Database Migration Service Release Notes

Features and enhancements

The following table shows the new features and enhancements introduced in AWS DMS version 3.4.4.

• AWS DMS now supports TLS encryption and TLS or SASL authentication using Amazon MSK and on-premises Kafka cluster as a target. For more information on using encryption and authentication for Kafka endpoints, see Connecting to Kafka using Transport Layer Security (TLS).

Issues resolved in AWS DMS 3.4.4

The issues resolved in AWS DMS 3.4.4 include the following:

• Improved AWS DMS logging on task failures when using Oracle endpoints.
• Improved AWS DMS task execution continues processing when Oracle source endpoints switch roles after Oracle Data Guard fail over.
• Improved error handling treats ORA—12561 as a recoverable error when using Oracle endpoints.
• Fixed an issue where EMPTY_BLOB() and EMPTY_CLOB() columns are migrated as null when using Oracle as a source.
• Fixed an issue where AWS DMS tasks fail to update records after add column DDL changes when using SQL Server as a source.
• Improved PostgreSQL as a source migration by supporting the TIMESTAMP WITH TIME ZONE data type.
• Fixed an issue where the afterConnectScript setting does not work during a full load when using PostgreSQL as a target.
• Introduced a new mapUnboundedNumericAsString setting to better handle the NUMERIC date type without precision and scale when using PostgreSQL endpoints.
• Fixed an issue where AWS DMS tasks fail with “0 rows affected” after stopping and resuming the task when using PostgreSQL as a source.
• Fixed an issue where AWS DMS fails to migrate the TIMESTAMP data type with the BC suffix when using PostgreSQL as a source.
• Fixed an issue where AWS DMS fails to migrate the TIMESTAMP value “±infinity” when using PostgreSQL as a source.
• Fixed an issue where empty strings are treated as NULL when using S3 as a source with the csvNullValue setting set to other values.
• Improved the timestampColumnName extra connection attribute in a full load with CDC to be sortable during CDC when using S3 as a target.
• Improved the handling of binary data types in hex format such as BYTE, BINARY, and BLOB when using S3 as a source.
• Fixed an issue where deleted records are migrated with special characters when using S3 as a target.
• Fixed an issue to handle empty key values when using Amazon DocumentDB (with MongoDB compatibility) as a target.
• Fixed an issue where AWS DMS fails to replicate NumberDecimal or Decimal128 columns when using MongoDB or Amazon DocumentDB (with MongoDB compatibility) as a source.
• Fixed an issue to allow CDC tasks to retry when there is a fail over on MongoDB or Amazon DocumentDB (with MongoDB compatibility) as a source.
• Added an option to remove the hexadecimal “0x” prefix to RAW data type values when using Kinesis, Kafka, or OpenSearch as a target.
• Fixed an issue where validation fails on fixed length character columns when using Db2 LUW as a source.
• Fixed an issue where validation fails when only the source data type or the target data type is FLOAT or DOUBLE.
• Fixed an issue where validation fails on NULL characters when using Oracle as a source.
• Fixed an issue where validation fails on XML columns when using Oracle as a source.
• Fixed an issue where AWS DMS tasks crash when there are nullable columns in composite keys using MySQL as a source.
• Fixed an issue where AWS DMS fails to validate both UNIQUEIDENTIFIER columns from SQL Server source endpoints and UUID columns from PostgreSQL target endpoints.
• Fixed an issue where a CDC task does not use an updated source table definition after it is modified.
• Improved AWS DMS fail over to treat task failures caused by an invalid user name or password as recoverable errors.
• Fixed an issue where AWS DMS tasks fail because of missing LSNs when using RDS for SQL Server as a source.

Original source Report a problem

Change in Handling of Large Numeric Values for Streaming Targets

In AWS DMS version 3.5.1, there is a change in how large integer and high-precision numeric values are handled when streaming data to targets like Kafka and Kinesis. Specifically, AWS DMS changed its internal data type representation, handling these values as INT64 instead of INT8. This shift can result in different data formats on the streaming endpoints, particularly when the values exceed the limits of INT8. Consequently, the representation of these numeric types may differ from their previous formatting when streamed to destinations like Kafka and Kinesis, potentially impacting downstream systems and processes that consume the data from these targets.

Summary of Change:

• In previous versions (e.g., 3.4.7/3.4.6), large integer values were represented as integers.
• Starting with version 3.5.1, these values may appear in scientific notation (e.g., 7.88129934789981E15), potentially leading to precision and formatting differences.

Affected Data Types

The recent change affects the representation of several numeric types when streamed to endpoints like Kafka and Kinesis. The impacted types are:

• Large integer types (e.g., bigint)
• Floating-point types (FLOAT, DOUBLE)
• High-precision decimal types (DECIMAL, NUMERIC)

Affected Scenarios:

• Full load migrations to streaming targets
• Change Data Capture (CDC) to streaming targets

This change specifically impacts streaming endpoints such as Kafka and Kinesis, while non-streaming targets remain unaffected.

To mitigate this change, you can implement a data type transformation that reverts to the previous formatting, representing large numbers as integers. However, it's important to note that this workaround may not be suitable for all scenarios, as it could potentially introduce limitations or compatibility issues.

