Skip to content

Cloud Paks🔗

Defines the Cloud Pak(s) which is/are layed out on the OpenShift cluster, typically in one or more OpenShift projects. The Cloud Pak definition represents the instance users connect to and which is responsible for managing the functional capabilities installed within the application.

Cloud Pak configuration🔗

cp4d🔗

Defines the Cloud Pak for Data instances to be configured on the OpenShift cluster(s).

cp4d:
- project: cpd
  operators_project: cpd-operators
  openshift_cluster_name: sample
  cp4d_version: latest
  use_fs_iam: False
  change_node_settings: True
  db2u_limited_privileges: False
  ibm_cert_manager: False
  accept_licenses: False
  openshift_storage_name: nfs-storage
  cp4d_entitlement: 
  - cpd-enterprise
  # - cpd-standard
  # - cognos-analytics
  - data-product-hub
  # - datastage
  # - ikc-premium
  # - ikc-standard
  # - openpages
  # - planning-analytics
  # - product-master
  # - speech-to-text
  # - text-to-speech
  - watson-assistant
  # - watson-discovery
  # - watsonx-ai
  # - watsonx-code-assistant-ansible
  # - watsonx-code-assistant-z
  # - watsonx-data
  # - watsonx-gov-mm
  # - watsonx-gov-rc
  # - watsonx-orchestrate  
  cp4d_production_license: True
  state: installed
  
  cartridges:
  - name: cpfs
  - name: cpd_platform

Properties🔗

Property Description Mandatory Allowed values
project Name of the OpenShift project of the Cloud Pak for Data instance Yes
operators_project Name of the OpenShift project that holds the operators. If not specified, the name will be project-operators No Valid OpenShift project name
openshift_cluster_name Name of the OpenShift cluster Yes, inferred from openshift Existing openshift cluster
cp4d_version Cloud Pak for Data version to install, this will determine the version for all cartridges. When specifying latest, the version will be determined from the olm-utils image Yes latest, x.y.z
sequential_install Deprecated property No True (default), False
use_fs_iam If set to True the deployer will enable Foundational Services IAM for authentication No False (default), True
use_cp_alt_repo When set to False, deployer will use use the alternative repo specified in cp_alt_repo resource No True (default), False
change_node_settings Controls whether the node settings using the machine configs will be applied onto the OpenShift cluster. No True, False
db2u_limited_privileges Depicts whether Db2U containers run with limited privileges. If they do (True), Deployer will create KubeletConfig and Tuned OpenShift resources as per the documentation. No False (default), True
ibm_cert_manager Specify if the IBM certificate manager must be installed instead of the Red Hat certificate Manager No False (default), True
accept_licenses Set to 'True' to accept Cloud Pak licenses. Alternatively the --accept-all-licenses can be used for the cp-deploy.sh command No True, False (default)
cp4d_entitlement Set to cpd-enterprise, cpd-standard, watsonx-data, watsonx-ai, watsonx-gov-mm, watsonx-gov-rc, dependent on the deployed license, multiple entitlements can be specified No For valid values, refer to product documentation
cp4d_production_license Whether the Cloud Pak for Data is a production license No True (default), False
state Indicated whether Cloud Pak for Data must be installed or removed No installed (default), removed
image_registry_name When using private registry, specify name of image_registry No
openshift_storage_name References an openshift_storage element in the OpenShift cluster that was defined for this Cloud Pak for Data instance. The name must exist under `openshift.[openshift_cluster_name].openshift_storage. No, inferred from openshift->openshift_storage
cartridges List of cartridges to install for this Cloud Pak for Data instance. See Cloud Pak for Data cartridges for more details Yes

cp4i🔗

Defines the Cloud Pak for Integration installation to be configured on the OpenShift cluster(s).

cp4i:
- project: cp4i
  openshift_cluster_name: {{ env_id }}
  openshift_storage_name: nfs-rook-ceph
  cp4i_version: 2021.4.1
  accept_licenses: False
  use_top_level_operator: False
  top_level_operator_channel: v1.5
  top_level_operator_case_version: 2.5.0
  operators_in_all_namespaces: True
 
  instances:
  - name: integration-navigator
    type: platform-navigator
    license: L-RJON-C7QG3S
    channel: v5.2
    case_version: 1.5.0

OpenShift projects🔗

The immediate content of the cp4i object is actually a list of OpenShift projects (namespaces). There can be more than one project and instances can be created in separate projects.

cp4i:
- project: cp4i
  ...

- project: cp4i-ace
  ...

- project: cp4i-apic
  ...

Operator channels, CASE versions, license IDs🔗

Before you run the Cloud Pak Deployer be sure that the correct operator channels are defined for the selected instance types. Some products require a license ID, please check the documentation of each product for the correct license. If you decide to use CASE files instead of the IBM Operator Catalog (more on that below) make sure that you selected the correct CASE versions - please refer: https://github.com/IBM/cloud-pak/tree/master/repo/case

CP4I main properties🔗

The following properties are defined on the project level:

Property Description Mandatory Allowed values
project The name of the OpenShift project that will be created and used for the installation of the defined instances. Yes
openshift_cluster_name Dynamically defined form the env_id parameter during the execution. Yes, inferred from openshift Existing openshift cluster
openshift_storage_name Reference to the storage definition that exists in the openshift object (please see above). The definition must include the class name of the file storage type and the class name of the block storage type. No, inferred from openshift->openshift_storage
cp4i_version The version of the Cloud Pak for Integration (e.g. 2021.4.1) Yes
use_case_files The property defines if the CASE files are used for installation. If it is True then the operator catalogs are created from the CASE files. If it is False, the IBM Operator Catalog from the entitled registry is used. No True, False (default)
accept_licenses Set to True to accept Cloud Pak licenses. Alternatively the --accept-all-licenses can be used for the cp-deploy.sh command Yes True, False
use_top_level_operator If it is True then the CP4I top-level operator that installs all other operators is used. Otherwise, only the operators for the selected instance types are installed. No True, False (default)
top_level_operator_channel Needed if the use_top_level_operator is True otherwise, it is ignored. Specifies the channel of the top-level operator. No
top_level_operator_case_version Needed if the use_top_level_operator is True otherwise, it is ignored. Specifies the CASE package version of the top-level operator. No
operators_in_all_namespaces It defines whether the operators are visible in all namespaces or just in the specific namespace where they are needed. No True, False (default)
instances List of the instances that are going to be created (please see below). Yes

Warning

Despite the properties use_case_files, use_top_level_operator and operators_in_all_namespaces are defined as optional, they are actually crucial for the way of execution of the installation process. If any of them is omitted, it is assumed that the default False value is used. If none of them exists, it means that all are False. In this case, it means that the IBM Operator Catalog is used and only the needed operators for specified instance types are installed in the specific namespace.

Properties of the individual instances🔗

The instance property contains one or more instances definitions. Each instance must have a unique name. There can be more the one instance of the same type.

