Skip to content

Latest commit

 

History

History
358 lines (292 loc) · 33.2 KB

README.md

File metadata and controls

358 lines (292 loc) · 33.2 KB
Raw

Red Hat OpenShift VPC cluster on IBM Cloud module

Graduated (Supported) latest release Renovate enabled semantic-release pre-commit Catalog release

Use this module to provision an IBM Cloud Red Hat OpenShift cluster on VPC Gen2. The module either creates the required Cloud Object Storage instance or uses an existing instance. The module also supports optionally passing a key management configuration for secret encryption and boot volume encryption.

Optionally, the module supports advanced security group management for the worker nodes, VPE, and load balancer associated with the cluster. This feature allows you to configure security groups for the cluster's worker nodes, VPE, and load balancer.

❗ Important: You can't update Red Hat OpenShift cluster nodes by using this module. The Terraform logic ignores updates to prevent possible destructive changes.

Before you begin

Overview

terraform-ibm-base-ocp-vpc

Usage

module "ocp_base" {
  source               = "terraform-ibm-modules/base-ocp-vpc/ibm"
  version              = "X.X.X" # Replace "X.X.X" with a release version to lock into a specific release
  cluster_name         = "example-cluster-name"
  resource_group_id    = "xxXXxxXXxXxXXXXxxXxxxXXXXxXXXXX"
  region               = "us-south"
  force_delete_storage = true
  vpc_id               = "79cxxxx-xxxx-xxxx-xxxx-xxxxxXX8667"
  # obtain the below values from the targeted VPC and adjust to the number of zones, subnets, subnet name, cidr_block, id, zone
  vpc_subnets          = {
    zone-1    = [
        {
            cidr_block = "192.168.32.0/22"
            id         = "0717-afc29fbb-0dbe-493a-a5b9-f3c5899cb8b9"
            zone       = "us-south-1"
        },
        {
            cidr_block = "192.168.36.0/22"
            id         = "0727-d65c1eda-9e38-4200-8452-cb8ff5bb3140"
            zone       = "us-south-2"
        },
        {
            cidr_block = "192.168.40.0/22"
            id         = "0737-9a823cd3-16bf-4ba4-a429-9e1fc7db74b8"
            zone       = "us-south-3"
        }
    ]
    zone-2 = [
        {
            cidr_block = "192.168.0.0/22"
            id         = "0717-846b9490-34ae-4a6c-8288-28112dca1ba3"
            zone       = "us-south-1"
        },
        {
            cidr_block = "192.168.4.0/22"
            id         = "0727-ef8db7f6-ffa5-4d8b-a317-4631741a45ee"
            zone       = "us-south-2"
        },
        {
            cidr_block = "192.168.8.0/22"
            id         = "0737-c9a6d871-d95b-4914-abf5-82c22f4161d1"
            zone       = "us-south-3"
        }
    ]
    zone-3 = [
        {
            cidr_block = "192.168.16.0/22"
            id         = "0717-d46e227c-89d4-4b02-9008-d03907a275b6"
            zone       = "us-south-1"
        },
        {
            cidr_block = "192.168.20.0/22"
            id         = "0727-93b1edcb-966c-4517-a7af-6ac63cd93adf"
            zone       = "us-south-2"
        },
        {
            cidr_block = "192.168.24.0/22"
            id         = "0737-807ec4f1-4d84-484e-b2f4-62dd5e431065"
            zone       = "us-south-3"
        }
    ]
  }
  worker_pools         = [
    {
      subnet_prefix    = "default"
      pool_name        = "default"
      machine_type     = "bx2.4x16"
      workers_per_zone = 2
    }
  ]
}

Secure by default cluster settings

In OCP version 4.15, outbound traffic is disabled by default. Learn more.

There is a provision to toggle outbound traffic by using the modules' disable_outbound_traffic_protection input. Refer Managing outbound traffic protection in VPC clusters.

