1. Documentation
×

Installation requirements for user-provisioned infrastructure on AWS

Before you install OKD on infrastructure that you provision, ensure that your Amazon Web Services (AWS) environment meets the installation requirements.

For a cluster that has user-provisioned infrastructure, you must deploy all of the required machines.

Required machines for cluster installation

You must specify the minimum required machines or hosts for your cluster so that your cluster remains stable if a node fails.

The smallest OKD clusters require the following hosts:

For a cluster that contains user-provisioned infrastructure, you must deploy all of the required machines.
Table 1. Minimum required hosts
Hosts Description

One temporary bootstrap machine

The cluster requires the bootstrap machine to deploy the OKD cluster on the three control plane machines. You can remove the bootstrap machine after you install the cluster.

Three control plane machines

The control plane machines run the Kubernetes and OKD services that form the control plane.

At least two compute machines, which are also known as worker machines.

The workloads requested by OKD users run on the compute machines.

To maintain high availability of your cluster, use separate physical hosts for these cluster machines.

The bootstrap and control plane machines must use Fedora CoreOS (FCOS) as the operating system. However, the compute machines can choose between Fedora CoreOS (FCOS), Fedora 8.6 and later.

See Red Hat Enterprise Linux technology capabilities and limits.

Minimum resource requirements for cluster installation

Each created cluster must meet minimum requirements so that the cluster runs as expected.

Table 2. Minimum resource requirements
Machine Operating System vCPU Virtual RAM Storage Input/Output Per Second (IOPS)

Bootstrap

FCOS

4

16 GB

100 GB

300

Control plane

FCOS

4

16 GB

100 GB

300

Compute

FCOS

2

8 GB

100 GB

300

  • One vCPU is equivalent to one physical core when simultaneous multithreading (SMT), or Hyper-Threading, is not enabled. When enabled, use the following formula to calculate the corresponding ratio: (threads per core × cores) × sockets = vCPUs.

  • OKD and Kubernetes are sensitive to disk performance, and faster storage is recommended, particularly for etcd on the control plane nodes which require a 10 ms p99 fsync duration. Note that on many cloud platforms, storage size and IOPS scale together, so you might need to over-allocate storage volume to obtain sufficient performance.

  • As with all user-provisioned installations, if you choose to use Fedora compute machines in your cluster, you take responsibility for all operating system life cycle management and maintenance, including performing system updates, applying patches, and completing all other required tasks. Use of Fedora 7 compute machines is deprecated and has been removed in OKD 4.10 and later.

For OKD version 4.22, FCOS is based on Fedora version 9.8, which has the micro-architecture requirements. The following list contains the minimum instruction set architectures (ISA) that each architecture requires:

  • x86-64 architecture requires x86-64-v2 ISA

  • ARM64 architecture requires ARMv8.0-A ISA

  • ppc64le architecture requires IBM® Power9 ISA

  • s390x architecture requires IBM® z14 ISA

For more information, see Architectures in the Fedora documentation.

If an instance type for your platform meets the minimum requirements for cluster machines, it is supported to use in OKD.

Additional resources

Tested instance types for AWS

There are several Amazon Web Services (AWS) instance types tested with OKD.

The following AWS instance types have been tested with OKD.

Use the machine types included in the following charts for your AWS instances. If you use an instance type that is not listed in the chart, ensure that the instance size you use matches the minimum resource requirements that are listed in the section named "Minimum resource requirements for cluster installation".

See the following machine types based on 64-bit x86 architecture:

  • c4.*

  • c5.*

  • c5a.*

  • i3.*

  • m4.*

  • m5.*

  • m5a.*

  • m6a.*

  • m6i.*

  • m7a.*

  • m7i.*

  • m7i-flex.*

  • r4.*

  • r5.*

  • r5a.*

  • r6i.*

  • t3.*

  • t3a.*

Tested instance types for AWS on 64-bit ARM infrastructures

There are several Amazon Web Services (AWS) 64-bit ARM instance types tested with OKD.

The following AWS 64-bit ARM instance types have been tested with OKD.