Naming convention for instance types🔗

For each instance definition, an instance type must be specified. We selected the type names that are as much as possible similar to the naming convention used in the Platform Navigator use interface. The following table shows all existing types:

Instance type Description/Product name
platform-navigator Platform Navigator
api-management IBM API Connect
automation-assets Automation assets a.k.a Asset repo
enterprise-gateway IBM Data Power
event-endpoint-management Event endpoint manager - managing asynchronous APIs
event-streams IBM Event Streams - Kafka
high-speed-transfer-server Aspera HSTS
integration-dashboard IBM App Connect Integration Dashboard
integration-design IBM App Connect Designer
integration-tracing Operations Dashboard
messaging IBM MQ

Platform navigator🔗

The Platform Navigator is defined as one of the instance types. There is typically only one instance of it. The exception would be an installation in two or more completely separate namespaces (see the CP4I documentation). Special attention is paid to the installation of the Navigator. The Cloud Pak Deployer will install the Navigator instance first, before any other instance, and it will wait until the instance is ready (this could take up to 45 minutes).

When the installation is completed, you will find the admin user password in the status/cloud-paks/cp4i--cp4i-PN-access.txt file. Of course, you can obtain the password also from the platform-auth-idp-credentials secret in ibm-common-services namespace.

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be platform-navigator
license License ID L-RJON-C7QG3S
channel Subscription channel v5.2
case_version CASE version 1.5.0

API management (IBM API Connect)🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be api-management
license License ID L-RJON-C7BJ42
version Version of API Connect 10.0.4.0
channel Subscription channel v2.4
case_version CASE version 3.0.5

Automation assets (Asset repo)🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be automation-assets
license License ID L-PNAA-C68928
version Version of Asset repo 2021.4.1-2
channel Subscription channel v1.4
case_version CASE version 1.4.2

Enterprise gateway (IBM Data Power)🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be enterprise-gateway
admin_password_secret The name of the secret where admin password is stored. The default name is used if you leave it empty.
license License ID L-RJON-BYDR3Q
version Version of Data Power 10.0-cd
channel Subscription channel v1.5
case_version CASE version 1.5.0

Event endpoint management🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be event-endpoint-management
license License ID L-RJON-C7BJ42
version Version of Event endpoint manager 10.0.4.0
channel Subscription channel v2.4
case_version CASE version 3.0.5

Event streams🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be event-streams
version Version of Event streams 10.5.0
channel Subscription channel v2.5
case_version CASE version 1.5.2

High speed transfer server (Aspera HSTS)🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be high-speed-transfer-server
aspera_key A license key for the Aspera software
redis_version Version of the Redis database 5.0.9
version Version of Aspera HSTS 4.0.0
channel Subscription channel v1.4
case_version CASE version 1.4.0

Integration dashboard (IBM App Connect Dashboard)🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be integration-dashboard
license License ID L-APEH-C79J9U
version Version of IBM App Connect 12.0
channel Subscription channel v3.1
case_version CASE version 3.1.0

Integration design (IBM App Connect Designer)🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be integration-design
license License ID L-KSBM-C87FU2
version Version of IBM App Connect 12.0
channel Subscription channel v3.1
case_version CASE version 3.1.0

Integration tracing (Operation dashborad)🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be integration-tracing
version Version of Integration tracing 2021.4.1-2
channel Subscription channel v2.5
case_version CASE version 2.5.2

Messaging (IBM MQ)🔗

Property Description Sample value for 2021.4.1
name Unique name within the cluster using only lowercase alphanumerics and "-"
type It must be messaging
queue_manager_name The name of the initial queue. Default is QUICKSTART
license License ID L-RJON-C7QG3S
version Version of IBM MQ 9.2.4.0-r1
channel Subscription channel v1.7
case_version CASE version 1.7.0

cp4waiops🔗

Defines the Cloud Pak for Watson AIOps installation to be configured on the OpenShift cluster(s). The following instances can be installed by the deployer:

  • AI Manager
  • Event Manager
  • Turbonomic
  • Instana
  • Infrastructure management
  • ELK stack (ElasticSearch, Logstash, Kibana)

Aside from the base install, the deployer can also install ready-to-use demos for each of the instances

cp4waiops:
- project: cp4waiops
  openshift_cluster_name: "{{ env_id }}"
  openshift_storage_name: auto-storage
  accept_licenses: False
 
  instances:
  - name: cp4waiops-aimanager
    kind: AIManager
    install: true
  ...

AIOPS main properties🔗

The following properties are defined on the project level:

Property Description Mandatory Allowed values
project The name of the OpenShift project that will be created and used for the installation of the defined instances. Yes
openshift_cluster_name Dynamically defined form the env_id parameter during the execution. No, only if mutiple OpenShift clusters defined Existing openshift cluster
openshift_storage_name Reference to the storage definition that exists in the openshift object (please see above). No, inferred from openshift->openshift_storage
accept_licenses Set to True to accept Cloud Pak licenses. Alternatively the --accept-all-licenses can be used for the cp-deploy.sh command Yes True, False

Service instances🔗

The project that is specified at the cp4waiops level defines the OpenShift project into which the instances of each of the services will be installed. Below is a list of instance "kinds" that can be installed. For every "service instance" there can also be a "demo content" entry to prepare the demo content for the capability.

AI Manager🔗

  instances:
  - name: cp4waiops-aimanager
    kind: AIManager
    install: true

    waiops_size: small
    custom_size_file: none
    waiops_name: ibm-cp-watson-aiops
    subscription_channel: v3.6
    freeze_catalog: false
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes AIManager
install Must the service be installed? Yes true, false
waiops_size Size of the install Yes small, tall, custom
custom_size_file Name of the file holding the custom sizes if waiops_size is custom No
waiops_name Name of the CP4WAIOPS instance Yes
subscription_channel Subscription channel of the operator Yes
freeze_catalog Freeze the version of the catalog source? Yes false, true
case_install Must AI manager be installed via case files? No false, true
case_github_url GitHub URL to download case file Yes if case_install is true
case_name Name of the case file Yes if case_install is true
case_version Version of the case file to download Yes if case_install is true
case_inventory_setup Case file operation to run for this service Yes if case_install is true cpwaiopsSetup

AI Manager - Demo Content🔗

  instances:
  - name: cp4waiops-aimanager-demo-content
    kind: AIManagerDemoContent
    install: true
    ...
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes AIManagerDemoContent
install Must the content be installed? Yes true, false

See sample config for remainder of properties.

