ocp_provision¤
Provision OCP cluster on IBM Cloud ROKS, ROSA, or DevIT Fyre.
Fyre clusters will be automatically reconfigured to enable NFS storage. By default this is made available via the nfs-client storage class and supports both ReadWriteOnce and ReadWriteMany access modes. The image-registry-storage PVC used by the OpenShift image registry component will also be reconfigured to use this storage class.
Role Variables¤
Most variables are provider-specific: the Conditional entries are only required when provisioning the matching cluster_type (and, for the Fyre sizing variables, the matching fyre_quota_type) — each entry states its exact condition.
Cluster Definition¤
cluster_type¤
Infrastructure provider type for cluster provisioning — one of fyre (IBM DevIT Fyre, internal development), roks (IBM Cloud ROKS classic infrastructure), roks-vpc (IBM Cloud ROKS on VPC infrastructure — modern, recommended; requires an existing VPC), roks-vpc-setup (creates a VPC, subnet, and public gateway — run this first, then use the output IDs with roks-vpc), rosa (Red Hat OpenShift Service on AWS), or ipi (Installer-Provisioned Infrastructure, bare metal/on-premises).
- Required
- Environment Variable:
CLUSTER_TYPE - Default: None
Determines the provisioning workflow and which provider-specific variables are required. Each type has different capabilities — for example GPU support is only available on roks and roks-vpc.
cluster_name¤
Name for the new cluster, used as the cluster identifier in the provider's system — must be unique within the provider's account/region and follow provider naming conventions (typically lowercase alphanumeric with hyphens; some providers impose length limits or character restrictions).
- Required
- Environment Variable:
CLUSTER_NAME - Default: None
Used for DNS, resource naming, and cluster identification.
ocp_version¤
OpenShift version to install — a specific version (e.g. 4.22, 4.22.2), the alias default (newest MAS-supported version), or the alias rotate (predetermined version by day of week, for testing only).
- Required
- Environment Variable:
OCP_VERSION - Default: None
The version must be compatible with MAS and available from the provider.
Warning
For ROKS (cluster_type=roks), the version must append the _openshift suffix — for example 4.22_openshift or 4.22.13_openshift.
ocp_fips_enabled¤
Whether to provision the cluster with FIPS (Federal Information Processing Standards) 140-2 cryptographic mode enabled — set to true for government or regulated environments that require FIPS 140-2 compliance.
- Optional
- Environment Variable:
OCP_FIPS_ENABLED - Default:
false
Applies to fyre and ipi cluster types. FIPS mode may impact performance and cannot be changed after cluster creation — only enable it if required for compliance.
ocp_storage_provider¤
Storage provider configuration for Fyre clusters — set to nfs to enable NFS storage; only applies when cluster_type=fyre, leave unset for other cluster types (ROKS, ROSA, IPI).
- Optional
- Environment Variable:
OCP_STORAGE_PROVIDER - Default: None
When set to nfs, an nfs-client storage class supporting both ReadWriteOnce (RWO) and ReadWriteMany (RWX) access modes is created connected to the Fyre infrastructure node, and the existing image registry PVC is deleted and recreated with NFS storage. See also fyre_nfs_image_registry_size.
GPU Node Support¤
GPU worker pools are supported on roks (classic) and roks-vpc (VPC) cluster types only.
ocp_provision_gpu¤
Whether to provision GPU-enabled worker nodes with the cluster — required for GPU-intensive workloads such as MAS Visual Inspection (MVI).
- Optional
- Environment Variable:
OCP_PROVISION_GPU - Default:
false
When true, a GPU worker pool is provisioned with gpu_workers nodes. On classic ROKS (roks) the GPU flavor is hardcoded to mg4c.32x384.2xp100; on VPC ROKS (roks-vpc) set gpu_flavor explicitly.
gpu_workerpool_name¤
Name for the GPU worker pool — use the default (gpu) for new GPU deployments, or set to an existing pool name to modify that pool rather than create a duplicate.
- Optional
- Environment Variable:
GPU_WORKERPOOL_NAME - Default:
gpu
Only applies when ocp_provision_gpu is true. If a GPU worker pool already exists with this name, it will be modified rather than a second pool created.
gpu_workers¤
Number of GPU-enabled worker nodes to provision in the GPU worker pool — a positive integer (e.g. 1, 2, 3).
