Two-way synchronization: ApsaraDB for MongoDB sharded clusters - Data Transmission Service

Prerequisites

Create the source and destination ApsaraDB for MongoDB sharded cluster instances. For more information, see Create a sharded cluster instance.
Important
- Obtain an endpoint for each shard node of the source ApsaraDB for MongoDB sharded cluster instance, including the source instance for a reverse task. Ensure that the accounts and passwords are the same for all shards. For more information, see Apply for an endpoint for a shard.
- For information about supported versions, see Overview of data synchronization scenarios.
We recommend that the destination ApsaraDB for MongoDB instance has at least 10% more storage space than is used by the source ApsaraDB for MongoDB instance.
Set the replication.oplogGlobalIdEnabled parameter to true for all shard and config server nodes of both the source and destination instances. For more information, see Configure database parameters.

Note
If this parameter is not set to true, the precheck fails or returns the error two-way mongo must have gid.
The required preparations depend on whether you use the schema synchronization feature.
- If you do not use the schema synchronization feature of DTS
  
  In the Configure Objects and Advanced Settings step, clear the Schema Synchronization checkbox under Synchronization Types. You must create the databases and collections to be sharded in the target ApsaraDB for MongoDB instance and configure data sharding based on your business requirements. We recommend that you enable the balancer and perform pre-sharding. For more information, see Configure sharding to maximize shard performance and How do I handle uneven data distribution in a sharded MongoDB cluster?.
- If you use the schema synchronization feature of DTS
  
  In the Configure Objects and Advanced Settings step, select the Schema Synchronization checkbox under Synchronization Types. We recommend that you enable the balancer and perform pre-sharding after schema synchronization is complete. For more information, see How do I handle uneven data distribution in a sharded MongoDB cluster?.
Note
- Configuring data sharding prevents all data from synchronizing to a single shard, which avoids performance bottlenecks. Enabling the balancer and performing pre-sharding helps prevent data skew.
- This topic provides an example of configuring a DTS task before purchase. In this scenario, you do not need to specify the number of shards in the source ApsaraDB for MongoDB (sharded cluster) instance.
  
  If you purchase a DTS task before you configure it, you must specify the correct number of shards at the time of purchase.