Event Manager🔗

  instances:
  - name: cp4waiops-eventmanager
    kind: EventManager
    install: true
    subscription_channel: v1.11
    starting_csv: noi.v1.7.0
    noi_version: 1.6.6
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes EventManager
install Must the service be installed? Yes true, false
subscription_channel Subscription channel of the operator Yes
starting_csv Starting Cluster Server Version Yes
noi_version Version of noi Yes

Event Manager Demo Content🔗

  instances:
  - name: cp4waiops-eventmanager
    kind: EventManagerDemoContent
    install: true
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes EventManagerDemoContent
install Must the content be installed? Yes true, false

Infrastructure Management🔗

  instances:
  - name: cp4waiops-infrastructure-management
    kind: InfrastructureManagement
    install: false
    subscription_channel: v3.5
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes InfrastructureManagement
install Must the service be installed? Yes true, false
subscription_channel Subscription channel of the operator Yes

ELK stack🔗

ElasticSearch, Logstash and Kibana stack.

  instances:
  - name: cp4waiops-elk
    kind: ELK
    install: false
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes ELK
install Must the service be installed? Yes true, false

Instana🔗

  instances:
  - name: cp4waiops-instana
    kind: Instana
    install: true
    version: 241-0

    sales_key: 'NONE'
    agent_key: 'NONE'

    instana_admin_user: "admin@instana.local"
    #instana_admin_pass: 'CHANGEME'
    
    install_agent: true

    integrate_aimanager: true
    #integrate_turbonomic: true
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes Instana
install Must the service be installed? Yes true, false
version Version of Instana to install No
sales_key License key to be configured No
agent_key License key for agent to be configured No
instana_admin_user Instana admin user to be configured Yes
instana_admin_pass Instana admin user password to be set (if different from global password) No
install_agent Must the Instana agent be installed? Yes true, false
integrate_aimanager Must Instana be integrated with AI Manager? Yes true, false
integrate_turbonomic Must Instana be integrated with Turbonomic? No true, false

Turbonomic🔗

  instances:
  - name: cp4waiops-turbonomic
    kind: Turbonomic
    install: true
    turbo_version: 8.7.0
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes Turbonomic
install Must the service be installed? Yes true, false
turbo_version Version of Turbonomic to install Yes

Turbonomic Demo Content🔗

  instances:
  - name: cp4waiops-turbonomic-demo-content
    kind: TurbonomicDemoContent
    install: true
    #turbo_admin_password: CHANGEME
    create_user: false
    demo_user: demo
    #turbo_demo_password: CHANGEME
Property Description Mandatory Allowed values
name Unique name within the cluster using only lowercase alphanumerics and "-" Yes
kind Service kind to install Yes TurbonomicDemoContent
install Must the content be installed? Yes true, false
turbo_admin_pass Turbonomic admin user password to be set (if different from global password) No
create_user Must the demo user be created? No false, true
demo_user Name of the demo user No
turbo_demo_password Demo user password if different from global password No

See sample config for remainder of properties.

cp4ba🔗

Defines the Cloud Pak for Business Automation installation to be configured on the OpenShift cluster(s).
See Cloud Pak for Business Automation for additional details.

---
cp4ba:
- project: cp4ba
  collateral_project: cp4ba-collateral
  openshift_cluster_name: "{{ env_id }}"
  openshift_storage_name: auto-storage
  accept_licenses: false
  state: installed
  cpfs_profile_size: small # Profile size which affect replicas and resources of Pods of CPFS as per https://www.ibm.com/docs/en/cpfs?topic=operator-hardware-requirements-recommendations-foundational-services

  # Section for Cloud Pak for Business Automation itself
  cp4ba:
    # Set to false if you don't want to install (or remove) CP4BA
    enabled: true # Currently always true
    profile_size: small # Profile size which affect replicas and resources of Pods as per https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=pcmppd-system-requirements
    patterns:
      foundation: # Foundation pattern, always true - https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__foundation
        optional_components:
          bas: true # Business Automation Studio (BAS) 
          bai: true # Business Automation Insights (BAI)
          ae: true # Application Engine (AE)
      decisions: # Operational Decision Manager (ODM) - https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__odm
        enabled: true
        optional_components:
          decision_center: true # Decision Center (ODM)
          decision_runner: true # Decision Runner (ODM)
          decision_server_runtime: true # Decision Server (ODM)
        # Additional customization for Operational Decision Management
        # Contents of the following will be merged into ODM part of CP4BA CR yaml file. Arrays are overwritten.
        cr_custom:
          spec:
            odm_configuration:
              decisionCenter:
                # Enable support for decision models
                disabledDecisionModel: false
      decisions_ads: # Automation Decision Services (ADS) - https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__ads
        enabled: true
        optional_components:
          ads_designer: true # Designer (ADS)
          ads_runtime: true # Runtime (ADS)
        gen_ai: # https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=services-configuring-generative-ai-secret
          endpoint: https://us-south.ml.cloud.ibm.com
          project_id: project_id          
          api_key: watsonx_ai_api_key
      content: # FileNet Content Manager (FNCM) - https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__ecm
        enabled: true
        optional_components:
          cmis: true # Content Management Interoperability Services (FNCM - CMIS)
          css: true # Content Search Services (FNCM - CSS)
          es: true # External Share (FNCM - ES)
          tm: true # Task Manager (FNCM - TM)
          ier: true # IBM Enterprise Records (FNCM - IER)
          icc4sap: false # IBM Content Collector for SAP (FNCM - ICC4SAP) - Currently not implemented
          ccxmo: true # Content Cortex for Microsoft Office (CCXMO)
          gen_ai: true # AI features of base CCX
          ccxai: true # Content Cortex Ai Services
        gen_ai:
          endpoint: https://us-south.ml.cloud.ibm.com/ml/v1/text/extractions
          space_id: space_id
          api_key: watsonx_ai_api_key
          cos_endpoint: cos_endpoint
          cos_api_key: cos_api_key
          cos_connection_id: cos_connection_id          
        ccxai_gen_ai:
          provider: watsonx
          endpoint: https://us-south.ml.cloud.ibm.com
          api_key: watsonx_ai_api_key
          model: openai/gpt-oss-120b
          watsonx_saas_config:
            project_id: project_id
      application: # Business Automation Application (BAA) - https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__baa
        enabled: true
        optional_components:
          app_designer: true # App Designer (BAA)
          ae_data_persistence: true # App Engine data persistence (BAA)
      document_processing: # Automation Document Processing (ADP) - https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__adp
        enabled: true
        optional_components: 
          document_processing_designer: true # Designer (ADP)
        # Additional customization for Automation Document Processing
        # Contents of the following will be merged into ADP part of CP4BA CR yaml file. Arrays are overwritten.
        cr_custom:
          spec:
            ca_configuration:
              ## NB: All config parameters for ADP are described here ==> https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=parameters-automation-document-processing
              ocrextraction:
                # [Tech Preview] OCR Engine 2 (IOCR) for ADP - Starts the Watson Document Understanding (WDU) pods to process documents.
                use_iocr: auto # Allowed values: auto, all, none. Refer to doc for option details: https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=parameters-automation-document-processing#:~:text=ocrextraction.use_iocr
                deep_learning_object_detection: # When enabled, ca_configuration.deeplearning parameters will be used (ignored otherwise), and deep-learning pods will be deployed to enhance object detection.
                  # If disabled, all training will automatically be done in "fast-training" mode and should finish in less than 10 min.
                  # Warn: If you enable this option and don't select the "fast training" mode in ADP before starting training, training could take hours (or more if you don't have GPUs).
                  #       See "Important" note here for usage recommandation on using "fast/deeplarning" training: https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/23.0.2?topic=project-creating-data-extraction-model#:~:text=Training%20takes%20time
                  enabled: true
              deeplearning: # Only used if deep_learning_object_detection is enabled. Configure usage of GPU-enabled Nodes.
                gpu_enabled: false # Use GPUs for deeplearning training instead of CPUs.
                nodelabel_key: nvidia.com/gpu.present
                nodelabel_value: "true"
                replica_count: 1 # Controls the number of deep learning pod replicas. NB: The number of GPUs available on your cluster should be ≥ to replica_count.
      workflow: # Business Automation Workflow (BAW) - https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__baw
        enabled: true
        optional_components:
          baw_authoring: true # Workflow Authoring (BAW) - always keep true if workflow pattern is chosen. BAW Runtime is not implemented.
          kafka: true # Will install a kafka cluster and enable kafka service for workflow authoring.
          workflow_assistant: true # Will enable Authoring assistant for workflow authoring.
          workplace_assistant: true # Will enable Workplace assistant.
        gen_ai: # https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/25.0.0?topic=customizing-enabling-generative-ai
          endpoint: https://us-south.ml.cloud.ibm.com
          project_id: project_id
          api_key: watsonx_ai_api_key
          model: openai/gpt-oss-120b
        wxo_service_instance_url: https://api.hostname/instances/tenant_id
  