Default Worker Pool management

You can manage the default worker pool using Terraform, and make changes to it through this module. This option is enabled by default. Under the hood, the default worker pool is imported as a ibm_container_vpc_worker_pool resource. Advanced users may opt-out of this option by setting import_default_worker_pool_on_create parameter to false. For most use cases it is recommended to keep this variable to true.

Important Considerations for Terraform and Default Worker Pool

Terraform Destroy

When using the default behavior of handling the default worker pool as a stand-alone ibm_container_vpc_worker_pool, you must manually remove the default worker pool from the Terraform state before running a terraform destroy command on the module. This is due to a known limitation in IBM Cloud.

Terraform CLI Example

For a cluster with 1 or more worker pools, follow these steps:

      $ terraform state list | grep ibm_container_vpc_worker_pool | grep default
        > module.ocp_base.data.ibm_container_vpc_worker_pool.all_pools["default"]
        > module.ocp_base.ibm_container_vpc_worker_pool.pool["default"]

      $ terraform state rm "module.ocp_base.ibm_container_vpc_worker_pool.pool[\"default\"]"

Schematics Example: For a cluster with 1 or more worker pools, follow these steps:

        $ ibmcloud schematics workspace state rm --id <workspace_id> --address "module.ocp_base.ibm_container_vpc_worker_pool.pool[\"default\"]"

Changes Requiring Re-creation of Default Worker Pool

If you need to make changes to the default worker pool that require its re-creation (e.g., changing the worker node operating_system), you must set the allow_default_worker_pool_replacement variable to true, perform the apply, and then set it back to false in the code before the subsequent apply. This is only necessary for changes that require the recreation the entire default pool and is not needed for scenarios that does not require recreating the worker pool such as changing the number of workers in the default worker pool.

This approach is due to a limitation in the Terraform provider that may be lifted in the future.

Advanced security group options

The Terraform module provides options to attach additional security groups to the worker nodes, VPE, and load balancer associated with the cluster.

The custom_sg example demonstrates how to use these capabilities.

See the IBM Cloud documentation on this topic here

Tip: The terraform-ibm-security-groups module can be used to create security groups and rules.

Worker nodes

  • Additional security groups can be specified at cluster creation time. These security groups are attached to all worker nodes of the cluster, including additional worker nodes/pools added after the creation of the cluster. See the variable custom_security_group_ids.
  • Additional security groups can be specified for specific worker pools. These security groups only apply to the worker pool. See the field additional_security_group_ids in the variable worker_pools.

In all cases, note that:

  • The default VPC security is no longer attached to the worker nodes.
  • You can opt-out of attaching the IBM-managed cluster security group (named kube-) through the flag attach_ibm_managed_security_group.
  • It is impossible to change the security groups associated with a cluster after the creation of that cluster.

VPEs (Virtual Private Endpoints)

  • The IBM Cloud OCP stack creates VPEs by default. Prior to version 4.14, a VPE to the master is created. From version 4.14, VPEs to the master, container registry, and IBM Cloud kube APIs are created.
  • You can attach additional security groups through the additional_vpe_security_group_ids variable.
  • The default IBM-managed security group is attached to those VPEs in all cases.

Load balancers

  • The IBM Cloud OCP stack manages the lifecycle of VPC Loadbalancers for your cluster. See the LoadBalancer section in the Understanding options for exposing apps.
  • By default, one load balancer is created at cluster creation for the default cluster ingress.
  • You can attach additional security groups using the additional_lb_security_group_ids variable. This set of security groups is attached to all loadbalancers managed by the cluster.
  • Important: If additional load balancers are added after creating the cluster, for example, by exposing a Kubernetes service of type LoadBalancer, update the number_of_lbs variable and re-run this module to attach the extra security groups to the newly created load balancer.
  • The default IBM-managed security group is attached to the LBs in all cases.

Troubleshooting

New kube_version message

  • When you run a terraform plan command, you might get a message about a new version of Kubernetes, as in the following example:

    kube_version = "4.12.16_openshift" -> "4.12.20_openshift"

Unless you have made equivalent changes to your configuration, or ignored the relevant attributes using ignore_changes, the following plan may include actions to undo or respond to these changes.