- Optional
- Environment Variable:
GPU_WORKERS - Default:
1
Use the default (1) for minimal GPU deployments or testing; increase for production MAS Visual Inspection deployments. GPU nodes use expensive hardware — only provision what you need.
gpu_flavor¤
GPU-enabled machine type for the GPU worker pool when using cluster_type=roks-vpc — for example gx2.8x64.v100 (8 vCPU, 64 GB RAM, 1x V100 GPU, default) or gx2.16x128.2v100 (16 vCPU, 128 GB RAM, 2x V100 GPUs).
- Optional
- Environment Variable:
GPU_FLAVOR - Default:
gx2.8x64.v100
Only applies when ocp_provision_gpu is true and cluster_type=roks-vpc; on classic ROKS (roks) the GPU flavor is hardcoded to mg4c.32x384.2xp100 and this variable is ignored.
Tip
GPU flavor availability varies by zone — verify availability before deploying with ibmcloud oc flavors --zone <zone> --provider vpc-gen2 | grep gpu.
ROKS¤
The following variables are only used when cluster_type is roks, roks-vpc, or roks-vpc-setup (unless noted otherwise).
ibmcloud_apikey¤
IBM Cloud API key for authentication, used by the ibmcloud CLI for all cluster operations — obtain it from IBM Cloud IAM (Identity and Access Management).
- Required (when
cluster_typeisroks,roks-vpc, orroks-vpc-setup) - Environment Variable:
IBMCLOUD_APIKEY - Default: None
The key must have cluster creation permissions in the target resource group (ibmcloud_resourcegroup). Keep API keys secure — use environment variables or secure vaults rather than hardcoding them in playbooks.
ibmcloud_endpoint¤
IBM Cloud API endpoint URL for authentication and cluster operations — override the default only for IBM Cloud dedicated or private environments, or for testing against staging environments.
- Optional
- Environment Variable:
IBMCLOUD_ENDPOINT - Default:
https://cloud.ibm.com
Determines which IBM Cloud environment is targeted for cluster provisioning. Used together with ibmcloud_apikey.
ibmcloud_resourcegroup¤
IBM Cloud resource group in which the ROKS cluster will be created — the group must exist before provisioning and ibmcloud_apikey must have permissions in it.
- Optional
- Environment Variable:
IBMCLOUD_RESOURCEGROUP - Default:
Default
Use the default for simple deployments; specify a different existing resource group for organizational separation.
ibmcloud_region¤
IBM Cloud region targeted when running VPC Infrastructure Service (ibmcloud is) commands — a valid region identifier such as us-south, eu-de, eu-gb, jp-tok, or au-syd.
- Required (when
cluster_type=roks-vpc-setup) - Optional (when
cluster_type=roks-vpc— auto-derived from the VPC CRN or fromroks_zone) - Environment Variable:
IBMCLOUD_REGION - Default: None
For roks-vpc-setup this must be provided explicitly, as it determines where all resources are created. For roks-vpc the region is resolved automatically in this order: roks_zone stripped of its zone suffix (e.g. eu-gb-1 → eu-gb), then the VPC's CRN extracted during the VPC name lookup (which overrides the zone-derived value if different) — an explicitly set ibmcloud_region always wins.
roks_zone¤
IBM Cloud availability zone where the ROKS cluster will be provisioned — classic format for roks (dal10, lon02, fra02, tok02) or VPC format for roks-vpc (us-south-1, eu-de-2, eu-gb-3, jp-tok-1 — always <region>-<n>).
- Optional
- Environment Variable:
ROKS_ZONE - Default:
dal10
The zone determines cluster location, which affects latency, data residency, and available services. For roks-vpc the zone must be within ibmcloud_region (e.g. region us-south → zone us-south-1), and not all roks_flavor values are available in all zones — verify availability in your target zone.
roks_flavor¤
Worker node machine type (flavor) for the ROKS cluster — classic flavors for roks (e.g. b3c.4x16, b3c.16x64.300gb, b3c.32x128) or VPC flavors for roks-vpc (e.g. bx2.4x16, bx2.16x64, bx2.48x192).