  # Section for IBM Process mining
  pm:
    # Set to false if you don't want to install (or remove) Process Mining
    enabled: true
    # Additional customization for Process Mining
    # Contents of the following will be merged into PM CR yaml file. Arrays are overwritten.
    cr_custom:
      spec:
        processmining:
          storage:
            # Disables redis to spare resources as per https://www.ibm.com/docs/en/process-mining/latest?topic=configurations-custom-resource-definition
            redis:
              install: false  

  # Section for IBM Robotic Process Automation
  rpa:
    # Set to false if you don't want to install (or remove) RPA
    enabled: true
    # Additional customization for Robotic Process Automation
    # Contents of the following will be merged into RPA CR yaml file. Arrays are overwritten.
    cr_custom:
      spec:
        # Configures the NLP provider component of IBM RPA. You can disable it by specifying 0. https://www.ibm.com/docs/en/rpa/latest?topic=platform-configuring-rpa-custom-resources#basic-setup
        sizeMapping:
          watson-nlp:
            replicas: 1

  # Section for IBM Business Automation Manager Open Editions
  bamoe:
    # Set to false if you don't want to install (or remove) BAMOE
    enabled: true

  # Section for IBM Content Assistant
  ica:
    # Set to false if you don't want to install (or remove) ICA
    enabled: true
    gen_ai:
      api_key: watsonx_ai_api_key
      space_id: space_id
      model: openai/gpt-oss-120b
      embedding_model: intfloat/multilingual-e5-large

  # Set to false if you don't want to install (or remove) CloudBeaver (PostgreSQL, DB2, MSSQL UI)
  cloudbeaver_enabled: true

  # Set to false if you don't want to install (or remove) Roundcube
  roundcube_enabled: true

  # Set to false if you don't want to install (or remove) AKHQ
  akhq_enabled: true

  # Set to false if you don't want to install (or remove) phpLDAPAdmin
  phpldapadmin_enabled: true

  # Set to false if you don't want to install (or remove) OpenSearch Dashboards
  opensearch_dashboards_enabled: true  

CP4BA main properties🔗

The following properties are defined on the project level.

Property Description Mandatory Allowed values
project The name of the OpenShift project that will be created and used for the installation of the defined instances. Yes Valid OCP project name
collateral_project The name of the OpenShift project that will be created and used for the installation of all collateral (prerequisites and extras). Yes Valid OCP project name
openshift_cluster_name Dynamically defined form the env_id parameter during the execution. No, only if multiple OpenShift clusters defined Existing openshift cluster
openshift_storage_name Reference to the storage definition that exists in the openshift object (please see above). No, inferred from openshift->openshift_storage
accept_licenses Set to true to accept Cloud Pak licenses. Alternatively the --accept-all-licenses can be used for the cp-deploy.sh command Yes true, false
state Set to installed to install enabled capabilities, set to removed to remove enabled capabilities. Yes installed, removed
cpfs_profile_size Profile size which affect replicas and resources of Pods of CPFS as per https://www.ibm.com/docs/en/cpfs?topic=operator-hardware-requirements-recommendations-foundational-services Yes starterset, small, medium, large

Cloud Pak for Business Automation properties🔗

Used to configure CP4BA.
Placed in cp4ba key on the project level.

Property Description Mandatory Allowed values
enabled Set to true to enable CP4BA. Currently always true. Yes true
profile_size Profile size which affect replicas and resources of Pods as per https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=pcmppd-system-requirements Yes small, medium, large
patterns Section where CP4BA patterns are configured. Please make sure to select all that is needed as a dependencies. Dependencies can be determined from documentation at https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments Yes Object - see details below

Foundation pattern properties🔗

Always configure in CP4BA.
https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__foundation
Placed in cp4ba.patterns.foundation key.

Property Description Mandatory Allowed values
optional_components Sub object for definition of optional components for pattern. Yes Object - specific to each pattern
optional_components.bas Set to true to enable Business Automation Studio Yes true, false
optional_components.bai Set to true to enable Business Automation Insights Yes true, false
optional_components.ae Set to true to enable Application Engine Yes true, false

Decisions pattern properties🔗

Used to configure Operation Decision Manager.
https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__odm
Placed in cp4ba.patterns.decisions key.

Property Description Mandatory Allowed values
enabled Set to true to enable decisions pattern. Yes true, false
optional_components Sub object for definition of optional components for pattern. Yes Object - specific to each pattern
optional_components.decision_center Set to true to enable Decision Center Yes true, false
optional_components.decision_runner Set to true to enable Decision Runner Yes true, false
optional_components.decision_server_runtime Set to true to enable Decision Server Yes true, false
cr_custom Additional customization for Operational Decision Management. Contents will be merged into ODM part of CP4BA CR yaml file. Arrays are overwritten. No Object

Decisions ADS pattern properties🔗

Used to configure Automation Decision Services.
https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__ads
Placed in cp4ba.patterns.decisions_ads key.

Property Description Mandatory Allowed values
enabled Set to true to enable decisions_ads pattern. Yes true, false
optional_components Sub object for definition of optional components for pattern. Yes Object - specific to each pattern
optional_components.ads_designer Set to true to enable Designer Yes true, false
optional_components.ads_runtime Set to true to enable Runtime Yes true, false
gen_ai Sub object for definition of GenAI connection. More on https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/24.0.0?topic=services-configuring-generative-ai-secret false Object
gen_ai.api_key Set to real value of your Watsonx.AI platform false Your real value, (default) watsonx_ai_api_key
gen_ai.endpoint Set to real value of your Watsonx.AI platform false Your real value, (default) https://us-south.ml.cloud.ibm.com
gen_ai.project_id Set to real value of your Watsonx.AI platform false Your real value, (default) project_id

Content pattern properties🔗

Used to configure FileNet Content Manager.
https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__ecm
Placed in cp4ba.patterns.content key.