Limitations

Type	Description
Source and destination database limits	Bandwidth requirement: The server where your source database resides must have sufficient outbound bandwidth. Otherwise, synchronization speed may be affected. The collections to synchronize must have a primary key or UNIQUE constraint, and the field values must be unique. Otherwise, duplicate data may appear in the destination database. Field names in the data must not contain the "." (dot) character; otherwise, data inconsistencies may occur. The _id field in the collections to synchronize must have unique values. Otherwise, data inconsistency may occur. If you synchronize at the collection level and need to edit objects (such as rename collections), one synchronization task supports up to 1,000 collections. If you exceed this limit, the task fails with an error after submission. To resolve this, split the collections into batches and configure multiple tasks. Or configure a full-database synchronization task. A single document in your source database cannot exceed 16 MB. Otherwise, the task fails. Your source database cannot be Azure Cosmos DB for MongoDB or Amazon DocumentDB (elastic cluster). Your source database must enable Oplog and retain Oplog for at least seven days. Otherwise, DTS may fail to capture data changes, causing task failure. In extreme cases, this may cause data inconsistency or loss. These issues are not covered by the DTS SLA. Important We recommend using Oplog to capture data changes. Change Streams do not support two-way synchronization. Two-way synchronization does not support scaling the number of shards for either the source or destination MongoDB sharded cluster. Otherwise, DTS tasks may fail and cause data inconsistency. The number of Mongos nodes in your source MongoDB sharded cluster instance cannot exceed 10. If the collections to synchronize contain TTL indexes, data inconsistency or latency may occur. Ensure no orphaned documents exist in the source and destination instances. Otherwise, data inconsistency or task failure may occur. For more information, see orphaned documents and How to clean orphaned documents in MongoDB (sharded cluster architecture). The source and destination databases must be ApsaraDB for MongoDB instances of the same architecture. Two-way synchronization is not supported for self-managed MongoDB instances or MongoDB instances of different architectures. Source database operation limits: During schema synchronization and full data synchronization, do not change the schema of databases or collections (including array-type data updates). Otherwise, the synchronization task may fail or cause data inconsistency between the source and destination databases. If you perform only full data synchronization, do not write new data to the source instance. Otherwise, data inconsistency may occur between the source and destination databases. While the synchronization instance is running, do not run commands that change data distribution on the objects to synchronize in the source database (for example, shardCollection, reshardCollection, unshardCollection, moveCollection, or movePrimary). Otherwise, data inconsistency may occur. If the Balancer in your source database balances data, latency may occur. DTS does not support connecting to MongoDB databases using SRV addresses. If the source database is MongoDB 5.0 or later and the destination database version is earlier than 5.0, you cannot sync `capped collection`s. If you attempt to sync them, the task fails or data inconsistency occurs between the source and destination databases. This is because, starting from MongoDB 5.0, the behavior of `capped collection`s changed to allow operations such as explicit deletion and increasing document size during updates, and earlier versions of the database kernel cannot support these new features.
Other limits	Before starting the task, add the shard key used by the destination to the data in the source. After starting the task, INSERT statements for the data to synchronize must include the shard key. UPDATE statements cannot modify the shard key. We recommend using the same MongoDB version for the source and destination databases. Or use a higher version for the destination to ensure compatibility. Synchronizing from a higher version to a lower version may cause compatibility issues. If the destination collection has a unique index or its `capped` property is set to `true`, concurrent replay is not supported during incremental synchronization. Only single-threaded writes are allowed. This may increase task latency. DTS does not support synchronizing data from the admin, config, and local databases. Transaction information is not preserved. Transactions in the source database become individual records in the destination database. When DTS writes data to the destination collection, if a primary key or unique key conflict occurs, DTS skips the write statement and keeps existing data in the destination collection. If your source MongoDB version is earlier than 3.6 and your destination MongoDB version is 3.6 or later, differences in execution plans may cause inconsistent field ordering in synchronized data. Field-value relationships remain unchanged. If your business uses text-matching queries on nested structures, assess the impact of inconsistent field ordering. Evaluate the performance of both the source and destination databases before synchronization. Run synchronization during off-peak hours. Otherwise, full data initialization consumes read and write resources, increasing database load. Full initialization runs INSERT operations concurrently. This creates fragmentation in the destination database collection. After full initialization, the collection space in the destination instance is larger than in the source instance. Do not write data to the destination database except through DTS during synchronization. Otherwise, data inconsistency may occur between the source and destination databases. For example, if you use DMS to perform online DDL operations while other data is written to the destination database, data loss may occur. A two-way synchronization task includes forward and reverse synchronization tasks. When you configure or reset the task, if the destination object of one task matches the synchronization object of the other task: Allow only one task to synchronize full and incremental data. The other task supports only incremental synchronization. Data from the source of the current task synchronizes only to the destination of the current task. It does not serve as source data for the other task. Because DTS writes data concurrently, the storage space used by the destination is 5% to 10% larger than the source. To get the count of documents in the destination MongoDB, use the syntax `db.$table_name.aggregate([{ $count:"myCount"}])`. Ensure the destination MongoDB does not have the same primary key (default is _id) as the source. Otherwise, data loss may occur. If the destination has the same primary key, clear related data in the destination without affecting your business (delete documents with the same _id value in the destination). Disable the Balancer in your source MongoDB database during full synchronization. Keep it disabled until all subtasks reach the incremental phase. Otherwise, data inconsistency may occur. For more information about managing the Balancer, see Manage the MongoDB Balancer. If you do not need the schema synchronization feature provided by DTS—for example, if data partitioning is already configured at the destination—do not select Configure Objects, and then select Synchronization Types and Schema Synchronization. Otherwise, sharding conflicts can lead to data inconsistency or task failure. If you switch traffic to the destination MongoDB database, ensure your business behavior meets MongoDB's requirements for sharded collections. DTS does not support synchronizing time-series collections introduced in MongoDB 5.0 and later. If a task fails, DTS support staff will attempt to restore it within eight hours. During restoration, they may restart the task or adjust its parameters. Note Only DTS task parameters are modified—not database parameters. Parameters that may be adjusted include those listed in Modify instance parameters.