Use the machine types included in the following charts for your AWS ARM instances. If you use an instance type that is not listed in the chart, ensure that the instance size you use matches the minimum resource requirements that are listed in "Minimum resource requirements for cluster installation".

See the following machine types based on 64-bit ARM architecture:

  • c6g.*

  • c7g.*

  • c8g.*

  • m6g.*

  • m7g.*

  • m7gd.*

  • r8g.*

Certificate signing requests management

On user-provisioned infrastructure, you must provide a mechanism for approving cluster certificate signing requests (CSRs) after installation when your cluster has limited access to automatic machine management.

The kube-controller-manager only approves the kubelet client CSRs. The machine-approver cannot guarantee the validity of a serving certificate that is requested by using kubelet credentials because it cannot confirm that the correct machine issued the request. You must determine and implement a method of verifying the validity of the kubelet serving certificate requests and approving them.

Required AWS infrastructure components

To install OKD on user-provisioned infrastructure in Amazon Web Services (AWS), you must manually create both the machines and their supporting infrastructure.

For more information about the integration testing for different platforms, see the OpenShift Container Platform 4.x Tested Integrations page.

By using the provided CloudFormation templates, you can create stacks of AWS resources that represent the following components:

  • An AWS Virtual Private Cloud (VPC)

  • Networking and load balancing components

  • Security groups and roles

  • An OKD bootstrap node

  • OKD control plane nodes

  • An OKD compute node

Or, you can manually create the components or you can reuse existing infrastructure that meets the cluster requirements. Review the CloudFormation templates for more details about how the components interrelate.

Other infrastructure components

Your OKD cluster on user-provisioned infrastructure in Amazon Web Services (AWS) requires additional infrastructure components, including a VPC, DNS entries, load balancers, security groups, and IAM roles.

  • A VPC

  • DNS entries

  • Load balancers (classic or network) and listeners

  • A public and a private Route 53 zone

  • Security groups

  • IAM roles

  • S3 buckets

If you are working in a disconnected environment, you cannot reach the public IP addresses for EC2, ELB, and S3 endpoints. Depending on the level to which you want to restrict internet traffic during the installation, the following configuration options are available:

Option 1: Create VPC endpoints

Create a VPC endpoint and attach it to the subnets that the clusters are using. Name the endpoints as follows:

  • ec2.<aws_region>.amazonaws.com

  • elasticloadbalancing.<aws_region>.amazonaws.com

  • s3.<aws_region>.amazonaws.com

With this option, network traffic remains private between your VPC and the required AWS services.

Option 2: Create a proxy without VPC endpoints

As part of the installation process, you can configure an HTTP or HTTPS proxy. With this option, internet traffic goes through the proxy to reach the required AWS services.

Option 3: Create a proxy with VPC endpoints

As part of the installation process, you can configure an HTTP or HTTPS proxy with VPC endpoints. Create a VPC endpoint and attach it to the subnets that the clusters are using. Name the endpoints as follows:

  • ec2.<aws_region>.amazonaws.com

  • elasticloadbalancing.<aws_region>.amazonaws.com

  • s3.<aws_region>.amazonaws.com

When configuring the proxy in the install-config.yaml file, add these endpoints to the noProxy field. With this option, the proxy prevents the cluster from accessing the internet directly. However, network traffic remains private between your VPC and the required AWS services.

You must provide a suitable VPC and subnets that allow communication to your machines.

Table 3. Required VPC components
Component AWS type Description

VPC

  • AWS::EC2::VPC

  • AWS::EC2::VPCEndpoint

You must provide a public VPC for the cluster to use. The VPC uses an endpoint that references the route tables for each subnet to improve communication with the registry that is hosted in S3.

Public subnets

  • AWS::EC2::Subnet

  • AWS::EC2::SubnetNetworkAclAssociation

Your VPC must have public subnets for between 1 and 3 availability zones and associate them with appropriate Ingress rules.