Property Description Mandatory Allowed values
enabled Set to true to enable content pattern. Yes true, false
optional_components Sub object for definition of optional components for pattern. Yes Object - specific to each pattern
optional_components.cmis Set to true to enable CMIS Yes true, false
optional_components.css Set to true to enable Content Search Services Yes true, false
optional_components.tm Set to true to enable Task Manager Yes true, false
optional_components.ier Set to true to enable IBM Enterprise Records Yes true, false
optional_components.icc4sap Set to true to enable IBM Content Collector for SAP. Currently not functional. Always false. Yes false
optional_components.ccxmo Set to true to enable Content Cortex for Microsoft Office Yes true, false
optional_components.gen_ai Set to true to enable Gen AI Yes true, false
optional_components.ccxai Set to true to enable Content Cortex AI Services Yes true, false
gen_ai Sub object for definition of LLM provider. For Enhanced Permanent Text Extration. Mandatory if gen_ai is set to true including sub items false Object
gen_ai.endpoint Set to real value of your Watsonx.AI platform false Your real value, (default) https://us-south.ml.cloud.ibm.com/ml/v1/text/extractions
gen_ai.space_id Set to real value of your Watsonx.AI platform false Your real value, (default) space_id
gen_ai.api_key Set to real value of your Watsonx.AI platform false Your real value, (default) api_key
gen_ai.cos_endpoint Set to real value of your Watsonx.AI platform, format https://bucket-name.s3.region.cloud-object-storage.appdomain.cloud false Your real value, (default) cos_endpoint
gen_ai.cos_api_key Set to real value of your Watsonx.AI platform false Your real value, (default) cos_api_key
gen_ai.cos_connection_id Set to real value of your Watsonx.AI platform false Your real value, (default) cos_connection_id
ccxai_gen_ai Sub object for definition of LLM provider for CCX AI. Mandatory if ccxai is set to true including sub items false Object
ccxai_gen_ai.provider Set to one fo the options false watsonx, azure
ccxai_gen_ai.endpoint Set to real value of your Watsonx.AI platform or Azure false Your real value, (default) https://us-south.ml.cloud.ibm.com
ccxai_gen_ai.api_key Set to real value of your Watsonx.AI platform or Azure false Your real value
ccxai_gen_ai.model Set to desired ID of foundation model of you rpatform false Your real value, (default) openai/gpt-oss-120b
ccxai_gen_ai.watsonx_saas_config Sub object for details for watsonx.ai LLM provider. Only rpovide if provider is watsonx false Your real value
ccxai_gen_ai.watsonx_saas_config.project_id Set to real value of your Watsonx.AI platform false Your real value

Application pattern properties🔗

Used to configure Business Automation Application.
https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__baa
Placed in cp4ba.patterns.application key.

Property Description Mandatory Allowed values
enabled Set to true to enable application pattern. Yes true, false
optional_components Sub object for definition of optional components for pattern. Yes Object - specific to each pattern
optional_components.app_designer Set to true to enable Application Designer Yes true, false
optional_components.ae_data_persistence Set to true to enable App Engine data persistence Yes true, false

Document Processing pattern properties🔗

Used to configure Automation Document Processing.
https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__baa
Placed in cp4ba.patterns.document_processing key.

Property Description Mandatory Allowed values
enabled Set to true to enable document_processing pattern. Yes true, false
optional_components Sub object for definition of optional components for pattern. Yes Object - specific to each pattern
optional_components.document_processing_designer Set to true to enable Designer Yes true
cr_custom Additional customization for Automation Document Processing. Contents will be merged into ADP part of CP4BA CR yaml file. Arrays are overwritten. No Object

Workflow pattern properties🔗

Used to configure Business Automation Workflow.
https://www.ibm.com/docs/en/cloud-paks/cp-biz-automation/latest?topic=deployment-capabilities-production-deployments#concept_c2l_1ks_fnb__baw
Placed in cp4ba.patterns.workflow key.

Property Description Mandatory Allowed values
enabled Set to true to enable workflow pattern. Yes true, false
optional_components Sub object for definition of optional components for pattern. Yes Object - specific to each pattern
optional_components.baw_authoring Set to true to enable Workflow Authoring. Currently always true. Yes true
optional_components.kafka Set to true to enable kafka service for workflow authoring. Yes true, false
optional_components.workflow_assistant Set to true to enable Authoring assistant for workflow authoring. Yes true, false
optional_components.workplace_assistant Set to true to enable Workplace assistant. Yes true, false
gen_ai Sub object for definition of GenAI connection. For Gen AI Service Flow, Authoring Assistant and Workplace Assistant false Object
gen_ai.api_key Set to real value of your Watsonx.AI platform false Your real value, (default) watsonx_ai_api_key
gen_ai.endpoint Set to real value of your Watsonx.AI platform false Your real value, (default) https://us-south.ml.cloud.ibm.com
gen_ai.project_id Set to real value of your Watsonx.AI platform false Your real value, (default) project_id
gen_ai.model Set to desired ID of foundation model (Only for Gen AI Service flow) false Your real value, (default) openai/gpt-oss-120b
wxo_service_instance_url Set to desired instace URL of you WXO instance false Your real value, (default) https://api.hostname/instances/tenant_id

Process Mining properties🔗

Used to configure IBM Process Mining.
Placed in pm key on the project level.

Property Description Mandatory Allowed values
enabled Set to true to enable process mining. Yes true, false
cr_custom Additional customization for Process Mining. Contents will be merged into PM CR yaml file. Arrays are overwritten. No Object

Robotic Process Automation properties🔗

Used to configure IBM Robotic Process Automation.
Placed in rpa key on the project level.

Property Description Mandatory Allowed values
enabled Set to true to enable rpa. Yes true, false
cr_custom Additional customization for Process Mining. Contents will be merged into RPA CR yaml file. Arrays are overwritten. No Object

Business Automation Manager Open Editions properties🔗

Used to configure Business Automation Manager Open Editions.
Placed in bamoe key on the project level.

Property Description Mandatory Allowed values
enabled Set to true to enable bamoe. Yes true, false

IBM Content Assistant properties🔗

Used to configure IBM Content Assistant.
Placed in ica key on the project level.

Property Description Mandatory Allowed values
enabled Set to true to enable ica. Yes true, false
gen_ai Sub object for definition of GenAI connection. More on https://www.ibm.com/docs/en/cacms/1.0.x?topic=software-configuring-service-id-your-watsonxai-account false Object
gen_ai.api_key Set to real value of your Watsonx.AI platform false Your real value
gen_ai.space_id Set to real value of your Watsonx.AI platform false Your real value
gen_ai.model Set to desired ID of foundation model false Your real value, (default) openai/gpt-oss-120b
gen_ai.embedding_model Set to desired ID of embedding model false Your real value, (default) intfloat/multilingual-e5-large

Other properties🔗

Used to configure extra UIs.
The following properties are defined on the project level.

