Configure On-Prem Runtime Images and Storage

Use this task to prepare runtime images and storage for an onboarded local cluster.

Applicable Roles

• Platform Operators, on-prem compute administrators, and storage administrators

Before You Start

1. Confirm that the target region and cluster are onboarded and nodes are healthy.
2. Confirm that cluster nodes can reach the image registry and storage services.
3. Prepare minimum-privilege registry and storage credentials and the certificate policy.
4. Define the test workload, image version, capacity, and access requirements.

Procedure

1. Register and Validate the Image Service

Open Image Services, enter the registry endpoint, authentication information, and associated region. After registration, review service and synchronization state. A registry that opens in a browser may still be unreachable from cluster nodes.

2. Synchronize or Upload Runtime Images

Open Image Management and synchronize the registry or upload an image record. Use stable labels that identify framework, version, hardware environment, and purpose. Do not rely only on latest in production.

3. Select Storage for the Workload

• Register Block Storage for independent persistent volumes.
• Register File Storage when multiple workloads share a directory.
• Register Object Storage for datasets, model packages, or output objects.

Enter connection information, capacity, access policy, and associated region, then return to the list and confirm the state.

4. Bind the Region and Run a Combined Validation

Confirm that image and storage components are associated with the target region or cluster. Create a minimal test workload and validate image pull, startup, volume mounting, or object access. After it ends, confirm data-retention and resource-reclaim behavior.

Before deleting or replacing an image service or storage component, check the component's downstream region, cluster, template, and workload dependencies. Remove or migrate those dependencies first; otherwise later image pulls, mounts, or object access may fail.

Completion Checklist

Purpose: These checks confirm that the current foundation supports a real workload. Do not create IDE, training, or inference instances at scale while any check fails.

Check	Pass Criteria
Image registry	Cluster nodes can reach it and authentication and synchronization are healthy.
Runtime image	A test instance can pull the explicit target version.
Storage capability	Target storage can be created, mounted, read/written, or accessed and reclaimed as configured.
Region binding	Images and storage are available in the correct region, cluster, and tenant scope.
Operations ownership	Owners for capacity, backup, deletion, and credential rotation are clear.

Troubleshooting

Symptom	Check First
Instance reports an image-pull failure	Full image path, tag, registry authentication, certificate, and node network
Image exists in the platform but users cannot select it	Synchronization, region binding, tenant permission, and label state
Volume cannot be created or mounted	CSI/storage driver, access mode, capacity, network, and authentication
Object or shared directory cannot be accessed	Endpoint, path or bucket, access policy, and tenant isolation
A component cannot be deleted or replaced safely	Region and cluster bindings, template or workload references, active mounts, and the component's current health state