tencent cloud

Tencent Cloud DataBuddy

Compute Resource Overview

ダウンロード
フォーカスモード
フォントサイズ
最終更新日: 2026-10-09 16:18:40
AI翻訳
This document describes the capability differences and selection recommendations of the four cluster types in DataBuddy, and explains the state lifecycle of compute resources.

Cluster Type Selection

DataBuddy provides four cluster types, each matching different workload characteristics:

Job Clusters

Suitable for automatically scheduled batch data processing, ETL jobs, and repeatable tasks.
Feature
Description
Engine
Spark
Billing Mode
Pay-as-you-go
Automatic Start/Stop
Not supported
Applicable Tasks
Notebook tasks, Python File tasks, and SQL tasks in Workflow
Network Settings
Supports inbound and outbound network connectivity. For details, see Compute Resource Network Settings.
Feature
Description
Engine
Ray
Billing Mode
Yearly/Monthly subscription
Automatic Start/Stop
Not supported
Applicable Tasks
Ray tasks in Workflow
Network Settings
Not supported.

Interactive Clusters

Suitable for multi-user collaborative ad hoc analysis and data exploration, with support for automatic stop when idle to save costs.
Feature
Description
Engine
Spark
Billing Mode
Pay-as-you-go
Automatic Start/Stop
Supported. You can configure automatic stop after N minutes without running tasks, and automatic start when tasks are running.
Applicable Tasks
SQL exploration, Dashboard, ChatBI, file upload to table
Network Settings
Supports inbound and outbound network connectivity. For details, see Compute Resource Network Settings.

Platform Clusters

Suitable for offline data ingestion, custom scripts, and platform-level service capabilities.
Feature
Description
Engine
DataX
Billing Mode
Pay-as-you-go
Automatic Start/Stop
Not supported
Applicable Tasks
Offline data ingestion tasks (table-to-table / table-to-Volume / file-to-table / file-to-Volume)
Network Settings
Supports outbound network connectivity. For details, see Compute Resource Network Settings.

Real-Time Clusters

Suitable for real-time data ingestion tasks, based on Oceanus (a stream computing engine).
Feature
Description
Engine
Flink(Oceanus)
Billing Mode
Pay-as-you-go
Automatic Start/Stop
Not supported
Applicable Tasks
Real-time whole-database multi-table ingestion, real-time sharded database and table ingestion
Network Settings
Supports outbound network connectivity. For details, see Compute Resource Network Settings.

Selection Recommendations

Select the appropriate cluster type based on your scenario, and avoid mixing types:
Application Scenario
Supported Cluster Type
Description
Batch processing tasks in workflows (Notebook, SQL, Python, Ray)
Job Cluster
Exclusive to a single job and automatically released after the job completes.
SQL exploration / Dashboard / ChatBI / file upload
Interactive Cluster
Shared by multiple users and isolated by CU; can be automatically stopped when idle.
Offline data ingestion (table / Volume / file)
Platform Cluster
Supports a wide range of offline data sources.
Real-time data ingestion (CDC, entire-database / sharded-database-and-table)
Real-time Cluster
Real-time engine, runs continuously.
Data source connectivity test
Platform clusters, real-time clusters
It is recommended to use the same cluster as the target task to ensure test accuracy.
Note:
Different cluster types differ significantly in their underlying engines, billing methods, and running modes, and cannot be migrated across cluster types. If the workload of a certain type of task changes, create a new cluster and migrate the task configuration.

CU and Billing

All clusters in DataBuddy use CU (Compute Unit) as the unified unit of measurement for compute resources:
1 CU ≈ 1 vCPU + 4 GB Memory
When you create a cluster, you can set a resource quota (CU limit), and billing is based on actual usage.
In a real-time cluster, in addition to the job usage itself, control usage (2 CU by default) and connector usage such as Meta Agent are also included.

Cluster State Lifecycle

Compute resources go through the following states throughout their lifecycle:
Status
Description
Applicable Cluster
Creating
After you click Create, the backend is initializing the cluster.
All
Pending start
Created successfully, not started yet (this state is available only for interactive clusters).
Interactive Cluster
Starting
After you click Start, the cluster is starting (this state is available only for interactive clusters).
Interactive Cluster
Running
The cluster is ready and can run tasks.
All
Stopping
After you click Stop (this state is available only for interactive clusters)
Interactive Cluster
Stopped
Stopped and can be restarted (this state is available only for interactive clusters).
Interactive Cluster
Failed
An error occurs during creation or running. A warning icon is displayed after the status. hover over the icon to view the specific cause.
All
Resizing
After you click Edit to modify the CU quota, the cluster is changing.
All
Isolating
The pay-as-you-go account has an overdue payment.
All
Deleting
After you click Delete, the cluster is being cleaned up.
All

States and Supported Operations

Supported operations in different states:
Status
Interactive Cluster
Job cluster / Platform cluster / Real-time cluster
Creating
-
-
Pending start
Start, Edit, Delete
Not applicable
Starting
-
Not applicable
Running
Stop, Edit
Top up, edit, and delete
Stopping.
-
Not applicable
Stopped
Start, Edit, Delete
Not applicable
Failed
Retry, Edit, Delete
Retry, Edit, Delete
Resizing
-
-
Isolated
Top up, delete
Top up, delete
Deleting
-
-
Note:
Job clusters, platform clusters, and real-time clusters enter the Running state immediately after they are created, without going through the Pending Start, Starting, or Stopped states. This is due to differences in the underlying engines. These three types of clusters are started on demand, and their resource usage is determined by the running status of the tasks.

References

ヘルプとサポート

この記事はお役に立ちましたか?

フィードバック