    A new version is detected because the Kubernetes master node is updated outside of Terraform, and the Terraform state is out of date with that version.

    The Kubernetes version is ignored in the module code, so the infrastructure will not be modified. The message identifies that drift exists in the versions, and after running the terraform apply command, the state will be refreshed.

Required IAM access policies

You need the following permissions to run this module.

  • Account Management
    • All Identity and Access Enabled service
      • Viewer platform access
    • All Resource Groups service
      • Viewer platform access
  • IAM Services
    • Cloud Object Storage service
      • Editor platform access
      • Manager service access
    • Kubernetes service
      • Administrator platform access
      • Manager service access
    • VPC Infrastructure service
      • Administrator platform access
      • Manager service access

Optionally, you need the following permissions to attach Access Management tags to resources in this module.

  • IAM Services
    • Tagging service
      • Administrator platform access

Note

  • One worker pool should always be named as default. Refer issue 2849 for further details.

Requirements

Name Version
terraform >= 1.3.0
ibm >= 1.68.0, < 2.0.0
kubernetes >= 2.16.1, < 3.0.0
null >= 3.2.1, < 4.0.0

Modules

Name Source Version
attach_sg_to_api_vpe terraform-ibm-modules/security-group/ibm 2.6.2
attach_sg_to_lb terraform-ibm-modules/security-group/ibm 2.6.2
attach_sg_to_master_vpe terraform-ibm-modules/security-group/ibm 2.6.2
attach_sg_to_registry_vpe terraform-ibm-modules/security-group/ibm 2.6.2
cos_instance terraform-ibm-modules/cos/ibm 8.11.10

Resources

Name Type
ibm_container_addons.addons resource
ibm_container_vpc_cluster.autoscaling_cluster resource
ibm_container_vpc_cluster.cluster resource
ibm_container_vpc_worker_pool.autoscaling_pool resource
ibm_container_vpc_worker_pool.pool resource
ibm_resource_tag.cluster_access_tag resource
ibm_resource_tag.cos_access_tag resource
kubernetes_config_map_v1_data.set_autoscaling resource
null_resource.config_map_status resource
null_resource.confirm_lb_active resource
null_resource.confirm_network_healthy resource
null_resource.reset_api_key resource
ibm_container_addons.existing_addons data source
ibm_container_cluster_config.cluster_config data source
ibm_container_cluster_versions.cluster_versions data source
ibm_container_vpc_worker_pool.all_pools data source
ibm_iam_account_settings.iam_account_settings data source
ibm_iam_auth_token.reset_api_key_tokendata data source
ibm_iam_auth_token.tokendata data source
ibm_is_lbs.all_lbs data source
ibm_is_virtual_endpoint_gateway.api_vpe data source
ibm_is_virtual_endpoint_gateway.master_vpe data source
ibm_is_virtual_endpoint_gateway.registry_vpe data source