- Optional
- Environment Variable:
ROKS_FLAVOR - Default:
b3c.16x64.300gb
The default (16 vCPU, 64 GB RAM, 300 GB storage) suits standard MAS deployments on classic ROKS. Larger flavors cost more but provide more resources.
Warning
The default is a classic flavor and will fail if used with roks-vpc. Always set an appropriate bx2.* VPC flavor when using cluster_type=roks-vpc — check availability with ibmcloud oc flavors --zone <zone> --provider vpc-gen2.
roks_workers¤
Number of worker nodes to provision in the ROKS cluster — a positive integer (minimum 1; 3 or more recommended for production high availability).
- Optional
- Environment Variable:
ROKS_WORKERS - Default:
3
Total cluster capacity = worker count x roks_flavor resources. More workers provide more capacity and better high availability but increase costs; GPU workers can be added separately via ocp_provision_gpu.
roks_flags¤
Additional command-line flags passed to the ibmcloud ks cluster create command — for example --disable-public-service-endpoint, --pod-subnet, or --service-subnet.
- Optional
- Environment Variable:
ROKS_FLAGS - Default:
''
Use for advanced configurations not covered by the other ROKS variables; consult the IBM Cloud Kubernetes Service documentation for available flags.
Warning
Use with caution — incorrect flags can cause provisioning failures.
ROKS VPC Setup¤
The following variables are only used when cluster_type = roks-vpc-setup.
Run this provider first to create the VPC infrastructure, then pass the output IDs to cluster_type=roks-vpc.
Also uses ibmcloud_apikey, ibmcloud_endpoint, ibmcloud_resourcegroup, and ibmcloud_region.
Note: Does not require ocp_version or roks_zone. Subnets are always created in all three zones of ibmcloud_region (<region>-1, <region>-2, <region>-3), matching the default IBM Cloud UI layout.
Resources created (all idempotent — existing resources with the same name are reused):
- VPC:
<roks_vpc_name> - Public gateways:
<roks_vpc_name>-pgw-<region>-1/2/3 - Subnets:
<roks_vpc_name>-subnet-<region>-1/2/3(256 IPs each, gateway attached inline) - COS instance:
<roks_vpc_name>-cos(Standard plan,premium-global-deployment)
Outputs printed at completion for use with cluster_type=roks-vpc:
ROKS_VPC_IDROKS_SUBNET_ID(one per zone)ROKS_COS_INSTANCE_CRN
roks_vpc_name¤
Base name applied to the new VPC and all of its associated resources: VPC (<name>), subnets (<name>-subnet-<region>-<n>), public gateways (<name>-pgw-<region>-<n>), and COS instance (<name>-cos).
- Optional
- Environment Variable:
ROKS_VPC_NAME - Default:
cluster_name
All resources are created idempotently — existing resources with the same name are reused. With cluster_type=roks-vpc, set this to the name used during roks-vpc-setup and the provider automatically resolves roks_vpc_id, roks_subnet_id, and roks_cos_instance_crn by name (see ROKS VPC); providing any of those directly overrides the lookup.
ROKS VPC¤
The following variables are only used when cluster_type = roks-vpc.
Also uses ibmcloud_apikey, ibmcloud_endpoint, ibmcloud_resourcegroup, ibmcloud_region, roks_zone, roks_flavor, roks_workers, and roks_flags.
Note: roks_zone must use VPC zone format (e.g. us-south-1, eu-de-2). roks_vpc_name (see ROKS VPC Setup) is used as the basis for the name-based lookups described below.
roks_vpc_id¤
IBM Cloud VPC ID — overrides the name-based lookup when provided, for example r006-xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx.
- Optional (if
roks_vpc_nameis provided) - Environment Variable:
ROKS_VPC_ID - Default: None
Use when the VPC was not created by roks-vpc-setup or when the name-based lookup is not suitable; otherwise the VPC is looked up by roks_vpc_name.
roks_subnet_id¤
ID of the VPC subnet to place worker nodes into — for example 0717-xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx.
