://enter-your-vault-url:8200"
      type  = "string"
    }
  ]

• follow the Jenkins X documentation around the installation of an external Vault instance.

To use AWS Secrets Manager instead of vault, set use_vault variable to false, and use_asm variable to true. You will also need a role that grants access to AWS Secrets Manager, this will be created for you by setting create_asm_role variable to true. Setting the above variables will add the asm role arn to the boot job service account, which is required for the boot job to interact with AWS secrets manager to populate secrets.

The module can install the nginx chart by setting create_nginx flag to true. Example can be found here. You can specify a nginx_values.yaml file or the module will use the default one stored here. If you are using terraform to create nginx resources, do not use the chart specified in the versionstream. Remove the entry in the helmfile.yaml referencing the nginx chart

path: helmfiles/nginx/helmfile.yaml

You can enable ExternalDNS with the enable_external_dns variable. This modifies the generated jx-requirements.yml file to enable External DNS when running jx boot.

If enable_external_dns is true, additional configuration is required.

If you want to use a domain with an already existing Route 53 Hosted Zone, you can provide it through the apex_domain variable:

This domain will be configured in the jx_requirements output in the following section:

ingress:
  domain: ${domain}
  ignoreLoadBalancer: true
  externalDNS: ${enable_external_dns}

If you want to use a subdomain and have this module create and configure a new Hosted Zone with DNS delegation, you can provide the following variables:

subdomain: This subdomain is added to the apex domain and configured in the resulting jx-requirements.yml file.

create_and_configure_subdomain: This flag instructs the script to create a new Route53 Hosted Zone for your subdomain and configure DNS delegation with the apex domain.

By providing these variables, the script creates a new Route 53 HostedZone that looks like <subdomain>.<apex_domain>, then it delegates the resolving of DNS to the apex domain. This is done by creating a NS RecordSet in the apex domain's Hosted Zone with the subdomain's HostedZone nameservers.

This ensures that the newly created HostedZone for the subdomain is instantly resolvable instead of having to wait for DNS propagation.

You can enable cert-manager to use TLS for your cluster through LetsEncrypt with the enable_tls variable.

LetsEncrypt has two environments, staging and production.

If you use staging, you will receive self-signed certificates, but you are not rate-limited, if you use the production environment, you receive certificates signed by LetsEncrypt, but you can be rate limited.

You can choose to use the production environment with the production_letsencrypt variable:

You need to provide a valid email to register your domain in LetsEncrypt with tls_email.

Customer has got signed certificates from CA and want to use it instead of LetsEncrypt certificates. Terraform creates k8s tls-ingress-certificates-ca secret with tls_key and tls_cert in default namespace. User should define:

enable_external_dns = true
apex_domain         = "office.com"
subdomain           = "subdomain"
enable_tls          = true
tls_email           = "[email protected]"

// Signed Certificate must match the domain: *.subdomain.office.com
tls_cert            = "/opt/CA/cert.crt"
tls_key             = "LS0tLS1C....BLRVktLS0tLQo="

This module can set up the resources required for running backups with Velero on your cluster by setting the flag enable_backup to true.

If your cluster is pre-existing and already contains a namespace named velero, then enabling backups will initially fail with an error that you are trying to create a namespace which already exists.

Error: namespaces "velero" already exists

If you get this error, consider it a warning - you may then adjust accordingly by importing that namespace to be managed by Terraform, deleting the previously existing ns if it wasn't actually in use, or setting enable_backup back to false to continue managing Velero in the previous manner.

The recommended way is to import the namespace and then run another Terraform plan and apply:

terraform import module.eks-jx.module.backup.kubernetes_namespace.velero_namespace velero

The configuration, as seen in Cluster provisioning, is not suited for creating and maintaining a production Jenkins X cluster. The following is a list of considerations for a production use case.

• Specify the version attribute of the module, for example:

  module "eks-jx" {
    source  = "jenkins-x/eks-jx/aws"
    version = "1.0.0"
    # insert your configuration
  }
  
  output "jx_requirements" {
    value = module.eks-jx.jx_requirements
  }

  Specifying the version ensures that you are using a fixed version and that version upgrades cannot occur unintended.

• Keep the Terraform configuration under version control by creating a dedicated repository for your cluster configuration or by adding it to an already existing infrastructure repository.

• Setup a Terraform backend to securely store and share the state of your cluster. For more information refer to Configuring a Terraform backend.

• Disable public API for the EKS cluster. If that is not not possible, restrict access to it by specifying the cidr blocks which can access it.

A "backend" in Terraform determines how state is loaded and how an operation such as apply is executed. By default, Terraform uses the local backend, which keeps the state of the created resources on the local file system. This is problematic since sensitive information will be stored on disk and it is not possible to share state across a team. When working with AWS a good choice for your Terraform backend is the s3 backend which stores the Terraform state in an AWS S3 bucket. The examples directory of this repository contains configuration examples for using the s3 backed.