Internet gateway

  • AWS::EC2::InternetGateway

  • AWS::EC2::VPCGatewayAttachment

  • AWS::EC2::RouteTable

  • AWS::EC2::Route

  • AWS::EC2::SubnetRouteTableAssociation

  • AWS::EC2::NatGateway

  • AWS::EC2::EIP

You must have a public internet gateway, with public routes, attached to the VPC. In the provided templates, each public subnet has a NAT gateway with an EIP address. These NAT gateways allow cluster resources, such as private subnet instances, to reach the internet and are not required for some restricted network or proxy scenarios.

Network access control

  • AWS::EC2::NetworkAcl

  • AWS::EC2::NetworkAclEntry

You must allow the VPC to access the following ports:

Port

Reason

80

Inbound HTTP traffic

443

Inbound HTTPS traffic

22

Inbound SSH traffic

1024 - 65535

Inbound ephemeral traffic

0 - 65535

Outbound ephemeral traffic

Private subnets

  • AWS::EC2::Subnet

  • AWS::EC2::RouteTable

  • AWS::EC2::SubnetRouteTableAssociation

Your VPC can have private subnets. The provided CloudFormation templates can create private subnets for between 1 and 3 availability zones. If you use private subnets, you must provide appropriate routes and tables for them.

Your DNS and load balancer configuration must use a public hosted zone and can use a private hosted zone similar to the one that the installation program uses if it provisions the cluster’s infrastructure. You must create a DNS entry that resolves to your load balancer. An entry for api.<cluster_name>.<domain> must point to the external load balancer, and an entry for api-int.<cluster_name>.<domain> must point to the internal load balancer.

The cluster also requires load balancers and listeners for port 6443, which the Kubernetes API and its extensions require, and port 22623, which the Ignition config files for new machines require. The targets are the control plane nodes. Port 6443 must be accessible to both clients external to the cluster and nodes within the cluster. Port 22623 must be accessible to nodes within the cluster.

Table 4. Required DNS and load balancing components
Component AWS type Description

DNS

AWS::Route53::HostedZone

The hosted zone for your internal DNS.

Public load balancer

AWS::ElasticLoadBalancingV2::LoadBalancer

The load balancer for your public subnets.

External API server record

AWS::Route53::RecordSetGroup

Alias records for the external API server.

External listener

AWS::ElasticLoadBalancingV2::Listener

A listener on port 6443 for the external load balancer.

External target group

AWS::ElasticLoadBalancingV2::TargetGroup

The target group for the external load balancer.

Private load balancer

AWS::ElasticLoadBalancingV2::LoadBalancer

The load balancer for your private subnets.

Internal API server record

AWS::Route53::RecordSetGroup

Alias records for the internal API server.

Internal listener

AWS::ElasticLoadBalancingV2::Listener

A listener on port 22623 for the internal load balancer.

Internal target group

AWS::ElasticLoadBalancingV2::TargetGroup

The target group for the internal load balancer.

Internal listener

AWS::ElasticLoadBalancingV2::Listener

A listener on port 6443 for the internal load balancer.

Internal target group

AWS::ElasticLoadBalancingV2::TargetGroup

The target group for the internal load balancer.

The control plane and worker machines require access to the following ports:

Table 5. Security groups
Group Type IP Protocol Port range

MasterSecurityGroup

AWS::EC2::SecurityGroup

icmp

0

tcp

22

tcp

6443

tcp

22623

WorkerSecurityGroup

AWS::EC2::SecurityGroup

icmp

0

tcp

22

BootstrapSecurityGroup

AWS::EC2::SecurityGroup

tcp

22

tcp

19531

The control plane machines require the following Ingress groups. Each Ingress group is an AWS::EC2::SecurityGroupIngress resource.

