Cross-region replication (CRR) enables automatic and asynchronous replication of objects from one bucket to another in a different region, either within the same account or across accounts. CRR synchronizes operations such as object creation, overwriting, and deletion, supporting compliance, latency, security, and availability requirements.
Scenarios
CRR is ideal for cross-region disaster recovery and data replication. Destination buckets contain exact replicas of source bucket objects, including names, versioning information, content, and metadata such as creation time, owner, user metadata, and access control lists (ACLs). You can tailor CRR rules to suit various business needs.
-
Compliance requirements
OSS's data redundancy mechanism is enhanced by CRR, which replicates data across geographically distant data centers to fulfill compliance mandates.
-
Minimal latency
For customers in different geographical locations, replicating objects to data centers closer to them reduces access latency.
-
Data backup and disaster recovery
To ensure data security and availability, data written to one data center is replicated to another, providing a backup in case of catastrophic events.
-
Data migration
Business needs may necessitate migrating data from one OSS data center to another.
-
Operational purposes
For compute clusters analyzing the same objects in different data centers, maintaining replicas in both regions is beneficial.
Features
CRR offers the following features:
-
Replication time control (RTC)
With RTC enabled, OSS replicates most objects within seconds and achieves 99.99% replication within 10 minutes, along with real-time monitoring of replication tasks.
-
Near real-time replication of data
CRR rules can be configured to replicate operations like creation, deletion, and modification of objects in near real time.
-
Data consistency
OSS ensures eventual consistency for replicated data, except when an object with the same name exists in the destination bucket, which may affect consistency.
-
Historical data migration
CRR can replicate both new data following rule configuration and historical data from the source bucket.
-
Replication progress query
You can track the latest replication time and view the progress of historical data migration as a percentage.
-
Versioning
CRR maintains eventual consistency for versioning-enabled buckets, replicating delete markers but not the deletion of specific object versions.
-
Transfer acceleration
Transfer acceleration can be used to speed up data transfer for CRR, especially across regions inside and outside the Chinese mainland. For more information, see transfer acceleration.
-
Replication of encrypted data
CRR supports replication of both unencrypted objects and those encrypted with SSE-KMS or SSE-OSS. For more information, see CRR combined with server-side encryption.
-
Event notification
Event notification rules can be set to receive updates on object creation, update, deletion, and overwriting during replication by specifying event types such as
ObjectReplication:ObjectCreated
,ObjectReplication:ObjectRemoved
, andObjectReplication:ObjectModified
. For more information, see real-time processing of OSS file changes through event notification.
Notes
Billing
-
CRR incurs traffic charges in OSS. For details on billing methods, see CRR traffic fees. The account owning the source bucket bears the cost for cross-account CRR traffic.
The account that owns the source bucket incurs charges for the traffic generated by CRR across accounts.
-
Each successful object replication counts as a request and is subject to API call fees. For details on billing methods, see request fees.
-
Enabling transfer acceleration results in additional fees. For details on billing methods, see transfer acceleration fees.
-
Activating RTC incurs extra charges. For details on billing methods, see CRR RTC fees.
-
Replicating Infrequent Access (IA) or Archive objects with CRR does not involve data retrieval operations, thus no retrieval fees are charged.
Replication time
CRR typically replicates data asynchronously in near real time. Replication time can vary from minutes to hours based on data size. If replication is slow, check for bandwidth limits and consider submitting a ticket to increase bandwidth for better efficiency.
You can monitor the Record the bandwidth usage of the bucket in each segment tab. This metric reflects the bandwidth used for data replication to the destination bucket.
Used Bandwidthcross-region replication inbound data on the OSS console'sRisk of overwriting objects with the same name
Configuring a CRR rule for two buckets may result in an object from the source bucket overwriting an existing object with the same name in the destination bucket.
Limits
Region limits
-
Transfer acceleration is required for CRR between regions inside and outside the Chinese mainland.
Rule limits
Data from a source bucket can be replicated across multiple destination buckets. By default, up to 100 CRR (Cross-Region Replication) rules can be configured for a single bucket. Additionally, a bucket can serve simultaneously as a source in one CRR rule and as a destination in another.
If you require assistance with many replication rules in your business scenario, please contact , technical support, or .
Operational limits
-
The source and destination buckets in a CRR rule must have the same versioning state. Versioning-suspended buckets cannot be used in CRR rules.
-
Once a CRR rule is configured, the versioning status of the involved buckets cannot be changed.
-
An object replicated from the source bucket may overwrite an object with the same name in the destination bucket.
-
Data in a source bucket can be replicated to multiple destination buckets. By default, you can configure up to 100 CRR rules for a bucket. A bucket can be specified as a source bucket in a CRR rule and a destination bucket in another CRR rule at the same time. If your business scenario involves a larger number of replication rules, contact technical support.
-
Objects of the Cold Archive or Deep Cold Archive storage class cannot be replicated from a source bucket to a destination bucket.
-
Appendable objects in a source bucket cannot be replicated to a destination bucket of the Cold Archive or Deep Cold Archivestorage class.