- Optional (resolved automatically from
roks_vpc_id+roks_zoneif not provided) - Environment Variable:
ROKS_SUBNET_ID - Default: None
When not provided, the subnet is resolved automatically by finding the subnet in the VPC whose zone matches roks_zone. This works correctly when the VPC was created by roks-vpc-setup, which creates exactly one subnet per zone; provide directly to override.
roks_cos_instance_crn¤
CRN of the IBM Cloud Object Storage instance — for example crn:v1:bluemix:public:cloud-object-storage:global:a/...:: — used because VPC-based ROKS clusters store the internal OpenShift image registry in COS.
- Optional (resolved automatically from
<roks_vpc_name>-cosif not provided) - Environment Variable:
ROKS_COS_INSTANCE_CRN - Default: None
When not provided, the CRN is looked up from the COS instance named <roks_vpc_name>-cos in the resource group; provide directly to override.
roks_secondary_storage¤
Optional secondary block storage for VPC worker nodes — for example 300gb.5iops-tier — attached to each worker node at provision time, providing additional ephemeral storage capacity beyond the 100GB boot disk.
- Optional
- Environment Variable:
ROKS_SECONDARY_STORAGE - Default:
''
Run ibmcloud ks flavor get --provider vpc-gen2 --zone <zone> --flavor <flavor> to list the available secondary storage configurations for your flavor.
ROSA¤
The following variables are only used when cluster_type = rosa.
rosa_token¤
Red Hat OpenShift Service on AWS (ROSA) authentication token, obtained from OpenShift Cluster Manager — required for all ROSA cluster operations.
- Required (when
cluster_type=rosa) - Environment Variable:
ROSA_TOKEN - Default: None
The token must be valid and have permissions to create clusters; without it ROSA cluster provisioning will fail. Tokens expire periodically — obtain a fresh token before provisioning and keep it secure.
rosa_cluster_admin_password¤
Password for the cluster-admin user account on the ROSA cluster — a string meeting OpenShift password requirements (typically 8+ characters).
- Optional
- Environment Variable:
ROSA_CLUSTER_ADMIN_PASSWORD - Default: None (auto-generated)
When unset, a secure password is auto-generated and saved to the rosa config file in rosa_config_dir. Consider using auto-generation for better security.
rosa_compute_nodes¤
Number of compute (worker) nodes in the ROSA cluster — a positive integer (minimum 2; 3 or more recommended for production high availability).
- Optional
- Environment Variable:
ROSA_COMPUTE_NODES - Default:
3
Total cluster capacity = node count x rosa_compute_machine_type resources. More nodes provide more capacity and better high availability but increase AWS costs.
rosa_compute_machine_type¤
AWS EC2 instance type for ROSA worker nodes — for example m5.2xlarge (development/testing), m5.4xlarge (default, 16 vCPU / 64 GB RAM, suitable for most MAS deployments), m5.8xlarge, or m5.12xlarge (larger workloads).
- Optional
- Environment Variable:
ROSA_COMPUTE_MACHINE_TYPE - Default:
m5.4xlarge
Verify instance type availability in your AWS region.
rosa_config_dir¤
Directory for storing ROSA cluster configuration files — an absolute filesystem path such as /tmp/rosa-config or ~/rosa-clusters.
- Optional
- Environment Variable:
ROSA_CONFIG_DIR - Default:
''
When set, cluster details (API endpoint, credentials) are saved to rosa-{{cluster_name}}-details.yaml in this directory, including the auto-generated rosa_cluster_admin_password if one was not specified; when unset, configuration files are not saved and details must be retrieved manually.
Warning
The config file contains sensitive information (the cluster-admin password). Ensure the directory has appropriate permissions.
Fyre¤
The following variables are only used when cluster_type = fyre. Fyre is an internal IBM development platform — access requires IBM credentials and appropriate permissions.
fyre_username¤
Fyre username for authentication with the IBM DevIT Fyre API — a valid Fyre username (IBM intranet ID).
- Required (when
cluster_type=fyre) - Environment Variable:
FYRE_USERNAME - Default: None
Paired with fyre_password; without valid credentials Fyre cluster provisioning will fail.
fyre_password¤
Fyre API key for authentication with the Fyre API, paired with fyre_username — obtained from the Fyre portal.