Table 6. Control plane Ingress
Ingress group Description IP protocol Port range

MasterIngressEtcd

etcd

tcp

2379- 2380

MasterIngressVxlan

Vxlan packets

udp

6081

MasterIngressWorkerVxlan

Vxlan packets

udp

6081

MasterIngressInternal

Internal cluster communication and Kubernetes proxy metrics

tcp

9000 - 9999

MasterIngressWorkerInternal

Internal cluster communication

tcp

9000 - 9999

MasterIngressKube

Kubernetes kubelet, scheduler and controller manager

tcp

10250 - 10259

MasterIngressWorkerKube

Kubernetes kubelet, scheduler and controller manager

tcp

10250 - 10259

MasterIngressIngressServices

Kubernetes Ingress services

tcp

30000 - 32767

MasterIngressWorkerIngressServices

Kubernetes Ingress services

tcp

30000 - 32767

MasterIngressGeneve

Geneve packets

udp

6081

MasterIngressWorkerGeneve

Geneve packets

udp

6081

MasterIngressIpsecIke

IPsec IKE packets

udp

500

MasterIngressWorkerIpsecIke

IPsec IKE packets

udp

500

MasterIngressIpsecNat

IPsec NAT-T packets

udp

4500

MasterIngressWorkerIpsecNat

IPsec NAT-T packets

udp

4500

MasterIngressIpsecEsp

IPsec ESP packets

50

All

MasterIngressWorkerIpsecEsp

IPsec ESP packets

50

All

MasterIngressInternalUDP

Internal cluster communication

udp

9000 - 9999

MasterIngressWorkerInternalUDP

Internal cluster communication

udp

9000 - 9999

MasterIngressIngressServicesUDP

Kubernetes Ingress services

udp

30000 - 32767

MasterIngressWorkerIngressServicesUDP

Kubernetes Ingress services

udp

30000 - 32767

The worker machines require the following Ingress groups. Each Ingress group is an AWS::EC2::SecurityGroupIngress resource.

Table 7. Worker Ingress
Ingress group Description IP protocol Port range

WorkerIngressVxlan

Vxlan packets

udp

6081

WorkerIngressWorkerVxlan

Vxlan packets

udp

6081

WorkerIngressInternal

Internal cluster communication

tcp

9000 - 9999

WorkerIngressWorkerInternal

Internal cluster communication

tcp

9000 - 9999

WorkerIngressKube

Kubernetes kubelet, scheduler, and controller manager

tcp

10250

WorkerIngressWorkerKube

Kubernetes kubelet, scheduler, and controller manager

tcp

10250

WorkerIngressIngressServices

Kubernetes Ingress services

tcp

30000 - 32767

WorkerIngressWorkerIngressServices

Kubernetes Ingress services

tcp

30000 - 32767

WorkerIngressGeneve

Geneve packets

udp

6081

WorkerIngressMasterGeneve

Geneve packets

udp

6081

WorkerIngressIpsecIke

IPsec IKE packets

udp

500

WorkerIngressMasterIpsecIke

IPsec IKE packets

udp

500

WorkerIngressIpsecNat

IPsec NAT-T packets

udp

4500

WorkerIngressMasterIpsecNat

IPsec NAT-T packets

udp

4500

WorkerIngressIpsecEsp

IPsec ESP packets

50

All

WorkerIngressMasterIpsecEsp

IPsec ESP packets

50

All

WorkerIngressInternalUDP

Internal cluster communication

udp

9000 - 9999

WorkerIngressMasterInternalUDP

Internal cluster communication

udp

9000 - 9999

WorkerIngressIngressServicesUDP

Kubernetes Ingress services

udp

30000 - 32767

WorkerIngressMasterIngressServicesUDP

Kubernetes Ingress services

udp

30000 - 32767

You must grant the machines permissions in AWS. The provided CloudFormation templates grant the machines Allow permissions for the following AWS::IAM::Role objects and provide an AWS::IAM::InstanceProfile for each set of roles. If you do not use the templates, you can grant the machines the following broad permissions or the following individual permissions.

Table 8. Roles and instance profiles
Role Effect Action Resource

Control plane

Allow

ec2:*

*

Allow

elasticloadbalancing:*

*

Allow

iam:PassRole

*

Allow

s3:GetObject

*

Worker

Allow

ec2:Describe*

*

Bootstrap

Allow

ec2:Describe*

*

Allow

ec2:AttachVolume

*

Allow

ec2:DetachVolume

*

Cluster machines

Your OKD cluster on user-provisioned infrastructure requires AWS::EC2::Instance objects for bootstrap, control plane, and compute machines.

You need AWS::EC2::Instance objects for the following machines:

  • A bootstrap machine. This machine is required during installation, but you can remove it after your cluster deploys.

