Cluster Monitoring

When self-built or connected clusters on the platform have deployed monitoring components and the monitoring components are running normally, you can view cluster monitoring data through a visual interface.

• Cluster Overview: View basic resource monitoring information of the cluster, including resource status, resource usage, resource trend statistics, network load statistics, and network health status.

• View Component Monitoring Data: View monitoring information of cluster control components to understand the running status and resource usage of the components in real-time.

Prerequisites

• Cluster status is normal.

• Monitoring components are deployed in the cluster and running normally.

Monitoring Data Operation Instructions

• Control Refresh: The monitoring data on the cluster overview and monitoring pages automatically refresh once when the page is opened. To refresh the monitoring data again, you can do so in two ways:

  • Manual Refresh: Click on at the bottom right of the page to manually refresh the monitoring data;

  • Set Auto-Refresh (default off): Click on to set the time interval for auto-refreshing the monitoring data.

• View Large Chart: Click on at the top right of the chart to view an enlarged chart with refined data display granularity.

• View/Set Legend: Click on at the top right of the chart to expand the legend in the monitoring chart. Click on a legend to hide/show the corresponding curve in the chart.

• View PromQL Expression: Hover over a chart and click on icon to view the PromQL expression.

Cluster Overview

1. In the left navigation bar, click Clusters > Cluster of clusters.

2. Click on the cluster name to be viewed.

   On the overview page, you can view basic information about the cluster, resource status on the cluster, current resource usage statistics, and corresponding trend statistical charts.

   Explanation of resource statistics:

   • Utilization Rate = Resource usage value / Total resource value

   • Request Rate = Resource request value / Total resource value

   • Allocation Rate = Allocated resource value / Total resource value

   • System Load: Average CPU load in 1 minute, 5 minutes, and 15 minutes. It reflects the system’s busyness by comparing the number of processes currently being executed or waiting to be executed by the CPU with the maximum number of processes that the CPU can execute.

   

View Component Monitoring Data

You can view monitoring information of core Kubernetes components in the cluster to help Platform Management personnel or cluster operators understand the running status and resource usage of the components in real-time, avoiding disruptions to cluster functionality due to abnormal component states.

Procedure of Operation

1. Click on the Monitoring tab at the top of the page to access the monitoring information page of the cluster components.

   Tip: Click on the secondary tabs named after the components to switch to the corresponding component’s monitoring information page.

   The monitoring information for each component mainly consists of Monitoring Overview, Component Pods List, and Monitoring Trend Statistics, as shown in the image below (using ApiServer as an example). Detailed explanations of monitoring information can be found in the Reference Information .

   

   Currently, components that support viewing monitoring overview and monitoring trend statistics data include:

   • ApiServer : Provides various HTTP Rest APIs for Kubernetes resource objects such as POST, DELETE, PUT, GET, and WATCH, serving as the data bus and data center for the entire system.

   • Controller Manager : Acts as the management control center within the cluster, responsible for managing nodes, Pod replicas, service endpoints, namespaces, service accounts, and resource quotas within the cluster. In case of a node failure, the Controller Manager promptly detects it and executes automated recovery processes to ensure the cluster operates as expected.

   • Scheduler : Operates on the master node and its core function is to listen to the apiserver to obtain Pods with PodSpec.NodeName empty, creating a binding for each such Pod to indicate which node it should be scheduled on.

   • ETCD : A highly available distributed key-value storage system providing features like data TTL expiration, data change monitoring, multi-values, directory watching, distributed lock atomic operations, used for service discovery, shared configuration, managing cluster node statuses, and ensuring node data consistency.

Reference Information

ApiServer

Below are the monitoring information field explanations related to the ApiServer component.

Table: ApiServer Monitoring Overview

Parameter	Description
Total CPU Usage	Total CPU usage by Pods running the component.
Total Memory Usage	Total memory usage by Pods running the component.
Total Restarts in 5 Minutes	Total number of restarts by Pods running the component in the last 5 minutes. : Restarts = 0; : 1 <= Restarts <= 3; : Restarts > 3.
Total Requests in 5 Minutes	Total number of HTTP requests received by Pods running the component in the last 5 minutes.

Table: ApiServer Pods List

Table: ApiServer Monitoring Trend Statistics

Controller Manager

Explanation of monitoring information fields related to the Controller Manager component is provided in the following table.

Table: Controller Manager Monitoring Overview

Parameter	Description
Total CPU Usage	Total CPU usage by Pods running the component.
Total Memory Usage	Total memory usage by Pods running the component.
Total Restarts in 5 Minutes	Total number of restarts by Pods running the component in the last 5 minutes. : Restarts = 0; : 1 <= Restarts <= 3; : Restarts > 3.

Table: Controller Manager Pods List

Table: Controller Manager Monitoring Trend Statistics

Scheduler

Scheduler

Scheduler Monitoring Overview

Table: Scheduler Monitoring Overview

Parameter	Description
Total CPU Usage	The total CPU usage by Pods running the component.
Total Memory Usage	The total memory usage by Pods running the component.
Total Restarts in the Last 5 Minutes	The total number of restarts by Pods running the component in the last 5 minutes. : Restart count = 0; : 1 <= Restart count <= 3; : Restart count > 3.
Total Schedules in the Last 5 Minutes	The total number of successful schedules by the component in the last 5 minutes.

Table: Scheduler Pods List

Table: Scheduler Monitoring Trend Statistics

ETCD

ETCD Basic Monitoring

Table: ETCD Monitoring Overview

Parameter	Description
Total CPU Usage	The total CPU usage by Pods running the component.
Total Memory Usage	The total memory usage by Pods running the component.
Total Restarts in the Last 5 Minutes	The total number of restarts by Pods running the component in the last 5 minutes. : Restart count = 0; : 1 <= Restart count <= 3; : Restart count > 3.
Total Leader Changes in the Last Day	The total number of leader changes by Pods running the component in the last day. Note: Frequent leader changes significantly affect ETCD performance. Unstable leaders may be due to network connection issues or high ETCD cluster load.

ETCD

ETCD Pods List

ETCD Monitoring Trend Statistics

ETCD Status

Table: ETCD Status

OS-Level CPU/Memory/IO Status

Table: OS-Level CPU/Memory/IO Status

Process-Level IO Status

Table: Process-Level IO Status

Collector Status

Table: Collector Status