Property Description Mandatory Allowed values
cloudbeaver_enabled Set to true to enable CloudBeaver (PostgreSQL, DB2, MSSQL UI). Yes true, false
roundcube_enabled Set to true to enable Roundcube. Client for mail. Yes true, false
akhq_enabled Set to true to enable AKHQ. Client for Kafka in CP4BA. Yes true, false
phpldapadmin_enabled Set to true to enable phpLDApAdmin. Client for OpenLDAP. Yes true, false
opensearch_dashboards_enabled Set to true to enable OpenSearch Dashboards. Client for OpenSearch. Yes true, false

ibm_bob🔗

Defines an IBM Bob self-hosted installation on an existing OpenShift cluster. The deployer installs IBM Bob with bobctl, which it takes from the versioned release bundle on GitHub. It downloads the bundle at run time, verifies its SHA-256 checksum and caches it in <status_dir>/downloads/ibm-bob/<version>/.

ibm_bob:
- project: ibm-bob
  operators_project: ibm-bob-operators
  openshift_cluster_name: "{{ env_id }}"
  openshift_storage_name: odf-storage
  ibm_bob_version: 2.0.0
  accept_licenses: false
  reuse_existing_cert_manager: false
  state: installed
  premium_packages:
    rag: false
    z_proxy: false
    understand: false

Prerequisites🔗

The cluster requirements of IBM Bob (supported OpenShift versions, node architecture, storage, cert-manager and permissions) depend on the IBM Bob version. See the IBM Bob prerequisites. The deployer checks them before it installs.

What the deployer does with them:

  • The file (RWX) and block (RWO) storage classes are resolved from the openshift_storage element that openshift_storage_name refers to.
  • If no cert-manager operator is found, the deployer installs the Red Hat cert-manager operator, as it does for Cloud Pak for Data. The IBM License Service is not installed.
  • The IBM entitlement key is read from the vault secret ibm_cp_entitlement_key. IBM Bob images are pulled from cp.icr.io with this key.

IBM Bob main properties🔗

Property Description Mandatory Allowed values
project The OpenShift project (bobctl: global.instanceNamespace) in which the IBM Bob instance is created. Yes
operators_project The OpenShift project (bobctl: global.operatorNamespace) for the IBM Bob operators. No, defaults to <project>-operators
openshift_cluster_name Name of the openshift element to install on. Yes, inferred from openshift Existing openshift cluster
openshift_storage_name References an openshift_storage element of the cluster. Provides the file and block storage classes. No, inferred from openshift -> openshift_storage
ibm_bob_version Version of IBM Bob, x.y.z, or latest for the highest published version (connected) or the highest downloaded bundle (air-gapped). If an older version is installed, the deployer upgrades it, see Upgrade. Yes 2.0.0, 2.0.1, latest
accept_licenses Set to True to accept the IBM Bob license. Alternatively use --accept-all-licenses with cp-deploy.sh. Yes (not required for state: removed) True, False
reuse_existing_cert_manager Set to True to reuse a working cert-manager that was installed outside OLM (for example with Helm or manifests) instead of installing the Red Hat cert-manager. The deployer checks that a cert-manager controller is running. No True, False (default)
state Desired state of the IBM Bob instance. Setting removed requires --confirm-destroy or global_config.confirm_destroy: true. No installed (default), removed
premium_packages Premium packages to enable, see below. No
model_gateway Model gateway configuration, see below. No
model_gateway_secret_names Maps env.<VAR> references in model_gateway to vault secret names. Only with env.<VAR> references
image_registry_name Private registry (image_registry object) to pull from. Required for air-gapped installations. Yes if CPD_AIRGAP is true
remove_model_weights When state: removed and openshift_ai_models is configured: the model weight PVCs and the models namespace are deleted too. Set to False to keep the downloaded weights (PVCs and namespace are kept). No True (default), False
ingress_cert TLS certificate for Bob's routes, see Ingress certificate below. No

Ingress certificate🔗

By default, Bob generates its own self-signed CA and issues a certificate for its routes. The deployer saves the CA certificate to <status_dir>/ibm-bob/<project>/bob-ca.crt; distribute it to every workstation that connects to IBM Bob.

To provision a publicly trusted certificate automatically, add an ingress_cert block with exactly one of:

Key Description
acme_email Email address for ACME HTTP-01 (Let's Encrypt by default). The deployer creates a namespace-scoped cert-manager Issuer and a Certificate covering Bob's three routes, waits for Ready, and calls bobctl setup-route. The cluster's apps domain must be reachable from the internet for the HTTP-01 challenge.
cluster_issuer Name of an existing cert-manager ClusterIssuer. The deployer creates the Certificate only; the ClusterIssuer must already be Ready.

acme_email and cluster_issuer are mutually exclusive. The optional acme_url key overrides the ACME directory URL (defaults to Let's Encrypt production; set to a staging or corporate ACME CA as needed).

# ACME HTTP-01 — Let's Encrypt (default) or corporate ACME CA
ibm_bob:
- project: ibm-bob
  ...
  ingress_cert:
    acme_email: ops@example.com
    # acme_url: https://acme-v02.api.letsencrypt.org/directory   # default

# Existing ClusterIssuer
ibm_bob:
- project: ibm-bob
  ...
  ingress_cert:
    cluster_issuer: my-corporate-issuer

Bob's route hostnames (covered by the certificate):

  • bob.<project>.<apps-domain>
  • api.bob.<project>.<apps-domain>
  • bob-keycloak.<project>.<apps-domain>

On state: removed, the deployer deletes the Certificate and, if ACME mode was used, the Issuer created by the deployer. The secret bob-cluster-ingress-tls is deleted together with the Bob namespace.

Premium packages🔗

Property Description Default
rag Enables bob.rag.enabled (retrieval-augmented generation, installs OpenSearch). False
z_proxy Enables bob.zProxy.enabled. Requires rag: True. False
understand Enables bob.understand.enabled. False

Model gateway🔗

If model_gateway is specified, the deployer writes it to model-gateway.yaml in the status directory and passes it to bobctl install --model-config. If it is omitted, IBM Bob is installed with an empty model gateway configuration, which you can set later with bobctl update-model-config. The structure of model_gateway follows the IBM Bob model gateway configuration.

Do not put secrets in the configuration. Refer to them as env.<VAR> and map each <VAR> to a vault secret:

  model_gateway:
    models:
    - model_name: premium-ide
      openai_compatible:
        api_key: env.AZURE_API_KEY
  model_gateway_secret_names:
    AZURE_API_KEY: ibm-bob-azure-api-key
./cp-deploy.sh vault set -vs=ibm-bob-azure-api-key=<value>

The deployer reads the values from the vault when it installs, and removes the temporary files that contain them afterwards.

Air-gapped installation🔗

For air-gapped clusters, set CPD_AIRGAP=true and provide an image_registry object with the private registry that holds the mirrored IBM Bob images. Set image_registry_name in the ibm_bob object to reference it.

Mirror the images before running the deployer:

# Step 1: download images from cp.icr.io to a local filesystem directory
CPD_AIRGAP=false CPD_SKIP_PORTABLE_REGISTRY=false \
  ./cp-deploy.sh env download -c <config-dir> -l <status-dir>

# Step 2: upload from the filesystem directory to the private registry
CPD_AIRGAP=true CPD_SKIP_PORTABLE_REGISTRY=false \
  ./cp-deploy.sh env download -c <config-dir> -l <status-dir>

Or mirror directly from the IBM registry to the private registry in one step:

CPD_AIRGAP=false CPD_SKIP_PORTABLE_REGISTRY=true \
  ./cp-deploy.sh env download -c <config-dir> -l <status-dir>

Store the private registry credentials in the vault before running the apply:

./cp-deploy.sh vault set -vs=image-registry-<name>=<user>:<password>

Then run the deployer with CPD_AIRGAP=true:

CPD_AIRGAP=true ./cp-deploy.sh env apply -c <config-dir> -l <status-dir>

Upgrade🔗

To upgrade IBM Bob, change ibm_bob_version to the new version (for example from 2.0.0 to 2.0.1, or to latest) and run the deployer again. Before it changes anything, the deployer compares the requested version with spec.version of the Bob custom resource bob-cr in the IBM Bob project (the label app.kubernetes.io/version is used when spec.version is not set):

Installed version Action
None (no Bob custom resource) Install with bobctl install.
Bob custom resource exists, version not recognized The deployer stops with an error, to avoid breaking the existing installation with bobctl install.
Same as requested Re-apply with bobctl install, as before.
Lower than requested Upgrade, see below.
Higher than requested The deployer stops with an error. Downgrading IBM Bob is not supported.

