> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gmicloud.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Managed GPU Clusters

> Provision and operate managed Kubernetes GPU clusters for large-scale AI workloads.

URL: `https://console.gmicloud.ai/user-console/ce/managed-clusters/clusters`

GMI Cloud's Managed GPU Cluster Service (MGCS) provides fully managed Kubernetes-based GPU clusters for large-scale AI workloads. Unlike individual containers, MGCS offers dedicated GPU worker nodes with Kubernetes orchestration, giving you cluster-level control over your compute resources.

Use [Cluster Requests](/cluster-engine/resources/cluster-requests) to launch new ones.

<Frame>
  <img src="https://mintcdn.com/gmicloud/wwBqG4D8qH5TFQuz/images/compute-managed-clusters.png?fit=max&auto=format&n=wwBqG4D8qH5TFQuz&q=85&s=53ac0b3c2f8abc8c35ab31a0b606d4ec" alt="Managed GPU Clusters" width="1749" height="954" data-path="images/compute-managed-clusters.png" />
</Frame>

## Toolbar

* Search by cluster name or full ID.
* **Data Center** dropdown filter.
* Status pills: **All**, **Running**, **Deleting**.
* **Request Cluster** (top-right), same flow as the Compute Home catalog.

## Empty state

> **No Clusters Found.** No clusters match your current filters. Try adjusting your search.

When populated, each cluster row exposes **name**, **ID**, **data center**, **node count**, **status**, and per-cluster actions (open, delete).

<Note>
  The request for GPU Cluster Service is not processed automatically. After submitting a request, please contact support to activate it.
</Note>

## View Managed GPU Clusters

1. Click "**Managed GPU Clusters**" in the left sidebar under the "Cluster" section

<img src="https://mintcdn.com/gmicloud/i7S_g2j1X9iiXhO1/assets/mgcs-sidebar-navigation.png?fit=max&auto=format&n=i7S_g2j1X9iiXhO1&q=85&s=d396a4b687c1dcd71c6a8301d778630e" alt="mgcs-sidebar-navigation.png" width="3024" height="1442" data-path="assets/mgcs-sidebar-navigation.png" />

2. The cluster list displays the following information for each cluster:

| **Column**             | **Description**                                  |
| ---------------------- | ------------------------------------------------ |
| **Name / ID**          | Cluster name and unique identifier               |
| **Kubernetes Version** | The version of Kubernetes running on the cluster |
| **Instance Type**      | GPU worker node specification                    |
| **Quantity**           | Number of worker nodes in the cluster            |
| **Billing Method**     | Pay as you go or Prepaid                         |
| **Status**             | Current cluster status                           |
| **Actions**            | Available management actions                     |

3. Use the filters to narrow down your clusters:
   * **Search**: Filter by cluster name or IP address
   * **Data Center**: Filter by data center location
   * **Cluster Status**: Filter by cluster status

## Request a New Cluster

To request a new managed GPU cluster, follow the 3-step configuration process:

1. Click the "**Request Cluster**" button on the Managed GPU Clusters page or the Requests page

<img src="https://mintcdn.com/gmicloud/i7S_g2j1X9iiXhO1/assets/mgcs-request-cluster-button.png?fit=max&auto=format&n=i7S_g2j1X9iiXhO1&q=85&s=9d40f307161ac9e2c96703344fb5aa33" alt="mgcs-request-cluster-button.png" width="3024" height="1442" data-path="assets/mgcs-request-cluster-button.png" />

### Step 1: Choose Your GPU Worker Node

Configure the compute resources for your cluster:

<img src="https://mintcdn.com/gmicloud/i7S_g2j1X9iiXhO1/assets/mgcs-request-cluster-step1.png?fit=max&auto=format&n=i7S_g2j1X9iiXhO1&q=85&s=7243f35ef709c1ec2d610c43a6163ce2" alt="mgcs-request-cluster-step1.png" width="3024" height="1442" data-path="assets/mgcs-request-cluster-step1.png" />

1. **Billing Method**: Select your preferred billing method
   * **Pay as you go**: Charged by the minute, no upfront cost, pay for what you use

2. **Data Center**: Choose your preferred data center location

<img src="https://mintcdn.com/gmicloud/i7S_g2j1X9iiXhO1/assets/mgcs-request-cluster-datacenter.png?fit=max&auto=format&n=i7S_g2j1X9iiXhO1&q=85&s=54d08f43f4a15b47b401e2b6b22b699d" alt="mgcs-request-cluster-datacenter.png" width="3024" height="1442" data-path="assets/mgcs-request-cluster-datacenter.png" />

3. **Kubernetes Version**: Select the Kubernetes version for your cluster (available after selecting a data center)

4. **Worker Node Specification**: Choose the GPU instance type for your worker nodes (available after selecting a data center)

5. **Worker Node Quantity**: Set the number of worker nodes for your cluster

6. Click "**Continue**" to proceed to the next step

### Step 2: OS Image

Select the operating system image for your worker nodes. The default is **Ubuntu 22.04 x86 64bits**.

### Step 3: Basic Information

Provide the basic information for your cluster, such as the cluster name.

After completing all steps, review the **Summary** panel on the right side which shows:

* Billing Method
* Data Center
* Kubernetes Version
* Worker Node Specification
* Worker Node Quantity
* OS Image
* **Estimated Monthly Cost** (List Price, Discount, and Estimated Total)

Click "**Submit**" to send your cluster request.

<Warning>
  The request for GPU Cluster Service is not processed automatically. After submitting the request, please contact support to activate it.
</Warning>

## View Cluster Requests

Track the status of your cluster requests on the Requests page.

1. Click "**Requests**" in the left sidebar under the "Cluster" section

<img src="https://mintcdn.com/gmicloud/i7S_g2j1X9iiXhO1/assets/mgcs-requests-sidebar-navigation.png?fit=max&auto=format&n=i7S_g2j1X9iiXhO1&q=85&s=8c1eb165fe657af7a85f900aaf097c36" alt="mgcs-requests-sidebar-navigation.png" width="3024" height="1442" data-path="assets/mgcs-requests-sidebar-navigation.png" />

2. Use the tabs to filter requests by status:
   * **All**: View all requests
   * **In Progress**: View requests currently being processed
   * **Error**: View requests that encountered errors