To use the s3 backend, you will need to create the bucket upfront. You need the S3 bucket as well as a Dynamo table for state locks. You can use terraform-aws-tfstate-backend to create these required resources.

You can save up to 90% of cost when you use Spot Instances. You just need to make sure your applications are resilient. You can set the ceiling spot_price of what you want to pay then set enable_spot_instances to true.

⚠️ Note: If the price of the instance reaches this point it will be terminated.

Worker Groups, the default worker node groups for this module, are based on an older AWS tool called "Launch Configurations" which have some limitations around Spot instances and delegating a percentage of a pool of workers to on-demand or spot instances, as well as issues when autoscaling is enabled.

The issue with autoscaling with the default worker group is that it is prone to autoscaling using Nodes from only a single AZ. AWS has a "AZRebalance" job that can run to help with this, but it is aggressive in removing nodes.

All of these issues can be resolved by using Worker Group Launch Templates instead, configured with a template for each Availability Zone. Using an ASG for each AZ bypasses the autoscaling issues in AWS. Furthermore, we are also able to specify several types of machines that are suitable for spot instances rather than just one. Using only one often results in Spot instances not being able to be provisioned, and this greatly reduces the occurence of this happening, as well as allowing for lower spot prices.

To use the Worker Group Launch Template, set the variable enable_worker_groups_launch_template to true, and define an array of instance types allowed.

When using autoscaling with Launch Templates per AZ, the min and max number of nodes is per zone. These values can be adjusted by using the variables lt_desired_nodes_per_subnet, lt_min_nodes_per_subnet, and lt_max_nodes_per_subnet

module "eks-jx" {
  source  = "jenkins-x/eks-jx/aws"
  enable_worker_groups_launch_template = true
  allowed_spot_instance_types          = ["m5.large", "m5a.large", "m5d.large", "m5ad.large", "t3.large", "t3a.large"]
  lt_desired_nodes_per_subnet          = 2
  lt_min_nodes_per_subnet              = 2
  lt_max_nodes_per_subnet              = 3
}

In order to prevent any interruption to service, you'll first want to enable Worker Group Launch Templates.

Once you've verified that you are able to see the new Nodes created by the Launch Templates by running kubectl get nodes, then you can remove the older Worker Group.

To remove the older worker group, it's recommended to first scale down to zero nodes, one at a time, by adjusting the min/max node capacity. Once you've scaled down to zero nodes for the original worker group, and your workloads have been scheduled on nodes created by the launch templates you can set enable_worker_group to false.

module "eks-jx" { source = "jenkins-x/eks-jx/aws" enable_worker_group = false enable_worker_groups_launch_template = true allowed_spot_instance_types = ["m5.large", "m5a.large", "m5d.large", "m5ad.large", "t3.large", "t3a.large"] lt_desired_nodes_per_subnet = 2 lt_min_nodes_per_subnet = 2 lt_max_nodes_per_subnet = 3 }

This module provisions self-managed worker nodes by default.

If you want AWS to manage the provisioning and lifecycle of worker nodes for EKS, you can opt for managed node groups.

They have the added benefit of running the latest Amazon EKS-optimized AMIs and gracefully drain nodes before termination to ensure that your applications stay available.

In order to provision EKS node groups create a main.tf with the following content:

module "eks-jx" {
  source  = "jenkins-x/eks-jx/aws"
  enable_worker_group = false
}

output "jx_requirements" {
  value = module.eks-jx.jx_requirements
}

output "vault_user_id" {
  value       = module.eks-jx.vault_user_id
  description = "The Vault IAM user id"
}

output "vault_user_secret" {
  value       = module.eks-jx.vault_user_secret
  description = "The Vault IAM user secret"
}

Note: EKS node groups now support using spot instances and launch templates (will be set accordingly with the use of the enable_spot_instances variable)

A single node group will be created by default when using EKS node groups. Supply values for the node_groups_managed variable to override this behaviour:

module "eks-jx" {
  source  = "jenkins-x/eks-jx/aws"
  enable_worker_group = false
  node_groups_managed = {
    node-group-name = {
      ami_type                = "AL2_x86_64"
      disk_size               = 50
      desired_capacity        = 3
      max_capacity            = 5
      min_capacity            = 3
      instance_types          = [ "m5.large" ]
      launce_template_id      = null
      launch_template_version = null

      k8s_labels = {
        purpose = "application"
      }
    },
    second-node-group-name = {
      # ...
    },
    # ...
  }
}

One can use launch templates with node groups by specifying the template id and version in the parameters.

resource "aws_launch_template" "foo" {
  name = "foo"
  # ...
}

module "eks-jx" {
  source  = "jenkins-x/eks-jx/aws"
  enable_worker_group = false
  node_groups_managed = {
    node-group-name = {
      ami_type                = "AL2_x86_64"
      disk_size               = 50
      desired_capacity        = 3
      max_capacity            = 5
      min_capacity            = 3
      instance_types          = [ "m5.large" ]
      launce_template_id      = aws_launch_template.foo.id
      launch_template_version = aws_launch_template.foo.latest_version

      k8s_labels = {
        purpose = "application"
      }
    },
    second-node-group-name = {
      # ...
    },
    # ...
  }
}