An upgrade does the following:

  1. Generates and applies the cluster-scoped resources of the new version, as for an install.
  2. Deletes the ClusterRole ibm-bob-role-<operators_project> and ClusterRoleBinding ibm-bob-rolebinding-<operators_project> that older versions created, if the new version no longer creates them.
  3. Runs bobctl upgrade. The values of the installed release are kept, so the model gateway, certificate and premium package settings of the installed instance are retained.
  4. Runs bobctl update-model-config when model_gateway or openshift_ai_models is configured, because bobctl upgrade does not take a model configuration.
  5. Waits until the Bob custom resource reports the new version, is Available for its current generation, and is not Degraded.

The instance is upgraded in place: the Keycloak admin credentials, the model weights and the other data of the instance are kept.

With --dry-run the deployer lists these steps in deployer-activities.log without running them, see Dry run.

Dry run🔗

Run cp-deploy.sh env apply --dry-run (or env destroy --dry-run) to review what the deployer would do with IBM Bob without changing the cluster or the vault. The deployer still reads the cluster (for example to detect an upgrade) and renders the configuration, but every command that would change something is written to <status_dir>/log/deployer-activities.log instead: the bobctl, oc and helm commands, the YAML of the objects it would create (namespaces, PVCs, ServingRuntimes, InferenceServices, model download jobs) and comments for the steps that cannot be shown as a command, such as waits and pull secrets. Pull secrets and other secrets are never written to the log, and the IBM entitlement key is not required. The log is recreated at the start of every run.

For state: removed the log lists the removal commands in the order in which they would run. Use a user with the cluster-reader role to run a dry run.

Model weights are not downloaded or uploaded in a dry run, so the log only contains the Job that would download them.

Uninstall🔗

To remove an IBM Bob instance, set state: removed. No storage or license properties are required:

ibm_bob:
- project: ibm-bob
  openshift_cluster_name: "{{ env_id }}"
  ibm_bob_version: 2.0.0
  state: removed

The deployer deletes the Bob CR, removes the Helm release, deletes both namespaces (project and operators_project) and removes the Keycloak admin vault key. cert-manager is left untouched.

Access information🔗

When the installation has completed, the smoke tests wait for the operators and the Bob custom resource to be ready. They then write the following to <status_dir>/cloud-paks/cloud-pak-deployer-info.txt, which is also displayed at the end of the run:

  • The gateway URL for IBM Bob clients (BOB_GATEWAY_URL for bob-shell, gatewayUrl for bob-ide).
  • The Keycloak URL, admin user and password (also stored in the vault as ibm_bob_keycloak_admin_<project>_<cluster>).
  • The URLs for premium packages (RAG URL, Internal search URL, Z Proxy URL, Z Lookup URL), if enabled and available in the Bob custom resource status.
  • When the default self-signed certificate is in use: the CA certificate is saved as <status_dir>/ibm-bob/<project>/bob-ca.crt. Distribute it to every workstation that connects to IBM Bob, because it is not publicly trusted.
  • When ingress_cert is configured: confirmation that the cert-manager certificate was issued and registered, including the hostnames it covers.

Adding users🔗

After installation, add users through the IBM Bob Admin UI at https://bob.<project>.<apps-domain>/admin. Log in with the temp-admin Keycloak credentials from the access information above. Navigate to Users & Teams, add each user by email address, assign their role (bob-admin or bob-user), and add them to a team.

Alternatively, add users directly in the Keycloak Admin Console at https://bob-keycloak.<project>.<apps-domain>: create a user in the bob realm, set a password, and assign the bob-admin or bob-user realm role.

Post-installation tasks🔗

Tasks such as LDAP and IDP configuration and customer-provided TLS certificates are performed with bobctl. See Post-installation tasks.

OpenShift AI model serving🔗

Instead of writing a model_gateway block by hand, the deployer can serve one of the open-weight models listed under Supported models on Red Hat OpenShift AI (KServe + vLLM) and generate the Bob model gateway configuration automatically. Configure this with openshift_ai_models in the ibm_bob object. openshift_ai_models and model_gateway are mutually exclusive.

What IBM Bob supports and requires for these models is documented only in the IBM Bob documentation: Required and supported models and Recommended vLLM runtime configuration. Check these pages before you choose a model.

ibm_bob:
- project: ibm-bob
  ...
  openshift_ai_models:
    core_model: nemotron-3-ultra       # see Supported models
    quantization: fp8                  # default depends on the model, see Supported models
    guardrail_model: openai/gpt-oss-20b  # default; configurable
    project: ibm-bob-models            # namespace for ServingRuntime / InferenceService / PVCs
    gpus_per_core_model: 8             # default depends on the model, see Supported models
    gpus_per_guardrail_model: 1
    storage_size:
      core: 700Gi                      # default depends on the model and quantization
      guardrail: 60Gi
    # runtime_image:                   # vLLM image per inference service
    #   core: vllm/vllm-openai:v0.32.0
    #   guardrail: vllm/vllm-openai:v0.32.0
    # huggingface_token_secret_name: ibm-bob-huggingface-token
    # vllm_extra_args: []
    # skip_gpu_check: false

The openshift object must have openshift_ai.install: auto (or True) and gpu.install: auto (or True). The deployer installs OpenShift AI and the NFD + NVIDIA GPU operators automatically when both are set to auto.

openshift_ai_models properties🔗

Property Description Mandatory Allowed values
core_model Core model to serve. Exactly one may be configured. Yes core_model value in Supported models
quantization Weight quantization variant. Not all variants are available for all models. No Quantizations in Supported models; default per model
guardrail_model Model ID for the guardrail (safety) model served on the security and background slots. No default openai/gpt-oss-20b
project OpenShift namespace for the model serving objects (PVCs, ServingRuntimes, InferenceServices). No default <ibm_bob.project>-models
gpus_per_core_model Number of GPUs requested by the core model InferenceService. The vLLM --tensor-parallel-size is set to the same number. No default per model, see Supported models
gpus_per_guardrail_model Number of GPUs requested by the guardrail model InferenceService. No default 1
storage_size.core PVC size for the core model weights (and draft model, if any). No default per model and quantization, see Supported models
storage_size.guardrail PVC size for the guardrail model weights. No default 60Gi
runtime_image.core vLLM container image for the core model ServingRuntime. By default the deployer uses vllm/vllm-openai:v0.32.0 and automatically mirrors it to your private registry for air-gapped installations. Use this property to override the image with a custom build or different version. No default vllm/vllm-openai:v0.32.0
runtime_image.guardrail vLLM container image for the guardrail model ServingRuntime. By default the deployer uses vllm/vllm-openai:v0.32.0 and automatically mirrors it to your private registry for air-gapped installations. No default vllm/vllm-openai:v0.32.0
huggingface_token_secret_name Vault secret name for the Hugging Face access token. Required for gated model repositories. Used during env download in both connected and air-gapped mode. No vault secret name
vllm_extra_args Extra vLLM CLI arguments appended to the arguments the deployer sets for the chosen model, so they take precedence. No list of strings
skip_gpu_check Skip the GPU capacity check. The model pods will be Pending until GPU nodes are added. Use for deployer flow testing on clusters without GPUs. No True, False (default)