- Required (when
cluster_type=fyre) - Environment Variable:
FYRE_APIKEY - Default: None
Without a valid API key Fyre cluster provisioning will fail. Keys may expire and need renewal — keep them secure and rotate regularly.
fyre_quota_type¤
Type of Fyre quota to use for cluster provisioning, which determines the billing method and required sizing variables — quick_burn (pre-defined cluster sizes, faster and simpler; requires fyre_cluster_size) or product_group (custom sizing with more control; requires fyre_worker_count, fyre_worker_cpu, and fyre_worker_memory).
- Required (when
cluster_type=fyre) - Environment Variable:
FYRE_QUOTA_TYPE - Default:
quick_burn
fyre_product_id¤
Product ID for internal IBM accounting and cost tracking — cluster costs are charged to this product ID, so an incorrect ID may cause provisioning to fail or incorrect billing.
- Required (when
cluster_type=fyre) - Environment Variable:
FYRE_PRODUCT_ID - Default: None
Required for both quota types; obtain the correct ID for your project from your IBM product team or manager.
fyre_site¤
Fyre datacenter site to provision the cluster in — a valid Fyre site code such as svl (San Jose/Silicon Valley, default and most commonly used), rtp, or raleigh.
- Optional
- Environment Variable:
FYRE_SITE - Default:
svl
The site determines cluster location, which affects network latency and available resources — not all sites support all features (see ocp_enable_ipv6).
cluster_platform¤
Fyre platform type for the cluster, passed as the platform field in the Fyre API request.
- Optional
- Environment Variable:
CLUSTER_PLATFORM - Default:
x
fyre_cluster_description¤
Human-readable description for the Fyre cluster, shown in the Fyre portal — any descriptive string (e.g. "MAS 9.0 testing", "Development cluster").
- Optional
- Environment Variable:
FYRE_CLUSTER_DESCRIPTION - Default:
No description provided
No functional impact on the cluster; useful for tracking and managing multiple clusters (include purpose, owner, or project information).
fyre_cluster_size¤
Pre-defined Fyre cluster size to use when provisioning with quick_burn quota — for example small (minimal testing), medium (standard development/testing), or large (more intensive workloads).
- Required (when
cluster_type=fyreandfyre_quota_type=quick_burn) - Environment Variable:
FYRE_CLUSTER_SIZE - Default:
medium
Quick burn sizes are pre-defined by Fyre and individual resources cannot be customized — for custom sizing use fyre_quota_type=product_group instead.
fyre_worker_count¤
Number of worker nodes to provision when using product_group quota with custom sizing — a positive integer (typically 2-10; use 2+ for development/testing and 3+ for high availability testing).
- Required (when
cluster_type=fyreandfyre_quota_type=product_group) - Environment Variable:
FYRE_WORKER_COUNT - Default:
2
More workers provide more resources but consume more quota; total cluster capacity = worker count x per-worker resources.
fyre_worker_cpu¤
CPU cores to assign to each worker node when using product_group quota — a positive integer, maximum 16 (Fyre limit).
- Required (when
cluster_type=fyreandfyre_quota_type=product_group) - Environment Variable:
FYRE_WORKER_CPU - Default:
8
Total cluster CPUs = fyre_worker_count x CPU per worker.
fyre_worker_memory¤
Memory in GB to assign to each worker node when using product_group quota — a positive integer, maximum 64 (Fyre limit).
- Required (when
cluster_type=fyreandfyre_quota_type=product_group) - Environment Variable:
FYRE_WORKER_MEMORY - Default:
32
Total cluster memory = fyre_worker_count x memory per worker.
fyre_worker_additional_disks¤
Additional disks to attach to every worker node — a comma-separated list of disk sizes in GB (e.g. 400, 400,400, 200,300,400).
- Optional
- Environment Variable:
FYRE_WORKER_ADDITIONAL_DISKS - Default: None
For example 400,400 adds two 400GB disks to each worker; total additional storage = disk sizes x fyre_worker_count. Useful for testing storage configurations or providing extra capacity.
fyre_worker_base_disk¤
Base disk size in GB for each worker node when using product_group quota — an integer between 40 and 500 (Fyre API limits).