⚠️ Note: EKS node groups are supported in kubernetes v1.14+ and platform version eks.3

When running EKS, authentication for the cluster is controlled by a configmap called aws-auth. By default, that should look something like this:

apiVersion: v1
data:
  mapAccounts: |
    []
  mapRoles: |
    - "groups":
      - "system:bootstrappers"
      - "system:nodes"
      "rolearn": "arn:aws:iam::777777777777:role/project-eks-12345"
      "username": "system:node:{{EC2PrivateDNSName}}"
  mapUsers: |
    []
kind: ConfigMap
metadata:
  name: aws-auth
  namespace: kube-system

When using this Terraform module, this AWS Auth configmap is generated for you via the EKS module that is used internally. Additional users, roles, and accounts may be mapped into this config map by providing the variables map_users, map_roles or map_accounts respectively.

To add an additional AWS IAM user named "patrick", you can create an aws_iam_user resource, and then use the map_users variable to allow Patrick to access EKS:

resource "aws_iam_user" "patrick" {
  name = "patrick"
}

module "eks-jx" {
  source  = "jenkins-x/eks-jx/aws"
  map_users = [
    {
      userarn  = aws_iam_user.patrick.arn
      username = aws_iam_user.patrick.name
      groups   = ["system:masters"]
    }
  ]
}

To map additional roles to the AWS Auth ConfigMap, use map_roles:

module "eks-jx" {
  source  = "jenkins-x/eks-jx/aws"
  map_roles = [
    {
      rolearn  = "arn:aws:iam::66666666666:role/role1"
      username = "role1"
      groups   = ["system:masters"]
    },
  ]
}

To map additional accounts to the AWS Auth ConfigMap, use map_accounts:

module "eks-jx" {
  source  = "jenkins-x/eks-jx/aws"
  map_accounts = [
    "777777777777",
    "888888888888",
  ]
}

Import a key pair or use an existing one and take note of the name. Set key_name and set enable_key_name to true.

Set volume_type to either standard, gp2 or io1 and volume_size to the desired size in GB. If chosing io1 set desired iops.

The existing nodes needs to be terminated and replaced with new ones if disk is needed to be resized. You need to execute the following command before terraform apply in order to replace the Auto Scaling Launch Configuration.

terraform taint module.eks-jx.module.cluster.module.eks.aws_launch_configuration.workers[0]

Creation of namespaces and service accounts using terraform is no longer required for JX3. To keep compatibility with JX2, a flag is_jx2 was introduced, in v1.6.0.

If you want to create the cluster in an existing VPC you can specify create_vpc to false and specify where to create the clsuter with vpc_id and subnets.

It is very common to have another module used to create EKS clusters for all your AWS accounts, in that case, you can set create_eks and create_vpc to false and cluster_name to the id/name of the EKS cluster where jx components need to be installed in. This will prevent creating a new vpc and eks cluster for jx. There are also flags to control the creation of IAM roles. See this for a complete example.

You can find examples for different configurations in the examples folder.

Each example generates a valid jx-requirements.yml file that can be used to boot a Jenkins X cluster.

list(object({
    name  = string
    value = string
    type  = string
  }))

list(object({
    provider_key_arn = string
    resources        = list(string)
  }))

[
  "0.0.0.0/0"
]

[
  "0.0.0.0/0"
]

[
  "/bin/bash",
  "-c"
]

list(object({
    rolearn  = string
    username = string
    groups   = list(string)
  }))

list(object({
    userarn  = string
    username = string
    groups   = list(string)
  }))

{
  "eks-jx-node-group": {}
}

[
  "10.0.4.0/24",
  "10.0.5.0/24",
  "10.0.6.0/24"
]

[
  "10.0.1.0/24",
  "10.0.2.0/24",
  "10.0.3.0/24"
]

This module sets up a series of IAM Policies and Roles. These roles will be annotated into a few Kubernetes Service accounts. This allows us to make use of IAM Roles for Sercive Accounts to set fine-grained permissions on a pod per pod basis. There is no way to provide your own roles or define other Service Accounts by variables, but you can always modify the modules/cluster/irsa.tf Terraform file.

At the moment, there is no release pipeline defined in jenkins-x.yml. A Terraform release does not require building an artifact; only a tag needs to be created and pushed. To make this task easier and there is a helper script release.sh which simplifies this process and creates the changelog as well:

./scripts/release.sh

This can be executed on demand whenever a release is required. For the script to work, the environment variable $GH_TOKEN must be exported and reference a valid GitHub API token.

Contributions are very welcome! Check out the Contribution Guidelines for instructions.