Billing

Synchronization type	Pricing
Schema synchronization and full data synchronization	Free of charge.
Incremental data synchronization	Charged. For more information, see Billing overview.

Supported two-way synchronization architectures

DTS supports two-way synchronization only between two ApsaraDB for MongoDB sharded cluster instances. Topologies with more than two ApsaraDB for MongoDB instances are not supported.

Supported conflict detection

To ensure data consistency, make sure that records with the same primary key, business primary key, or unique key are updated on only one of the two synchronization nodes.

DTS uses conflict detection and resolution to maximize the stability of two-way synchronization instances. DTS can detect the following types of conflicts:

Uniqueness conflicts from INSERT operations
If a record that you try to insert into the destination instance conflicts with an existing record, DTS automatically ignores the INSERT operation.
Mismatched records from UPDATE operations
If the record that you try to update does not exist in the destination instance or conflicts with another record, DTS automatically ignores the UPDATE operation.
Non-existent records for DELETE operations
If the record that you try to delete does not exist in the destination instance, DTS automatically ignores the DELETE operation.

Important

Due to factors like system time differences and synchronization latency, DTS cannot guarantee that the conflict detection mechanism will prevent all data conflicts. When using two-way synchronization, you must adjust your business logic to ensure that records with the same primary key, business primary key, or unique key are updated on only one node at a time.
For the data synchronization conflicts described above, DTS provides a default resolution policy, Ignore, which cannot be changed.

Synchronization types

Type	Description
schema synchronization	Synchronizes the schemas of specified objects from the source ApsaraDB for MongoDB instance to the destination ApsaraDB for MongoDB instance.
full data synchronization	Synchronizes all existing data of specified objects from the source ApsaraDB for MongoDB instance to the destination ApsaraDB for MongoDB instance. Note Full data synchronization is supported for databases and collections.
incremental data synchronization	After full data synchronization completes, this process continuously synchronizes incremental updates from the source ApsaraDB for MongoDB instance to the destination ApsaraDB for MongoDB instance. Incremental data synchronization does not replicate databases created after the task starts. The following changes are replicated: CREATE COLLECTION and CREATE INDEX DROP DATABASE, DROP COLLECTION, and DROP INDEX RENAME COLLECTION Note The RENAME COLLECTION operation is not replicated if the `dropTarget` option is set to `true`. Document inserts, updates, and deletes within a collection. Note For document updates, it replicates only operations that use the `$set` command.

Procedure

Important

This section uses the scenario where you configure a DTS task before you purchase it as an example. In this case, you do not need to specify the number of Shards in the source ApsaraDB for MongoDB (sharded cluster). If you purchase the DTS task before you configure it, you must correctly specify the number of Shards during the purchase.

Go to the data synchronization task list page in the destination region. You can do this in one of two ways.
DTS console
1. Log on to the DTS console.
2. In the navigation pane on the left, click Data Synchronization.
3. In the upper-left corner of the page, select the region where the synchronization instance is located.
DMS console
Note
The actual steps may vary depending on the mode and layout of the DMS console. For more information, see Simple mode console and Customize DMS console layout and style.
1. Log on to the DMS console.
2. In the top menu bar, choose Data + AI > DTS (DTS) > Data Synchronization.
3. To the right of Data Synchronization Tasks, select the region of the synchronization instance.
Click Create Task to open the task configuration page.

Configure the source and destination databases.

Warning

After you select the source and destination instances, review the Limits at the top of the page. Otherwise, the task may fail or data inconsistency may occur.