- Optional
- Environment Variable:
FYRE_WORKER_BASE_DISK - Default: None
When unset, the Fyre API default (250 GB per worker) applies. Reducing this can significantly save disk quota on development and test clusters — for example a 4-worker cluster defaults to at least 1000 GB of worker base disks.
fyre_master_base_disk¤
Base disk size in GB for each master node when using product_group quota — an integer between 40 and 500 (Fyre API limits).
- Optional
- Environment Variable:
FYRE_MASTER_BASE_DISK - Default: None
When unset, the Fyre API default (100 GB per master) applies. As with fyre_worker_base_disk, reducing this saves disk quota on development and test clusters.
fyre_nfs_image_registry_size¤
Size of NFS storage allocated for the OpenShift image registry when NFS storage is configured on Fyre clusters — a Kubernetes storage size such as 50Gi, 100Gi, or 200Gi.
- Optional
- Environment Variable:
FYRE_NFS_IMAGE_REGISTRY_SIZE - Default:
100Gi
Only applies when ocp_storage_provider is nfs. The size is limited by available storage on the Fyre infrastructure node; the image registry stores all container images used in the cluster, so increase for clusters with many container images.
ocp_enable_ipv6¤
Whether to enable IPv6 networking for the Fyre cluster — set to true for IPv6 testing.
- Optional
- Environment Variable:
OCP_ENABLE_IPV6 - Default:
false
Warning
IPv6 is only supported at the RTP (Raleigh) Fyre site (fyre_site must be rtp) — it will fail at other sites.
IPI¤
These variables are only used when cluster_type = ipi.
Note
IPI stands for Installer Provisioned Infrastructure. OpenShift offers two possible deployment methods: IPI and UPI (User Provisioned Infrastructure). The difference is the degree of automation and customization. IPI will not only deploy OpenShift but also all infrastructure components and configurations.
ipi_platform¤
Cloud platform for the IPI deployment — aws (default, tested and supported) or gcp; other platforms supported by openshift-install may work but are untested.
- Optional
- Environment Variable:
IPI_PLATFORM - Default:
aws
Determines which cloud provider is used and which provider-specific variables are required: aws requires aws_access_key_id / aws_secret_access_key; gcp requires gcp_service_account_file and ipi_gcp_projectid.
ipi_region¤
Cloud platform region where the IPI cluster will be deployed — a region valid for the selected platform, such as us-east-1, us-west-2, or eu-west-1 for AWS.
- Optional
- Environment Variable:
IPI_REGION - Default:
us-east-1
The region determines cluster location, which affects latency, data residency, and service availability — verify that your target region supports the required instance types.
ipi_base_domain¤
Base DNS domain for the OpenShift cluster — a domain you control, such as example.com or ocp.mycompany.com.
- Required (when
cluster_type=ipi) - Environment Variable:
IPI_BASE_DOMAIN - Default: None
Cluster URLs will be subdomains of this base domain (e.g. api.clustername.example.com). The openshift-install process creates DNS records in this domain, so you must have control over it and the ability to configure DNS.
ipi_pull_secret_file¤
Path to the file containing your Red Hat OpenShift pull secret (required to pull OpenShift container images during installation) — an absolute path to the pull secret JSON file, downloaded from the Red Hat Hybrid Cloud Console.
- Required (when
cluster_type=ipi) - Environment Variable:
IPI_PULL_SECRET_FILE - Default: None
Must be a valid, non-expired pull secret; without one cluster installation fails when attempting to pull container images. Keep the file secure — it contains authentication credentials.
ipi_dir¤
Working directory for the IPI installation process — an absolute filesystem path containing the openshift-install executable, configuration files, and generated logs.
- Optional
- Environment Variable:
IPI_DIR - Default:
~/openshift-install
cluster_name is used in the subdirectory structure beneath this directory. The directory contains sensitive information (kubeconfig, credentials) — ensure appropriate permissions and preserve it for cluster management and troubleshooting.
sshKey¤
Public SSH key to install on all cluster nodes, enabling SSH access via bastion host for troubleshooting — a valid SSH public key string such as ssh-rsa AAAAB3....