Inputs

Name Description Type Default Required
access_tags A list of access tags to apply to the resources created by the module, see https://cloud.ibm.com/docs/account?topic=account-access-tags-tutorial for more details list(string) [] no
additional_lb_security_group_ids Additional security groups to add to the load balancers associated with the cluster. Ensure that the number_of_lbs is set to the number of LBs associated with the cluster. This comes in addition to the IBM maintained security group. list(string) [] no
additional_vpe_security_group_ids Additional security groups to add to all existing load balancers. This comes in addition to the IBM maintained security group. 
object({
    master   = optional(list(string), [])
    registry = optional(list(string), [])
    api      = optional(list(string), [])
  })
 {} no
addons Map of OCP cluster add-on versions to install (NOTE: The 'vpc-block-csi-driver' add-on is installed by default for VPC clusters and 'ibm-storage-operator' is installed by default in OCP 4.15 and later, however you can explicitly specify it here if you wish to choose a later version than the default one). For full list of all supported add-ons and versions, see https://cloud.ibm.com/docs/containers?topic=containers-supported-cluster-addon-versions 
object({
    debug-tool                = optional(string)
    image-key-synchronizer    = optional(string)
    openshift-data-foundation = optional(string)
    vpc-file-csi-driver       = optional(string)
    static-route              = optional(string)
    cluster-autoscaler        = optional(string)
    vpc-block-csi-driver      = optional(string)
    ibm-storage-operator      = optional(string)
  })
 {} no
allow_default_worker_pool_replacement (Advanced users) Set to true to allow the module to recreate a default worker pool. Only use in the case where you are getting an error indicating that the default worker pool cannot be replaced on apply. Once the default worker pool is handled as a stand-alone ibm_container_vpc_worker_pool, if you wish to make any change to the default worker pool which requires the re-creation of the default pool set this variable to true. bool false no
attach_ibm_managed_security_group Specify whether to attach the IBM-defined default security group (whose name is kube-) to all worker nodes. Only applicable if custom_security_group_ids is set. bool true no
cluster_config_endpoint_type Specify which type of endpoint to use for for cluster config access: 'default', 'private', 'vpe', 'link'. 'default' value will use the default endpoint of the cluster. string "default" no
cluster_name The name that will be assigned to the provisioned cluster string n/a yes
cluster_ready_when The cluster is ready when one of the following: MasterNodeReady (not recommended), OneWorkerNodeReady, Normal, IngressReady string "IngressReady" no
cos_name Name of the COS instance to provision for OpenShift internal registry storage. New instance only provisioned if 'enable_registry_storage' is true and 'use_existing_cos' is false. Default: '<cluster_name>_cos' string null no
custom_security_group_ids Security groups to add to all worker nodes. This comes in addition to the IBM maintained security group if attach_ibm_managed_security_group is set to true. If this variable is set, the default VPC security group is NOT assigned to the worker nodes. list(string) null no
disable_outbound_traffic_protection Whether to allow public outbound access from the cluster workers. This is only applicable for ocp_version 4.15 bool false no
disable_public_endpoint Whether access to the public service endpoint is disabled when the cluster is created. Does not affect existing clusters. You can't disable a public endpoint on an existing cluster, so you can't convert a public cluster to a private cluster. To change a public endpoint to private, create another cluster with this input set to true. bool false no
enable_registry_storage Set to true to enable IBM Cloud Object Storage for the Red Hat OpenShift internal image registry. Set to false only for new cluster deployments in an account that is allowlisted for this feature. bool true no
existing_cos_id The COS id of an already existing COS instance to use for OpenShift internal registry storage. Only required if 'enable_registry_storage' and 'use_existing_cos' are true string null no
force_delete_storage Flag indicating whether or not to delete attached storage when destroying the cluster - Default: false bool false no
ignore_worker_pool_size_changes Enable if using worker autoscaling. Stops Terraform managing worker count bool false no
import_default_worker_pool_on_create (Advanced users) Whether to handle the default worker pool as a stand-alone ibm_container_vpc_worker_pool resource on cluster creation. Only set to false if you understand the implications of managing the default worker pool as part of the cluster resource. Set to true to import the default worker pool as a separate resource. Set to false to manage the default worker pool as part of the cluster resource. bool true no
kms_config Use to attach a KMS instance to the cluster. If account_id is not provided, defaults to the account in use. 
object({
    crk_id           = string
    instance_id      = string
    private_endpoint = optional(bool, true) # defaults to true
    account_id       = optional(string)     # To attach KMS instance from another account
    wait_for_apply   = optional(bool, true) # defaults to true so terraform will wait until the KMS is applied to the master, ready and deployed
  })
 null no
manage_all_addons Instructs Terraform to manage all cluster addons, even if addons were installed outside of the module. If set to 'true' this module will destroy any addons that were installed by other sources. bool false no
number_of_lbs The number of LBs to associated the additional_lb_security_group_names security group with. number 1 no
ocp_entitlement Value that is applied to the entitlements for OCP cluster provisioning string null no
ocp_version The version of the OpenShift cluster that should be provisioned (format 4.x). If no value is specified, the current default version is used. You can also specify default. This input is used only during initial cluster provisioning and is ignored for updates. To prevent possible destructive changes, update the cluster version outside of Terraform. string null no
operating_system The operating system of the workers in the default worker pool. If no value is specified, the current default version OS will be used. See https://cloud.ibm.com/docs/openshift?topic=openshift-openshift_versions#openshift_versions_available . string null no
region The IBM Cloud region where the cluster will be provisioned. string n/a yes
resource_group_id The Id of an existing IBM Cloud resource group where the cluster will be grouped. string n/a yes
tags Metadata labels describing this cluster deployment, i.e. test list(string) [] no
use_existing_cos Flag indicating whether or not to use an existing COS instance for OpenShift internal registry storage. Only applicable if 'enable_registry_storage' is true bool false no
use_private_endpoint Set this to true to force all api calls to use the IBM Cloud private endpoints. bool false no
verify_worker_network_readiness By setting this to true, a script will run kubectl commands to verify that all worker nodes can communicate successfully with the master. If the runtime does not have access to the kube cluster to run kubectl commands, this should be set to false. bool true no
vpc_id Id of the VPC instance where this cluster will be provisioned string n/a yes
vpc_subnets Metadata that describes the VPC's subnets. Obtain this information from the VPC where this cluster will be created 
map(list(object({
    id         = string
    zone       = string
    cidr_block = string
  })))
 n/a yes
worker_pools List of worker pools 
list(object({
    subnet_prefix = optional(string)
    vpc_subnets = optional(list(object({
      id         = string
      zone       = string
      cidr_block = string
    })))
    pool_name         = string
    machine_type      = string
    workers_per_zone  = number
    resource_group_id = optional(string)
    operating_system  = optional(string)
    labels            = optional(map(string))
    minSize           = optional(number)
    maxSize           = optional(number)
    enableAutoscaling = optional(bool)
    boot_volume_encryption_kms_config = optional(object({
      crk             = string
      kms_instance_id = string
      kms_account_id  = optional(string)
    }))
    additional_security_group_ids = optional(list(string))
  }))
 n/a yes
worker_pools_taints Optional, Map of lists containing node taints by node-pool name map(list(object({ key = string, value = string, effect = string }))) null no

