Deploy a Workload
Create a new Pod and deploy a container-based workload.
Prerequisites
- Account with AI Training access
- GPU/CPU quota allocated for Pod deployment
Step 1. Open Deploy New Pod
- Navigate to Infrastructure → Workloads.
- Click Deploy New Pod in the upper right.
Step 2. Basic Information
| Field | Description |
|---|---|
| Pod Name | User-defined Pod name |
| Namespace | Auto-generated from project ID (e.g. abc123 → project-abc123) |
Step 3. GPU Configuration
| Field | Description |
|---|---|
| Compute type | CPU Only or GPU (N/N Available) |
| GPU Count | Slider 1–4 (enabled when GPU is selected) |
Step 4. Container Image
| Option | Description |
|---|---|
| Select Template | Card selection — image and ports auto-detected |
| Custom | Free text input |
| Field | Description |
|---|---|
| Ports | External-facing ports. Auto-detected from templates; empty for custom |
| Environment variables | Key-Value environment variable pairs |
Step 5. Advanced Options
| Field | Default | Description |
|---|---|---|
| CPU Request | 100m | Minimum guaranteed CPU |
| CPU Limit | Auto | Maximum CPU |
| Memory Request | 128Mi | Minimum guaranteed memory |
| Memory Limit | Auto | Maximum memory |
| Storage | 10Gi | Container ephemeral disk |
| Mount Volume | — | Select existing volume or no volume |
| Max Running Time | Default / Custom / No limit | Default: admin setting |
| Notify before termination | — | Alert N minutes before termination (min 5 min) |
| Select Registry Auth | — | Credentials for pulling private images |
| Start Command | — | Overrides the image's default CMD |
Step 6. Complete Deployment
Review the Summary panel on the right, then click Create.