  • Three control plane machines. A control plane machine set does not govern the control plane machines.

  • Compute machines. You must create at least two compute machines, which are also known as worker machines, during installation. A compute machine set does not govern these machines.

Required AWS permissions for the IAM user

To deploy all components of an OKD cluster, you must grant the all the required permissions to the IAM user that you create in Amazon Web Services (AWS).

Your IAM user must have the permission tag:GetResources in the region us-east-1 to delete the base cluster resources. As part of the AWS API requirement, the OKD installation program performs various actions in this region.

When you attach the AdministratorAccess policy to the IAM user that you create in Amazon Web Services (AWS), you grant that user all of the required permissions. To deploy all components of an OKD cluster, the IAM user requires the following permissions:

Required EC2 permissions for installation

  • ec2:AttachNetworkInterface

  • ec2:AuthorizeSecurityGroupEgress

  • ec2:AuthorizeSecurityGroupIngress

  • ec2:CopyImage

  • ec2:CreateNetworkInterface

  • ec2:CreateSecurityGroup

  • ec2:CreateTags

  • ec2:CreateVolume

  • ec2:DeleteSecurityGroup

  • ec2:DeleteSnapshot

  • ec2:DeleteTags

  • ec2:DeregisterImage

  • ec2:DescribeAccountAttributes

  • ec2:DescribeAddresses

  • ec2:DescribeAvailabilityZones

  • ec2:DescribeDhcpOptions

  • ec2:DescribeImages

  • ec2:DescribeInstanceAttribute

  • ec2:DescribeInstanceCreditSpecifications

  • ec2:DescribeInstances

  • ec2:DescribeInstanceTypes

  • ec2:DescribeInstanceTypeOfferings

  • ec2:DescribeInternetGateways

  • ec2:DescribeKeyPairs

  • ec2:DescribeNatGateways

  • ec2:DescribeNetworkAcls

  • ec2:DescribeNetworkInterfaces

  • ec2:DescribePrefixLists

  • ec2:DescribePublicIpv4Pools (only required if publicIpv4Pool is specified in install-config.yaml)

  • ec2:DescribeRegions

  • ec2:DescribeRouteTables

  • ec2:DescribeSecurityGroupRules

  • ec2:DescribeSecurityGroups

  • ec2:DescribeSubnets

  • ec2:DescribeTags

  • ec2:DescribeVolumes

  • ec2:DescribeVpcAttribute

  • ec2:DescribeVpcClassicLink

  • ec2:DescribeVpcClassicLinkDnsSupport

  • ec2:DescribeVpcEndpoints

  • ec2:DescribeVpcs

  • ec2:DisassociateAddress (only required if publicIpv4Pool is specified in install-config.yaml)

  • ec2:GetEbsDefaultKmsKeyId

  • ec2:ModifyInstanceAttribute

  • ec2:ModifyNetworkInterfaceAttribute

  • ec2:RevokeSecurityGroupEgress

  • ec2:RevokeSecurityGroupIngress

  • ec2:RunInstances

  • ec2:TerminateInstances

Required permissions for creating network resources during installation

  • ec2:AllocateAddress

  • ec2:AssociateAddress

  • ec2:AssociateDhcpOptions

  • ec2:AssociateRouteTable

  • ec2:AttachInternetGateway

  • ec2:CreateDhcpOptions

  • ec2:CreateInternetGateway

  • ec2:CreateNatGateway

  • ec2:CreateRoute

  • ec2:CreateRouteTable

  • ec2:CreateSubnet

  • ec2:CreateVpc

  • ec2:CreateVpcEndpoint

  • ec2:ModifySubnetAttribute

  • ec2:ModifyVpcAttribute

If you use an existing Virtual Private Cloud (VPC), your account does not require these permissions for creating network resources.
Required Elastic Load Balancing permissions (ELB) for installation

