All Products
Search
Document Center

ApsaraMQ for RocketMQ:Global Replicator overview

Last Updated:Mar 11, 2026

When your business spans multiple regions, keeping messages and consumer progress in sync across ApsaraMQ for RocketMQ instances requires building and maintaining custom replication pipelines. Global Replicator eliminates this overhead by providing built-in cross-region message replication with seconds-level latency, supporting geo-disaster recovery, active geo-redundancy, and centralized data aggregation. Global Replicator is compatible with both ApsaraMQ for RocketMQ and open-source Apache RocketMQ clusters (version 4.4.0 or later).

What gets replicated

Data typeSupported
Normal messagesYes
Ordered messagesYes
Scheduled or delayed messagesYes
Transactional messagesYes
Consumer progress (offsets)Yes

Global Replicator supports both one-way and two-way synchronization. Two-way synchronization prevents circular replication by default.

If the source instance encounters an error, the quick synchronization feature syncs consumer progress from the destination group back to the source instance, reducing duplicate message consumption.

Message synchronization of ApsaraMQ for RocketMQ

How it works

Global Replicator replicates messages asynchronously through connectors. These connectors scale in seconds, apply rule-based tags to messages, and support resumable upload and download.

Cross-region network connectivity between instances is handled by Cloud Enterprise Network (CEN), delivering message synchronization with seconds-level latency.

During synchronization, Global Replicator adds tags to each message based on the tag key and value specified in user attributes. Consumer applications filter these tags using the SQL-92 filtering method to control whether they receive all data or only locally produced data.

Architecture for cross-region synchronization of data in ApsaraMQ for RocketMQ
Note

Global Replicator consumes read and write resources on the ApsaraMQ for RocketMQ instances involved. Evaluate your instance computing specifications before enabling the feature.

Use cases

All clusters involved in Global Replicator are independent instances with full read and write capabilities. Messages replicate asynchronously between them.

Data aggregation (one-way)

Aggregate messages from multiple regions or instances into a central region for unified processing. Each business unit processes data on the nearest server, avoiding cross-network latency. One-way synchronization replicates data from distributed regions and instances -- including open-source Apache RocketMQ and ApsaraMQ for RocketMQ clusters -- to a single central location.

  • Typical industries: Banking, securities, insurance

  • Architecture pattern: Multiple units, single center

Architecture for data aggregation in ApsaraMQ for RocketMQ

Geo-disaster recovery (one-way)

Maintain a standby instance in a secondary region that stays in sync through one-way replication. If the primary data center or region fails, switch traffic to the standby to restore service quickly.

  • Typical industries: All

  • Architecture pattern: Two regions, two data centers

Architecture for geo-disaster recovery in ApsaraMQ for RocketMQ

Under normal operation, no applications run in the secondary region, which reduces resource consumption and cost. During a failover, start the applications in the secondary region and use the consumer offset resetting feature to consume a minimum number of historical messages on the source instance.

Important

Because replication is asynchronous, a small amount of data may not have reached the standby instance at the time of failure. Implement message idempotence in your business logic to handle potential duplicates after failover.

Active geo-redundancy (two-way)

Deploy applications in multiple regions simultaneously, with two-way synchronization keeping messages in sync across all active instances. Both regions serve production traffic at all times, ensuring business continuity if one region becomes unavailable.

  • Typical industries: Finance, energy, and other industries where service continuity is critical

  • Architecture pattern: Two regions, two data centers

Architecture for active geo-redundancy in ApsaraMQ for RocketMQ

Global Replicator automatically tags synchronized messages. Producer applications require no code changes. Consumer applications use the SQL-92 filtering method to select either full data or only local data based on these tags.

During a regional failure, producer applications route all traffic to the healthy instance. Adjust the filtering conditions on consumer applications, and use the consumer offset resetting feature to consume messages from other regions with minimal duplication. Design your business logic to be idempotent so that duplicate messages do not affect downstream processing.

Message tagging details

The instance that an application connects to is determined by the configured endpoint. During synchronization, Global Replicator adds tags to each replicated message based on the tag key and value in the user attributes. Consumer applications use these tags to control whether they receive all data or only locally produced data.

For details on tagging behavior and consumer configuration, see Usage notes.

Adding tags to messages during synchronization in ApsaraMQ for RocketMQ

Benefits

  • Low-code setup: Replicate messages between instances without building custom synchronization pipelines.

  • Flexible configuration: Set up one-way or two-way synchronization and use tag-based filtering for selective message consumption.

  • Seconds-level cross-region latency: Built on EventBridge, Global Replicator delivers stable, low-latency message synchronization across regions.

Billing

Global Replicator incurs costs from two services:

  • EventBridge: Global Replicator uses EventBridge for message synchronization. For pricing details, see EventBridge billing.

  • ApsaraMQ for RocketMQ: Replication consumes read resources on the source instance and write resources on the destination instance. For pricing details, see ApsaraMQ for RocketMQ billing.

Supported editions, versions, and regions

Editions and versions

Edition / versionOne-way syncTwo-way sync
ApsaraMQ for RocketMQ 5.0YesYes
ApsaraMQ for RocketMQ 4.0 (all editions except Standard)YesYes
ApsaraMQ for RocketMQ 4.0 (Standard Edition)YesNo
ApsaraMQ for RocketMQ Enterprise Platinum Edition with Apache RocketMQ clusters (version 4.4.0 or later)YesYes

Global Replicator supports synchronizing data from one ApsaraMQ for RocketMQ instance to multiple destination instances.

Region mappings

Global Replicator is available in all supported regions of ApsaraMQ for RocketMQ. Global Replicator supports cross-region data communication over internal networks. The following table lists the supported source-to-destination region mappings.

SourceDestination
China (Hangzhou), China (Shanghai), China (Shenzhen), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Hohhot), China (Chengdu), China (Ulanqab), China (Heyuan), China (Guangzhou), China (Fuzhou - Local Region) Closing DownChina (Hangzhou), China (Shanghai), China (Shenzhen), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Hohhot), China (Chengdu), China (Ulanqab), China (Heyuan), China (Guangzhou), China (Fuzhou - Local Region) Closing Down
China (Hong Kong)China (Hong Kong)
Japan (Tokyo)Japan (Tokyo)
South Korea (Seoul)South Korea (Seoul)
SingaporeSingapore
Malaysia (Kuala Lumpur)Malaysia (Kuala Lumpur)
Indonesia (Jakarta)Indonesia (Jakarta)
Philippines (Manila)Philippines (Manila)
Thailand (Bangkok)Thailand (Bangkok)
Germany (Frankfurt)Germany (Frankfurt)
UK (London)UK (London)
US (Silicon Valley)US (Silicon Valley), US (Virginia)
US (Virginia)US (Silicon Valley), US (Virginia)
UAE (Dubai)UAE (Dubai)
SAU (Riyadh - Partner Region)SAU (Riyadh - Partner Region)

See also