This document describes how to view resource usage trends, load status, and task running status on the compute resource details page, helping you evaluate resource utilization and adjust CU quotas as needed.
Overview
The compute resource details page provides two types of information:
Category | Content. |
Configuration Details | Basic cluster information (resource name, ID, type, quota, billing mode, creator, update time, description) and edit entry |
Monitoring Metric | Resource usage trends, load trends, and task running status |
Monitoring metrics are used to answer two core questions:
What is the current resource utilization of the cluster? Does it need to be scaled out or scaled in?
What tasks are running or queued on the cluster? Do you need to terminate abnormal tasks?
Accessing the Monitoring Page
1. Select Compute Resources in the left sidebar.
2. Click the name of the target resource in the list to go to the details page.
Resource Usage Trends
The top section displays a trend chart of resource usage and load:
Metric | Description | Calculation Formula |
Resource usage | Number of CUs currently consumed | Total CUs actually occupied by all tasks in the cluster |
Load | Cluster resource usage rate | Current CPU usage ÷ total CPU quota |
Time Range
Option | Description |
Last 24 hours | Default Value |
Last 7 days | Query data from the last 7 days. |
Last 30 days | Query data from the last 30 days. |
Custom | Custom time range, with data query support for up to the last year. |
Switching Views
You can switch between the Usage / Load views above the chart, with a minimum granularity of 5 seconds.
Note:
When the load remains above 80%, consider scaling out the cluster (increase the CU quota limit in Edit Cluster). When the load remains below 20% for a long period, consider scaling in the cluster to reduce costs.
Task List
The following section displays the task running status on the current cluster by task type. The task list displayed varies by cluster type:
Cluster Type | Task categories displayed |
Job Cluster | Notebook tasks, Python File tasks, Ray tasks, SQL tasks |
Interactive Cluster | SQL exploratory queries, Dashboard queries, ChatBI queries |
Platform Cluster | Offline data ingestion tasks |
Real-time Cluster | Real-time whole-database multi-table ingestion tasks, real-time sharded database and table ingestion tasks (only running tasks are displayed) |
Two tabs, Running and Queued, are displayed under each task type (real-time clusters only have the "Running" Tab). You can search by task name or instance ID, and manually refresh the list.
Viewing Configuration Details
In Configuration Details, you can view the complete configuration information of the cluster:
Field | Description |
Compute Resource Status | |
Compute Resource Name/ID | The name is entered by you; the ID is automatically generated; quick copy is supported. |
Compute Resource Type | Job cluster / Interactive cluster / Platform cluster / Real-time cluster |
Resource quota | Configured CU quota or single-instance quota / Maximum concurrency per instance / Number of instances (corresponding fields displayed by cluster type) |
Automatic Start/Stop | Displayed only for interactive clusters: "Automatically stops after N minutes without running tasks, and automatically starts when tasks are running." |
Billing Mode | Pay-as-you-go |
Created By/Updated At | Cluster creator and last updated time |
Compute Resource Description | The description you entered. Click to enter edit mode. |
The upper right corner of the details page supports the following operations (availability depends on the cluster type and status, see Status and Executable Operations for details): Start and Stop (interactive clusters only)
Editing
Deletion
Top Up (available for pay-as-you-go clusters in the isolated state)
Permissions
Refresh
Use Limits
Real-time Task List and Running Status: The list data is refreshed every 5 seconds. To view the latest status immediately, click Refresh in the upper right corner of the list.
Cross-Cluster View: Cluster monitoring displays only tasks on the current cluster. To view the global task running status across clusters, go to the Ops monitoring page of the corresponding task module (such as Workflow Run Records). Historical Data Retention: A custom time range supports querying data from up to the past year. If the range is exceeded, use the historical details in the corresponding task module.
FAQs
Q1: Why is the resource usage in the monitoring chart not exactly the same as the cost on the bill?
The monitoring chart displays the instantaneous CU usage (sampled at 5-second granularity). The cost on the bill is the cumulative CU * hours after per-second integration. Short-term fluctuations in instantaneous values do not directly equate to cost changes.
Q2: Why does an interactive cluster have a long-term load below 20% but still incur high costs?
In the Running state, an interactive cluster retains the minimum CU usage of a single instance, so a base fee is incurred even when no tasks are running. We recommend that you enable Auto-Stop When Idle to automatically stop the cluster when it is idle and avoid the base fee.
Q3: Can I view the resource usage of a specific task during a historical period?
Cluster monitoring displays only tasks that are currently running or queued. To view the resource usage of historical tasks, go to the run instance details page of the corresponding task or Workflow Run Records. References