Docs

Torch Object Storage

Data subscription

Creating data subscription rules

You can set up data subscription rules to automatically copy data from the source service to a TOS bucket, making data acquisition easier and more efficient. This topic describes how to do so.

Notes

When using the data subscription feature, pay attention to the following:

Supported regions: Currently, China (Beijing), China (Shanghai), and China (Guangzhou) are supported. If you need to use data subscription in other regions, please contact your account manager for assistance.

Supported bucket types: Currently, only FNS buckets can be selected as the destination bucket. HNS buckets are not supported. For more information, see Bucket type.

Out-of-order replication: Objects may not be replicated to your destination bucket in the same order they were written to the source.

Guaranteed success rate: This feature uses asynchronous replication. While it does not ensure 100% replication success, it offers a reliability of at least 99.9% and an automatic retry mechanism.

Supported inference endpoint types: This feature only supports content generated via VeArk inference endpoints, including both preset and customized ones.

Historical data: A data subscription rule only applies to data generated after its creation. It does not apply to data that existed before the rule was created.

Scope of application: This feature only applies to scenarios that meet all of the following conditions:

Data source: Data for subscription must come from the VeArk platform.

Supported models:

Model category	Model name
Image generation	doubao-seedream-5.0
		doubao-seedream-4.5
		doubao-seedream-4.0
		doubao-seedream-5.0-pro
		doubao-seedream-5.0-flash
Video generation	doubao-seedance-2.0
		doubao-seedance-2.0-fast
		doubao-seedance-2.0-mini
		doubao-seedance-2.5

Prerequisites

Before creating a data subscription rule, make sure that:

You have created a TOS bucket to store the subscribed data. For instructions, see Creating buckets.

You have permission to write data to your bucket via data subscription. When creating a data subscription rule for the first time, click One-click authorization in the console to create a service-linked role (TosSubscribeToArkRole) and grant it the following key permissions:

tos:PutObject: Used to write VeArk outputs to the specified bucket.

You have activated VeArk for your account.

Steps

In the left navigation pane, click Data subscription.

On the data subscription page, click Create data subscription.

On the Create data subscription rules page, configure the following parameters.

Parameter	Description
Rule name	Enter a name for your data subscription rule. Naming rules: Rule names can contain Chinese characters and English letters. They cannot exceed 255 characters in length. Each rule name must be unique within the same account.
Enabled status	The status of the data subscription rule. Enabled (Default): The rule takes effect immediately after creation and begins replicating eligible data. Disabled: The rule is created, but remains inactive. It can be manually enabled later.
Data source	Currently, only ARK is supported.
Inference access point	The inference endpoints supported by VeArk. You can select multiple endpoints. tip Inference endpoints must be unique across different rules.
Destination bucket region	The region of the bucket where the replicated objects are stored. The available destination buckets depend on the region you select.
Destination bucket	The name of the destination bucket. Select an existing bucket in the current region to store VeArk outputs. The source data and the destination bucket must be located in the same country. Cross-border replication is not supported. Currently, only FNS buckets can be selected as the destination bucket.
Destination storage type	The storage type used when the replicated objects are stored in the destination bucket. The following storage types are currently available for data subscription: Standard Storage (Default), Infrequent Access Storage, Flashback Archive, Archive, Cold Archive, and Deep Cold Archive. We recommend choosing a storage type based on access frequency and cost.
Destination object renaming rules	To organize the object structure within the destination bucket, you can use a fixed string as the prefix for the replicated object name. For example, use my-aigc-images/ to store all outputs under that directory. Example: Suppose the name of the source object is task123.jpg and the prefix is set to generated/. The resulting object name would be generated/task123.jpg. tip Follow these guidelines when specifying a prefix: Prefixes can contain Chinese characters, letters, numbers, and common symbols. The prefix length must not exceed 500 characters. Prefixes cannot start with a forward slash (/) or backslash (\). Consecutive forward slashes (/) are not allowed. .. cannot be used as a directory name. If the name of a replicated object is empty, the replication task will fail. If you do not define a prefix rule, the source object name is copied as is.
Overwrite namesake object policy	Specify how to handle an existing object with the same name in the destination bucket. Overwrite (Default): Replace the destination object with the source object. Skip: Keep the destination object and discard the source object for this replication.

Click Select to finish creating the rule.

Related operations

On the data subscription page, you can perform the following actions:

Enable/disable: Toggle the switch in the Status column of a rule to quickly activate or pause it.

Edit: Click Edit in the Actions column of a rule to modify its configurations, such as inference endpoint and fixed prefix.

Delete: Click Delete in the Actions column to permanently delete a rule that you no longer need.

Last updated: 2026.10.22 17:37:07