  • elasticloadbalancing:AddTags

  • elasticloadbalancing:ApplySecurityGroupsToLoadBalancer

  • elasticloadbalancing:AttachLoadBalancerToSubnets

  • elasticloadbalancing:ConfigureHealthCheck

  • elasticloadbalancing:CreateListener

  • elasticloadbalancing:CreateLoadBalancer

  • elasticloadbalancing:CreateLoadBalancerListeners

  • elasticloadbalancing:CreateTargetGroup

  • elasticloadbalancing:DeleteLoadBalancer

  • elasticloadbalancing:DeregisterInstancesFromLoadBalancer

  • elasticloadbalancing:DeregisterTargets

  • elasticloadbalancing:DescribeInstanceHealth

  • elasticloadbalancing:DescribeListeners

  • elasticloadbalancing:DescribeLoadBalancerAttributes

  • elasticloadbalancing:DescribeLoadBalancers

  • elasticloadbalancing:DescribeTags

  • elasticloadbalancing:DescribeTargetGroupAttributes

  • elasticloadbalancing:DescribeTargetHealth

  • elasticloadbalancing:ModifyLoadBalancerAttributes

  • elasticloadbalancing:ModifyTargetGroup

  • elasticloadbalancing:ModifyTargetGroupAttributes

  • elasticloadbalancing:RegisterInstancesWithLoadBalancer

  • elasticloadbalancing:RegisterTargets

  • elasticloadbalancing:SetLoadBalancerPoliciesOfListener

  • elasticloadbalancing:SetSecurityGroups

OKD uses both the ELB and ELBv2 API services to provision load balancers. The permission list shows permissions required by both services. A known issue exists in the AWS web console where both services use the same elasticloadbalancing action prefix but do not recognize the same actions. You can ignore the warnings about the service not recognizing certain elasticloadbalancing actions.
Required IAM permissions for installation

  • iam:AddRoleToInstanceProfile

  • iam:CreateInstanceProfile

  • iam:CreateRole

  • iam:DeleteInstanceProfile

  • iam:DeleteRole

  • iam:DeleteRolePolicy

  • iam:GetInstanceProfile

  • iam:GetRole

  • iam:GetRolePolicy

  • iam:GetUser

  • iam:ListInstanceProfilesForRole

  • iam:ListRoles

  • iam:ListUsers

  • iam:PassRole

  • iam:PutRolePolicy

  • iam:RemoveRoleFromInstanceProfile

  • iam:SimulatePrincipalPolicy

  • iam:TagInstanceProfile

  • iam:TagRole

  • If you specify an existing IAM role in the install-config.yaml file, the following IAM permissions are not required: iam:CreateRole,iam:DeleteRole, iam:DeleteRolePolicy, and iam:PutRolePolicy.

  • If you have not created a load balancer in your AWS account, the IAM user also requires the iam:CreateServiceLinkedRole permission.
Required Route 53 permissions for installation