Outputs

Name Description
api_vpe Info about the api VPE, if it exists. For more info about schema, see https://registry.terraform.io/providers/IBM-Cloud/ibm/latest/docs/data-sources/is_virtual_endpoint_gateway
cluster_crn CRN for the created cluster
cluster_id ID of cluster created
cluster_name Name of the created cluster
cos_crn CRN of the COS instance
ingress_hostname The hostname that was assigned to your Ingress subdomain.
kms_config KMS configuration details
master_status The status of the Kubernetes master.
master_url The URL of the Kubernetes master.
master_vpe Info about the master, or default, VPE. For more info about schema, see https://registry.terraform.io/providers/IBM-Cloud/ibm/latest/docs/data-sources/is_virtual_endpoint_gateway
ocp_version Openshift Version of the cluster
operating_system The operating system of the workers in the default worker pool.
private_service_endpoint_url Private service endpoint URL
public_service_endpoint_url Public service endpoint URL
region Region cluster is deployed in
registry_vpe Info about the registry VPE, if it exists. For more info about schema, see https://registry.terraform.io/providers/IBM-Cloud/ibm/latest/docs/data-sources/is_virtual_endpoint_gateway
resource_group_id Resource group ID the cluster is deployed in
vpc_id ID of the clusters VPC
vpe_url The virtual private endpoint URL of the Kubernetes cluster.
workerpools Worker pools created

Contributing

You can report issues and request features for this module in GitHub issues in the module repo. See Report an issue or request a feature.

To set up your local development environment, see Local development setup in the project documentation.