Category	Parameter	Description
N/A	Task Name	DTS automatically generates a task name. We recommend that you specify a descriptive name for easy identification. The name does not need to be unique.
Source Database	Select Existing Connection	Select the registered database instance with DTS from the drop-down list. The database information below is automatically configured. Note In the DMS console, this configuration item is Select a DMS database instance. If you have not registered the database instance or do not need to use a registered instance, manually configure the database information below.
	Database Type	Select MongoDB.
	Connection Type	Select Alibaba Cloud Instance.
	Instance Region	Select the region where the source ApsaraDB for MongoDB instance is located.
	Replicate Data Across Alibaba Cloud Accounts	For this example, select No, as the database instance belongs to the current Alibaba Cloud account.
	Architecture	Select Sharded Cluster.
	Migration Method	Select Oplog.
	Instance ID	Select the instance ID of the source ApsaraDB for MongoDB instance.
	Authentication Database	Enter the name of the database to which the database account of the source ApsaraDB for MongoDB instance belongs. The default value is `admin`.
	Database Account	Enter the database account of the source ApsaraDB for MongoDB instance. The account must have the read permission on the databases to be synchronized, and on the `config`, `admin`, and `local` databases.
	Database Password	Enter the password for the specified database account.
	Shard account	Enter the shard account for the source ApsaraDB for MongoDB instance.
	Shard password	Enter the shard password for the source ApsaraDB for MongoDB instance.
	Encryption	DTS supports three connection methods: Non-encrypted, SSL-encrypted, and Mongo Atlas SSL. The options for Encryption vary based on the selected Access Method and Architecture. The options displayed in the console prevail. Note A MongoDB database where the Architecture is Sharded Cluster and the Migration Method is Oplog does not support SSL-encrypted. If the source is a self-managed MongoDB database (Access Method is not Alibaba Cloud Instance) with a Replica Set architecture, and you select SSL-encrypted, DTS also allows you to upload a CA certificate to verify the connection.
Destination Database	Select Existing Connection	Select the registered database instance with DTS from the drop-down list. The database information below is automatically configured. Note In the DMS console, this configuration item is Select a DMS database instance. If you have not registered the database instance or do not need to use a registered instance, manually configure the database information below.
	Database Type	Select MongoDB.
	Connection Type	Select Alibaba Cloud Instance.
	Instance Region	Select the region where the destination ApsaraDB for MongoDB instance is located.
	Replicate Data Across Alibaba Cloud Accounts	For this example, select No, as the database instance belongs to the current Alibaba Cloud account.
	Architecture	Select Sharded Cluster.
	Instance ID	Select the instance ID of the destination ApsaraDB for MongoDB instance.
	Authentication Database	Enter the name of the database to which the database account of the destination ApsaraDB for MongoDB instance belongs. The default value is `admin`.
	Database Account	Enter the database account of the destination ApsaraDB for MongoDB instance. The account must have the `dbAdminAnyDatabase` permission, the `readWrite` permission on the destination database, and the `read` permission on the `local` database.
	Database Password	Enter the password for the specified database account.
	Encryption	DTS supports three connection methods: Non-encrypted, SSL-encrypted, and Mongo Atlas SSL. The options for Encryption vary based on the selected Access Method and Architecture. The options displayed in the console prevail. Note MongoDB databases with an Architecture of Sharded Cluster do not support SSL-encrypted. If the destination is a self-managed MongoDB database (Access Method is not Alibaba Cloud Instance) with a Replica Set, and you select SSL-encrypted, DTS also allows you to upload a CA certificate to verify the connection.

After you complete the configuration, click Test Connectivity and Proceed at the bottom of the page.
Note
Ensure that the IP address blocks of the DTS service are added to the security settings of the source and destination databases, either automatically or manually, to allow access from DTS servers. For more information, see Add the IP address whitelist of DTS servers.

Configure the task objects.

On the Configure Objects page, specify the objects to synchronize.

