POC Permissions

The identity that runs install.sh needs these permissions. On a new cluster, a permission check confirms them before Terraform creates anything, and names any that is missing.

New cluster

Permissions to create VPC, EC2, Elastic IP, IAM role, S3, and EKS resources. AdministratorAccess is the simplest; PowerUserAccess isn’t enough, because it can’t create IAM roles. The permission check confirms these actions:

Service Actions

EC2 / VPC

ec2:CreateVpc, ec2:CreateSubnet, ec2:CreateInternetGateway, ec2:CreateNatGateway, ec2:AllocateAddress, ec2:CreateRouteTable, ec2:CreateLaunchTemplate, ec2:CreateTags

EKS

eks:CreateCluster, eks:CreateNodegroup, eks:CreateAddon, eks:CreatePodIdentityAssociation

IAM

iam:CreateRole, iam:AttachRolePolicy, iam:PutRolePolicy, iam:PassRole

S3

s3:CreateBucket, s3:PutBucketOwnershipControls, s3:PutBucketPublicAccessBlock, s3:PutEncryptionConfiguration

The check itself

iam:SimulatePrincipalPolicy

Use the same AWS identity for Terraform and install.sh. EKS makes only the identity that created the cluster its admin.

If you are sure you have the permissions but the check reports some as missing (it can’t evaluate every organization policy), add skip_permission_check = true to terraform.tfvars and run ./install.sh again. A permission that really is missing then fails part-way through the install.

Existing cluster

Terraform isn’t used, so no cloud permissions are needed to create resources. You need:

  • A kubeconfig context with cluster-admin, because the install creates CRDs, a storage class, and cluster roles.

  • Access to the storage for the Kloudfuse node. On AWS, the node’s IAM role needs the S3 actions, and on Google Cloud the node’s service account needs the bucket roles, listed in Existing cluster requirements.

  • On Azure, Network Contributor for the cluster’s identity on the resource group of the static public IP.