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

# CKS Cluster Overview

> Determine whether a CKS cluster suits your business and learn the workflow from preparing the runtime environment to maintaining workloads.

CKS is an unmanaged Kubernetes cluster provided by the GPU Cloud. It is suitable for running training, inference, data processing, or online services in containers, while using Kubernetes to centrally schedule computing resources, organize workloads, and maintain the runtime environment.

## Applicable Scenarios

When your workloads require container orchestration, coordinated execution of multiple tasks, or ongoing maintenance of online services, you can choose a CKS cluster.

| Business Requirement | Role of a CKS Cluster |
| - | - |
| Run training, inference, or batch processing tasks in parallel | Use workloads such as Pods, Jobs, and CronJobs to organize tasks and view their status centrally. |
| Continuously run online inference services or business applications | Use resources such as Deployments and StatefulSets to maintain replicas, configurations, and runtime status. |
| Share computing resources across teams or multiple projects | Use namespaces, node pools, and access credentials to define resource and management boundaries. |
| Deliver business environments through images | Package the runtime, dependencies, and code into container images, then submit them to the cluster for execution. |
| Provide shared data for workloads | Use Kubernetes storage objects such as PVCs, PVs, and StorageClasses to access persistent storage. |
| Centrally observe and maintain business operations | Assess operational status across clusters, nodes, workloads, monitoring, logs, and alarms. |

If you need only one or a small number of compute resources for development and debugging, temporary experiments, single-node training, or manual O\&M, you can prioritize CES instances. For more information, see [Select Computing Resources](https://docs.scitix.ai/gpu-cloud/get-started/gpu-cloud-overview.md#select-computing-resources).

## Usage Process

1. [Create a CKS cluster](https://docs.scitix.ai/gpu-cloud/cluster/create-kubernetes-cluster.md) and select the Kubernetes version, compute resources, and initial node pool.
2. [View cluster status and configuration](https://docs.scitix.ai/gpu-cloud/cluster/view-cluster-status-and-configuration.md) and confirm that the cluster is available.
3. Depending on the access method, [create a cluster access role](https://docs.scitix.ai/gpu-cloud/cluster/create-cluster-access-role.md) or [access the cluster using KubeConfig](https://docs.scitix.ai/gpu-cloud/cluster/access-cluster-with-kubeconfig.md).
4. Scale nodes and install components as required by your workloads, and prepare namespaces, configurations, and persistent storage.
5. Select a workload suitable for your business requirements and submit the YAML. Configure Ingress when you need to provide an entry point for HTTP or HTTPS services.
6. Monitor business operations through workload status, monitoring, logs, and alerts, and clean up resources that are no longer needed when operations end.

## Core Resources

| Object | Purpose |
| - | - |
| Cluster | The management boundary for Kubernetes resources and configurations. |
| Node pool | Organizes a group of nodes with the same specifications, billing method, and OS, and can be used to divide resource pools for different services. |
| Node | Provides the compute resources required to run Pods and other Kubernetes workloads. |
| Namespace | Isolates and organizes workloads, configurations, and storage claims within a cluster. |
| Workload | Includes Kubernetes resources such as Pods, Deployments, StatefulSets, DaemonSets, Jobs, and CronJobs. |
| Configuration | ConfigMaps store non-sensitive configuration, while Secrets store sensitive information such as passwords, tokens, and certificates. |
| Persistent storage | Workloads request storage through PVCs and mount bound storage to Pods. |
| Application | A release deployed to a cluster through Helm. |

## Maintenance Responsibilities and Precautions

* A CKS cluster is an unmanaged cluster. You must maintain the Kubernetes resources, workloads, and runtime environment within the cluster.
* Node specifications and inventory may affect the creation of clusters or node pools.
* Creating, modifying, or deleting Kubernetes resources may affect running tasks. Before performing any actions, confirm the resource scope, dependencies, and data impact.
* Deleting a cluster releases its nodes and deletes the Kubernetes resources and node disk data within the cluster. Data in persistent storage is retained.

## Related Documentation

* [Expand and Maintain Cluster Nodes](https://docs.scitix.ai/gpu-cloud/cluster/expand-and-maintain-cluster-nodes.md)
* [Select and Install Cluster Components](https://docs.scitix.ai/gpu-cloud/cluster/choose-and-install-cluster-components.md)
* [CKS cluster persistent storage overview](https://docs.scitix.ai/gpu-cloud/cluster/persistent-storage-overview.md)
* [Deploy and Troubleshoot Workloads](https://docs.scitix.ai/gpu-cloud/cluster/deploy-and-troubleshoot-workloads.md)
* [View Cluster Overview Monitoring](https://docs.scitix.ai/gpu-cloud/monitoring/view-cluster-overview-monitoring.md)
* [Query and Analyze Logs](https://docs.scitix.ai/gpu-cloud/log/query-logs.md)
* [Alerts Overview](https://docs.scitix.ai/gpu-cloud/alarm/alarm-overview.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.