Parameter	Description
Synchronization Types	DTS always selects Incremental Data Synchronization. By default, you must also select Schema Synchronization and Full Data Synchronization. After the precheck, DTS initializes the destination cluster with the full data of the selected source objects, which serves as the baseline for subsequent incremental synchronization.
Processing Mode of Conflicting Tables	Precheck and Report Errors: Checks if the destination database contains collections with the same names. If a collection with a duplicate name is found, an error is reported during the precheck phase and the data synchronization task does not start. Otherwise, the precheck is successful. Note If you cannot delete or rename the collection with the same name in the destination database, you can use the object name mapping feature to change the collection name in the destination database. For more information, see Set the name of a synchronization object in the destination instance. Ignore Errors and Proceed: Skips checking for collections with the same name in the destination database. Warning Selecting Ignore Errors and Proceed may cause data inconsistency and put your business at risk. For example: If a record in the destination database has the same primary key or unique key value as a record in the source database, the record in the destination database is retained. The record from the source database is not synchronized to the destination database. Data initialization may fail, only some data may be synchronized, or the synchronization may fail.
Synchronization Topology	Select Two-way Synchronization.
Exclude DDL Operations	Yes: DDL operations are not synchronized. No: DDL operations are synchronized. Note To ensure stability in bidirectional synchronization, DDL operations are synchronized only in the forward direction, not in reverse.
Conflict Resolution Policy	If a conflict described in How conflict detection works occurs, select a conflict resolution strategy that suits your business requirements. TaskFailed (Fail and report an error upon conflict) The task will error out and stop if a conflict occurs. The task enters a failed state and requires manual intervention for recovery. Ignore (Keep the existing record in the destination instance) The system skips the conflicting statement and continues the synchronization. The existing record in the destination instance is retained. Overwrite (Overwrite the conflicting record in the destination instance) If a conflict occurs, the system overwrites the conflicting record in the destination instance with the data from the source. Note This scenario supports only Ignore.
Source Objects	In the Source Objects box, click the objects, and then click to move them to the Selected Objects box. Note You can select objects at the DATABASE or COLLECTION level.
Selected Objects	To set the name of a synchronization object in the destination instance, right-click the object in the Selected Objects box to modify it. For more information, see Object name mapping. To remove a synchronization object, click it in the Selected Objects box, and then click to move it to the Source Objects box. Note To select incremental operations at the database or collection level, right-click the desired object in the Selected Objects box and make your selections in the dialog box that appears. To filter data (supported during full synchronization but not incremental synchronization), right-click the desired collection in the Selected Objects box and configure the settings in the dialog box that appears. For instructions, see Configure filter conditions. If you use the object name mapping feature to specify a database or collection to receive data, synchronization may fail for other objects that depend on the mapped object.

Click Next: Advanced Settings.

Parameter	Description
Dedicated Cluster for Task Scheduling	By default, DTS uses a shared cluster for tasks, so you do not need to make a selection. For greater task stability, you can purchase a dedicated cluster to run the DTS synchronization task. For more information, see What is a DTS dedicated cluster?.
Retry Time for Failed Connections	If the connection to the source or destination database fails after the synchronization task starts, DTS reports an error and immediately begins to retry the connection. The default retry duration is 720 minutes. You can customize the retry time to a value from 10 to 1,440 minutes. We recommend a duration of 30 minutes or more. If the connection is restored within this period, the task resumes automatically. Otherwise, the task fails. Note If multiple DTS instances (e.g., Instance A and B) share a source or destination, DTS uses the shortest configured retry duration (e.g., 30 minutes for A, 60 for B, so 30 minutes is used) for all instances. DTS charges for task runtime during connection retries. Set a custom duration based on your business needs, or release the DTS instance promptly after you release the source/destination instances.
Retry Time for Other Issues	If a non-connection issue (e.g., a DDL or DML execution error) occurs, DTS reports an error and immediately retries the operation. The default retry duration is 10 minutes. You can also customize the retry time to a value from 1 to 1,440 minutes. We recommend a duration of 10 minutes or more. If the related operations succeed within the set retry time, the synchronization task automatically resumes. Otherwise, the task fails. Important The value of Retry Time for Other Issues must be less than that of Retry Time for Failed Connections.
Enable Throttling for Full Data Synchronization	During full data synchronization, DTS consumes read and write resources from the source and destination databases, which can increase their load. To mitigate pressure on the destination database, you can limit the migration rate by setting Queries per second (QPS) to the source database, RPS of Full Data Migration, and Data migration speed for full migration (MB/s). Note This parameter is available only if Synchronization Types is set to Full Data Synchronization. You can also adjust the rate of full data synchronization when the synchronization instance is running.
Only one data type for primary key _id in a table of the data to be synchronized	Specify whether the data types of the `_id` primary key are unique within each collection to be synchronized. Important Select an option based on your actual data. An incorrect selection may lead to data loss. This parameter is available only if Synchronization Types includes Full Data Synchronization. Yes: The data types are unique. During the full synchronization phase, DTS does not scan the data types of the primary keys in the source data. For each collection, DTS synchronizes only the data corresponding to one primary key data type. No: The data types are not unique. During the full synchronization phase, DTS scans the data types of the primary keys in the source data and synchronizes all the data.
Enable Throttling for Incremental Data Synchronization	You can also limit the incremental synchronization rate to reduce pressure on the destination database by setting RPS of Incremental Data Synchronization and Data synchronization speed for incremental synchronization (MB/s).
Environment Tag	You can select an environment tag to identify the instance based on your requirements. This is not required for this example.
Configure ETL	Choose whether to enable the extract, transform, and load (ETL) feature. For more information, see What is ETL? Valid values: Yes: Enables the ETL feature. Enter data processing statements in the code editor. For more information, see Configure ETL in a data migration or data synchronization task. No: Disables the ETL feature.
Monitoring and Alerting	Choose whether to set up alerts. If the synchronization fails or the latency exceeds the specified threshold, DTS sends a notification to the alert contacts. No: No alerts are configured. Yes: Configures alerts. You must also set the alert threshold and alert notifications. For more information, see Configure monitoring and alerting during task configuration.

