View Cluster Overview Monitor

You can view the CKS cluster overview monitoring to understand the overall status of cluster nodes, Pods, compute resources, memory, and I/O. You can access full monitoring from the OBSERVABILITY > Monitor page, or access the monitoring view for a target cluster from the Clusters page.

​

View Monitoring Data

1. Log in to the GPU Cloud console and go to OBSERVABILITY > Monitor. To view it from the cluster entry point, go to the Clusters page, click the name of the target cluster to open the cluster workspace, and then go to OBSERVABILITY > Monitor. When accessed from the cluster workspace, monitoring is scoped to the current cluster.
2. Select Cluster Overview under Resource Type.
3. Select a Region and the target CKS cluster.
4. To narrow the scope, select a namespace or Pod.
5. Select a time range.
6. View the overview cards and monitoring charts. The cluster overview provides the following information:

   Information	Description
   Node Quantity	The number of nodes in the current cluster.
   Pod Quantity	The number of pods within the current filter scope.
   GPU Computing Power Utilization	GPU computing power utilization within the current filter scope.
   CPU Utilization	CPU utilization within the current filter scope.

   Cluster overview monitoring includes the following metric groups:

   Group	Metrics	Description
   Compute utilization	Average CPU utilization (%), average GPU compute utilization (%), total GPU memory usage (GiB)	View overall CPU, GPU compute, and GPU memory usage for the cluster. The average CPU utilization chart can display both the req and limit curves.
   Memory	Average memory utilization (%), average GPU memory utilization (%)	View overall memory and GPU memory usage for the cluster. The average memory utilization chart can display both the req and limit curves.
   I/O throughput	Average network I/O (Mbps), average disk I/O (IO/s)	View overall network and disk I/O status for the cluster. The average disk I/O chart can display both the Read and Write curves.

   If some charts display No Data, check whether the selected scope contains workload or resource usage data of the corresponding type. If a chart fails to load, retain the resource, time range, and error information, and contact technical support.
7. To retrieve the latest data, click Refresh.

When viewing metrics, focus on trends in average CPU utilization, average GPU compute utilization, average memory utilization, average GPU memory utilization, average network I/O, and average disk I/O. When troubleshooting workload issues, evaluate them together with workload status, logs, and alarm history.