- Optional
- Environment Variable:
SSH_PUB_KEY - Default:
''
SSH access to nodes is typically not needed for normal operations of managed clusters; only set this if you need direct node access for debugging or advanced cluster management.
ipi_controlplane_type¤
Machine type for control plane (master) nodes — a machine type valid for the selected platform, such as m5.2xlarge, m5.4xlarge (default, 16 vCPU / 64 GB RAM), or m5.8xlarge for AWS.
- Optional
- Environment Variable:
IPI_CONTROLPLANE_TYPE - Default:
m5.4xlarge
Control plane nodes run cluster management services — an undersized control plane can impact cluster stability, so do not undersize for production. See also ipi_controlplane_replicas.
ipi_controlplane_replicas¤
Number of control plane (master) nodes to provision — an odd positive integer (1, 3, 5, 7) as required for etcd quorum; 3 is recommended for production.
- Optional
- Environment Variable:
IPI_CONTROLPLANE_REPLICAS - Default:
3
Warning
Must be an odd number for etcd quorum. Using 1 provides no high availability — use 3 for production.
ipi_compute_type¤
Machine type for compute (worker) nodes — a machine type valid for the selected platform, such as m5.2xlarge (development/testing), m5.4xlarge (default, 16 vCPU / 64 GB RAM, suitable for most MAS deployments), m5.8xlarge, or m5.12xlarge (larger workloads) for AWS.
- Optional
- Environment Variable:
IPI_COMPUTE_TYPE - Default:
m5.4xlarge
Consider total cluster capacity (type x ipi_compute_replicas) when choosing.
ipi_compute_replicas¤
Number of compute (worker) nodes to provision — a positive integer (minimum 2; 3 or more recommended for production high availability).
- Optional
- Environment Variable:
IPI_COMPUTE_REPLICAS - Default:
3
Total cluster capacity = worker count x ipi_compute_type resources.
ipi_rootvolume_size¤
Root volume size in GiB for cluster nodes — a positive integer such as 120, 200, or 500, covering the OS, container images, and ephemeral storage.
- Optional
- Environment Variable:
IPI_ROOTVOLUME_SIZE - Default: Platform default (typically 120GB)
Platform defaults are typically sufficient — only increase for clusters with many container images or workloads with high ephemeral storage needs. Insufficient space can cause node issues.
AWS¤
The following variables are only used when cluster_type = ipi and ipi_platform = aws.
aws_access_key_id¤
AWS access key ID for authentication — a string of typically 20 characters starting with AKIA, associated with an IAM user or role.
- Required (when
cluster_type=ipiandipi_platform=aws) - Environment Variable:
AWS_ACCESS_KEY_ID - Default: None
The IAM user/role must have extensive permissions (VPC, EC2, ELB, Route53, IAM, etc.) to create the cluster infrastructure — consider using a dedicated IAM user for cluster provisioning.
aws_secret_access_key¤
AWS secret access key for authentication, paired with aws_access_key_id — a string of typically 40 characters.
- Required (when
cluster_type=ipiandipi_platform=aws) - Environment Variable:
AWS_SECRET_ACCESS_KEY - Default: None
Warning
Keep secret keys secure. Never commit them to version control — use environment variables or secure vaults, and rotate keys regularly.
GCP¤
The following variables are only used when cluster_type = ipi and ipi_platform = gcp.
gcp_service_account_file¤
Path to the GCP service account credentials file — an absolute path to a service account JSON key file, downloaded from GCP IAM.
- Required (when
cluster_type=ipiandipi_platform=gcp) - Environment Variable:
GOOGLE_APPLICATION_CREDENTIALS - Default: None
The service account must have extensive permissions (Compute, Networking, IAM, etc.) to create instances and networking resources; keep the file secure.
ipi_gcp_projectid¤
GCP project where the IPI cluster will be deployed — a valid, existing GCP project ID.
- Required (when
cluster_type=ipiandipi_platform=gcp) - Environment Variable:
GOOGLE_PROJECTID - Default: None
All cluster resources are created in, and costs billed to, this project. The project must exist before provisioning and gcp_service_account_file must have permissions in it.
Example Playbook¤
- hosts: localhost
vars:
cluster_type: roks
cluster_name: mycluster
ocp_version: 4.10
ibmcloud_apikey: xxxxx
roles:
- ibm.mas_devops.ocp_provision
License¤
EPL-2.0