Click Data Verification to configure a data verification task.
To use the data verification feature, see Configure data verification.

Save the task and perform a precheck.
- To view the parameters for configuring this instance via an API operation, hover over the Next: Save Task Settings and Precheck button and click Preview OpenAPI parameters in the tooltip.
- If you have finished viewing the API parameters, click Next: Save Task Settings and Precheck at the bottom of the page.
Note
- Before a synchronization task starts, DTS performs a precheck. You can start the task only if the precheck passes.
- If the precheck fails, click View Details next to the failed item, fix the issue as prompted, and then rerun the precheck.
- If the precheck generates warnings:
  For non-ignorable warning, click View Details next to the item, fix the issue as prompted, and run the precheck again.
  For ignorable warnings, you can bypass them by clicking Confirm Alert Details, then Ignore, and then OK. Finally, click Precheck Again to skip the warning and run the precheck again. Ignoring precheck warnings may lead to data inconsistencies and other business risks. Proceed with caution.

Purchase the instance.

When the Success Rate reaches 100%, click Next: Purchase Instance.

On the Purchase page, select the billing method and link specifications for the data synchronization instance. For more information, see the following table.

Category	Parameter	Description
New Instance Class	Billing Method	Subscription: You pay upfront for a specific duration. This is cost-effective for long-term, continuous tasks. Pay-as-you-go: You are billed hourly for actual usage. This is ideal for short-term or test tasks, as you can release the instance at any time to save costs.
	Resource Group Settings	The resource group to which the instance belongs. The default is default resource group. For more information, see What is Resource Management?.
	Instance Class	DTS offers synchronization specifications at different performance levels that affect the synchronization rate. Select a specification based on your business requirements. For more information, see Data synchronization link specifications.
	Subscription Duration	In subscription mode, select the duration and quantity of the instance. Monthly options range from 1 to 9 months. Yearly options include 1, 2, 3, or 5 years. Note This option appears only when the billing method is Subscription.

Read and select the checkbox for Data Transmission Service (Pay-as-you-go) Service Terms.
Click Buy and Start, and then click OK in the OK dialog box.
You can monitor the task progress on the data synchronization page.

Configure the reverse synchronization task.
1. Wait for the forward synchronization task's Status to become Running, which indicates that initialization is complete.
2. Locate the reverse synchronization task and click Configure Task.
3. Configure the reverse synchronization task by following Step 3 to Step 6.
  Important
  
  When configuring the reverse task, ensure the source and destination instances are swapped: the forward task's destination is the reverse task's source, and vice versa. Carefully verify the consistency of instance information, such as database names, accounts, and passwords.
  
  We recommend that you do not change mapping names when you configure the reverse task. Otherwise, data inconsistency may occur.
  
  You cannot modify the Instance Region for the source and destination databases of the reverse synchronization task. Fewer parameters are required for the reverse task than for the forward task. The actual options are displayed in the console.
  
  The Processing Mode of Conflicting Tables setting for the reverse task ignores objects that were synchronized by the forward task.
  
  The reverse synchronization task does not support synchronizing objects that are in the Selected Objects list of the forward task.
  
  The reverse synchronization task automatically filters DDL operations.
4. When the Success Rate is 100%, click Back.
After the second task is configured, wait for the Status of both links to become Running. At this point, the bidirectional data synchronization is fully configured.