Skip to content

Understanding Assets

Docs

Assets are preconfigured building blocks that simplify workload creation and enable sharing of standardized configurations across teams. This section walks you through creating and using the four main asset types in Run:AI.

What are Assets?

Assets allow you to: - Standardize Configurations: Create reusable templates for environments and resources - Simplify Workload Creation: Reduce complexity for end users - Share Resources: Enable teams to use consistent setups - Maintain Control: Define scope and permissions for asset usage

Asset Types

Run:AI supports four main asset types:

  1. Environments: Container images, tools, and runtime configurations
  2. Compute Resources: GPU, CPU, and memory specifications
  3. Data Sources: Storage connections for datasets and model artifacts
  4. Credentials: Secure access to services and sensitive data

Prerequisites

Before creating assets, ensure you have: - Admin or asset creation permissions in your project - Access to the Run:AI user interface - Understanding of your workload requirements

Using Assets in Workloads

Once you've created assets, you can use them when creating workloads:

1. Create a Workload with Assets

  1. Navigate to Workload manager → Workloads
  2. Click "+ NEW WORKLOAD"
  3. Select your workload type

2. Apply Your Assets

Environment Selection: - Choose your custom environment from the dropdown

Compute Resources: - Select your predefined compute resource

Data Sources: - Mount your PVC and Git data sources

Credentials (if needed): - Reference any credentials for secure access

3. Verify Asset Usage

Once your workload is running:

# Check mounted data sources
ls -la /mnt/datasets
ls -la /mnt/sample-data

# Verify environment variables from credentials
echo $WANDB_API_KEY

Best Practices

Asset Organization

  1. Naming Conventions: Use descriptive, consistent names 

    # Good examples
pytorch-gpu-training-env
large-memory-compute-resource
shared-datasets-pvc

# Avoid
env1
resource-temp
data

  2. Scope Management:

  3. Use Project scope for team-specific assets

  4. Use Cluster scope for organization-wide resources

  5. Documentation: Always add meaningful descriptions to your assets

Security Considerations

  1. Credential Management:
  2. Never store credentials in plain text
  3. Use appropriate scoping to limit access

  4. Regularly rotate sensitive credentials

  5. Data Access:

  6. Ensure data sources have proper access controls
  7. Use read-only mounts when write access isn't needed

Troubleshooting

Common Issues

Asset Not Visible: - Check asset scope matches your project access - Verify you have proper permissions

Mount Failures: 

# Check PVC status
kubectl get pvc -n runai-<project-name>

# Verify storage class availability
kubectl get storageclass

Image Access Issues: 

# Test image accessibility
docker pull <image-name>

# Check image registry permissions

Next Steps

Now that you understand assets:

  1. Create your first assets using the detailed guides for each type
  2. Experiment with different asset combinations in workloads
  3. Share assets with team members for collaboration
  4. Optimize resource allocations based on actual usage