Supported models🔗

The Cloud Pak Deployer currently supports the following core models. For what IBM Bob requires of these models, see the IBM Bob documentation.

core_model Quantization Hugging Face repository Default GPUs Default storage_size.core
nemotron-3-ultra fp8 (default) RedHatAI/NVIDIA-Nemotron-3-Ultra-550B-A55B-FP8-dynamic 8 700Gi
nemotron-3-ultra nvfp4 RedHatAI/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 8 400Gi
nemotron-3.5-super nvfp4 (default, only) nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4 2 100Gi
laguna-s-2.1 fp8 (default) poolside/Laguna-S-2.1-FP8 8 150Gi
laguna-s-2.1 nvfp4 poolside/Laguna-S-2.1-NVFP4 8 120Gi
mistral-medium-3.5 fp8 (default, only) mistralai/Mistral-Medium-3.5-128B 8 300Gi

The guardrail model openai/gpt-oss-20b serves the security and background gateway slots (1 GPU, 60Gi).

The vLLM arguments are the values recommended on the IBM Bob vLLM page. nemotron-3.5-super is not on that page; its arguments are from the Hugging Face model card (2x H100 or 1x H200 recipe, so gpus_per_core_model: 1 is also valid on an H200). Adjust the arguments for your hardware with vllm_extra_args. The following notes are specific to the deployer:

  • The Laguna S 2.1 and Mistral Medium 3.5 speculative decoding draft models (dFlash and Eagle) are downloaded together with the model weights into a sub-directory of the model PVC. The Nemotron models use built-in speculative decoding.
  • The nvfp4 quantization of Laguna S 2.1 runs on NVIDIA Blackwell GPUs only. If vLLM crashes with DFlash at the default sequence limit, add --max-num-seqs=32 to vllm_extra_args.
  • All models are served with the upstream vllm/vllm-openai:v0.32.0 image (or the image configured via runtime_image). The vLLM images that Red Hat ships with OpenShift AI are not used because they contain older vLLM versions than these models need. In an air-gapped environment, the deployer automatically mirrors the runtime image, and the ubi9/ubi image of the temporary pods that load the model weights, to your private registry alongside the IBM Bob images during env download.

What the deployer does🔗

env download (both modes):

  1. Downloads model weights from Hugging Face into status_dir/downloads/ibm-bob/<version>/models/ using the hf CLI (huggingface_hub). A local marker file makes re-runs a no-op. The status directory is packaged by env download and carried to the air-gapped site.
  2. Mirrors the vLLM runtime container image (vllm/vllm-openai:v0.32.0 or user-configured runtime_image) and the registry.access.redhat.com/ubi9/ubi:latest image, which the temporary pods use to copy the model weights into the PVCs, to local disk (downloads/ibm-bob/<version>/images/vllm/) or directly to the private registry when image_registry_name is configured.

env apply:

  1. Installs OpenShift AI (KServe) and the NFD + NVIDIA GPU operators when openshift_ai.install and gpu.install are auto.
  2. Creates a namespace and two PVCs (one per model) using the file storage class.
  3. In connected mode: runs a Kubernetes Job to download weights from Hugging Face directly into the PVCs.
  4. In air-gapped mode: copies weights from the local status directory into the PVCs file by file with oc cp (files already on the PVC with the same size are skipped, so an interrupted copy resumes; progress is written to <status_dir>/log/ibm-bob/models-copy-<core|guardrail>.log), then writes a PVC-side marker. If the local weights are missing (step 1 was not run), fails with instructions to run env download first.
  5. Creates a vLLM ServingRuntime and InferenceService per model with the recommended arguments (plus any vllm_extra_args).
  6. Waits for each InferenceService to have a resolvable URL, and for Ready status (unless skip_gpu_check is true).
  7. Renders the Bob model gateway config (slots premium-ide, security, background and a static router) and passes it to bobctl install --model-config.

Vault token for Hugging Face🔗

Set the token before running the deployer:

./cp-deploy.sh vault set -vs=ibm-bob-huggingface-token=<your-hf-token>

Then reference it in the config:

  openshift_ai_models:
    core_model: nemotron-3-ultra
    huggingface_token_secret_name: ibm-bob-huggingface-token

Air-gapped installation🔗

The deployer follows the same two-phase pattern used for container images.

Connected side — run env download:

# Standard two-phase (status dir is packaged and carried to the air-gapped site)
CPD_AIRGAP=false ./cp-deploy.sh env download -c <config-dir> -l <status-dir>

# Direct air-gap (skip portable packaging, download and apply on same network path)
CPD_AIRGAP=false CPD_SKIP_PORTABLE_REGISTRY=true \
  ./cp-deploy.sh env download -c <config-dir> -l <status-dir>

This downloads model weights (and Bob container images if image_registry_name is set) into <status-dir>/downloads/ibm-bob/<version>/models/. For gated models, set huggingface_token_secret_name first:

./cp-deploy.sh vault set -vs=ibm-bob-huggingface-token=<your-hf-token>

Air-gapped side — carry the status dir and run env apply:

For the standard two-phase flow, the status directory (which includes the downloaded weights) is packaged by env download and transferred to the air-gapped site. Then:

CPD_AIRGAP=true ./cp-deploy.sh env apply -c <config-dir> -l <status-dir>

The deployer copies the weights from the status directory into the cluster PVCs automatically. If the weights are not present (step 1 was skipped or the status dir was not carried over), the deployer fails with instructions.

Testing without GPUs (skip_gpu_check)🔗

To verify the deployer flow on a cluster without GPU nodes:

  openshift_ai_models:
    core_model: nemotron-3-ultra
    skip_gpu_check: true

The deployer runs the full flow: OpenShift AI check, PVC creation, weight download (connected) or status-dir upload (air-gapped), ServingRuntime and InferenceService creation, gateway generation, and Bob installation. The model pods remain Pending as expected; this does not affect Bob itself. In a dry run, nothing of this is created: the objects are written to deployer-activities.log.

Removal🔗

When state: removed and openshift_ai_models is set, the deployer deletes the InferenceService and ServingRuntime objects. PVCs are kept — the downloaded weights are expensive to re-download. Delete them manually when no longer needed:

oc delete pvc ibm-bob-core-model ibm-bob-guardrail-model -n ibm-bob-models