  • route53:ChangeResourceRecordSets

  • route53:ChangeTagsForResource

  • route53:CreateHostedZone

  • route53:DeleteHostedZone

  • route53:GetChange

  • route53:GetHostedZone

  • route53:ListHostedZones

  • route53:ListHostedZonesByName

  • route53:ListResourceRecordSets

  • route53:ListTagsForResource

  • route53:UpdateHostedZoneComment

Required Amazon Simple Storage Service (S3) permissions for installation

  • s3:CreateBucket

  • s3:DeleteBucket

  • s3:GetAccelerateConfiguration

  • s3:GetBucketAcl

  • s3:GetBucketCors

  • s3:GetBucketLocation

  • s3:GetBucketLogging

  • s3:GetBucketObjectLockConfiguration

  • s3:GetBucketPolicy

  • s3:GetBucketRequestPayment

  • s3:GetBucketTagging

  • s3:GetBucketVersioning

  • s3:GetBucketWebsite

  • s3:GetEncryptionConfiguration

  • s3:GetLifecycleConfiguration

  • s3:GetReplicationConfiguration

  • s3:ListBucket

  • s3:PutBucketAcl

  • s3:PutBucketPolicy

  • s3:PutBucketTagging

  • s3:PutEncryptionConfiguration

S3 permissions that cluster Operators require

  • s3:DeleteObject

  • s3:GetObject

  • s3:GetObjectAcl

  • s3:GetObjectTagging

  • s3:GetObjectVersion

  • s3:PutObject

  • s3:PutObjectAcl

  • s3:PutObjectTagging

Required permissions to delete base cluster resources

  • autoscaling:DescribeAutoScalingGroups

  • ec2:DeleteNetworkInterface

  • ec2:DeletePlacementGroup

  • ec2:DeleteVolume

  • elasticloadbalancing:DeleteTargetGroup

  • elasticloadbalancing:DescribeTargetGroups

  • iam:DeleteAccessKey

  • iam:DeleteUser

  • iam:DeleteUserPolicy

  • iam:ListAttachedRolePolicies

  • iam:ListInstanceProfiles

  • iam:ListRolePolicies

  • iam:ListUserPolicies

  • s3:DeleteObject

  • s3:ListBucketVersions

  • tag:GetResources

Required permissions to delete network resources

  • ec2:DeleteDhcpOptions

  • ec2:DeleteInternetGateway

  • ec2:DeleteNatGateway

  • ec2:DeleteRoute

  • ec2:DeleteRouteTable

  • ec2:DeleteSubnet

  • ec2:DeleteVpc

  • ec2:DeleteVpcEndpoints

  • ec2:DetachInternetGateway

  • ec2:DisassociateRouteTable

  • ec2:ReleaseAddress

  • ec2:ReplaceRouteTableAssociation

If you use an existing VPC, your account does not require these permissions to delete network resources. Instead, your account only requires the tag:UntagResources permission to delete network resources.
Optional permissions for installing a cluster with a custom Key Management Service (KMS) key

  • kms:CreateGrant

  • kms:Decrypt

  • kms:DescribeKey

  • kms:Encrypt

  • kms:GenerateDataKey

  • kms:GenerateDataKeyWithoutPlainText

  • kms:ListGrants

  • kms:RevokeGrant

If you provide an Amazon Machine Image (AMI) that is encrypted with a customer-managed key, you must provide the kms:ReEncrypt* permissions in addition to these permissions.
Required permissions to delete a cluster with shared instance roles

  • iam:UntagRole

Required permissions to delete a cluster with shared instance profiles

  • tag:UntagResources

Additional IAM and S3 permissions that are required to create manifests

  • iam:GetUserPolicy

  • iam:ListAccessKeys

  • iam:PutUserPolicy

  • iam:TagUser

  • s3:AbortMultipartUpload

  • s3:GetBucketPublicAccessBlock

  • s3:ListBucket

  • s3:ListBucketMultipartUploads

  • s3:PutBucketPublicAccessBlock

  • s3:PutLifecycleConfiguration

If you are managing your cloud provider credentials with mint mode, the IAM user also requires the iam:CreateAccessKey and iam:CreateUser permissions.
Optional permissions for instance and quota checks for installation

  • servicequotas:ListAWSDefaultServiceQuotas

Optional permissions for the cluster owner account when installing a cluster on a shared VPC

  • sts:AssumeRole

Required permissions for enabling Bring your own public IPv4 addresses (BYOIP) feature for installation

  • ec2:DescribePublicIpv4Pools

  • ec2:DisassociateAddress

Required permissions to install a cluster with dual-stack networking

  • ec2:DescribeEgressOnlyInternetGateways

  • ec2:CreateEgressOnlyInternetGateway

Required permissions to delete a cluster with dual-stack networking

  • ec2:DeleteEgressOnlyInternetGateway

Obtaining an AWS Marketplace image

If you are deploying an OKD cluster using an AWS Marketplace image, you must first subscribe through AWS. Subscribing to the offer provides you with the AMI ID that the installation program uses to deploy compute nodes.

You should only modify the FCOS image for compute machines to use an AWS Marketplace image. Control plane machines and infrastructure nodes do not require an OKD subscription and use the public RHCOS default image by default, which does not incur subscription costs on your AWS bill. Therefore, you should not modify the cluster default boot image or the control plane boot images. Applying the AWS Marketplace image to them will incur additional licensing costs that cannot be recovered.
Prerequisites

  • You have an AWS account to purchase the offer. This account does not have to be the same account that is used to install the cluster.

Procedure

  1. Complete the OKD subscription from the AWS Marketplace.