Recommendations:

• Test your specific use case in a non-production environment before deploying AWS DMS version 3.5.1 or later to identify and address any impacts of this change.
• Affected customers can implement the change-data-type transformation workaround, if applicable, to revert to the previous formatting for large numbers as integers. However, this approach may not suit all scenarios.

We are reviewing this behavior to ensure consistent data type handling across endpoints in future releases.

The following table shows the new features and enhancements introduced in AWS Database Migration Service (AWS DMS) version 3.5.1.

New feature or enhancement Description Support for PostgreSQL 15.x AWS DMS version 3.5.1 supports PostgreSQL version 15.x. For more information, see Using PostgreSQL as a source and Using PostgreSQL as a target. Support for Amazon DocumentDB Elastic Clusters with sharded collections AWS DMS version 3.5.1 supports Amazon DocumentDB Elastic Clusters with sharded collections. For more information, see Using Amazon DocumentDB as a target for AWS Database Migration Service. Support for using Amazon Amazon Redshift Serverless as a target endpoint Support for using Amazon Amazon Redshift Serverless as a target endpoint. For more information, see Using an Amazon Redshift database as a target for AWS Database Migration Service. Enhanced PostgreSQL target endpoint settings for providing Babelfish support Enhanced PostgreSQL target endpoint settings for providing Babelfish support. For more information, see Using a PostgreSQL database as a target for AWS Database Migration Service. AWS DMS 3.5.1 improves the methodology of handling open transactions when starting a CDC-Only task from the Start Position for an Oracle source AWS DMS 3.5.1 improves the methodology of handling open transactions when starting a CDC-Only task from the Start Position for an Oracle source. For more information, see OpenTransactionWindow in the Endpoint settings when using Oracle as a source for AWS DMS section. Support for using Amazon Timestream as a target endpoint Support for using Amazon Timestream as a target endpoint. For more information, see Using Amazon Timestream as a target for AWS Database Migration Service.

AWS DMS version 3.5.1 includes the following resolved issues:

• Large numeric value handling change
  The representation of large numeric values on streaming targets has been updated. Review the 'Handling of Large Numeric Values in Streaming Targets' documentation for details on potential impacts.
• Fixed an issue for Oracle source where CDC-only tasks had continuously growing inactive sessions, resulting in the following exception: ORA-00020: maximum number of processes exceeded on the source database.
• Fixed an issue for DocumentDB as a target where UPDATE statements were not properly replicated in some scenarios.
• Improved error handling for the data validation feature to properly fail the task when data validation is disabled for validation-only tasks.
• Fixed an issue for Amazon Redshift target where the DMS task would not retry applying changes on the target when the target has ParallelApplyThreads set greater than zero after connection termination, which would result in data loss.
• Fixed an issue for MySQL to MySQL replication of mediumtext data types with full-LOB mode.
• Fixed an issue with redundant calls to retrieve other tables' metadata, which resulted in degraded performance and latency after a table's metadata had been altered.
• Fixed an issue where non-standard UPDATE operations caused tasks to terminate unexpectedly without generating error messages.
• Fixed an issue in Batch Optimized Apply mode where Large Object (LOB) lookup would fail to find a record when a DELETE operation and an INSERT operation were combined into a single UPDATE operation.
• Addressed an issue where tasks involving numerous transformations would crash during startup.
• Fixed a failure to capture special format Data Definition Language (DDL) changes during the Change Data Capture (CDC) phase.
• Fixed an issue for skipping cached changes in case of an unexpected interruption while applying those changes.
• Fixed an issue for MongoDB source where AWS DMS task would fail to resume after AWS DMS upgrade when AWS DMS swap file was present.
• Fixed an issue for MySQL source where the JSON data type was not being hadled propely with Batch Apply enabled.
• Fixed an issue for SQL Server to PostgreSQL migration where data validaiton would report false positives in certain situations.
• Fixed an issue for Amazon Redshift target where the EmptyAsNull ECA would not work correctly.
• Enhanced logging for SQL Server source to unclude the storage unit value.
• Fixed an issue where for targets using CSV files to load data AWS DMS task was showing a memory leak.
• Fixed an issue for Amazon S3 target where CdcMaxBatchInterval and CdcMinFileSize were not respected when cdcInsertAndUpdateOnly setting was enabled.
• Enhanced logging for Kafka target
• Enhanced logging for Oracle source with binary reader to properly indicate tables being skipped due to missing primary keys.
• Enhanced logging for SQL Server source in AlwaysOn configuration to properly indicate missing permissions.
• Enhanced logging for migrations with disabled DDL replication to indicate unexpected target table structure after its modified outside of AWS DMS.
• Fixed an issue for Db2 target where the task would fail when the AWS DMS status table is enabled.
• Fixed an issue for MongoDB / Amazon DocumentDB endpoints where the credentials could not be retrieved from Secret Manager which resulted in an error.
• Fixed an issue for MongoDB / Amazon DocumentDB where the task would fail with ParallelApply enabled while replicating certain sequence of events.
• Enhanced logging for Amazon Redshift target to include more detailed information in default logging level.
• Fixed an issue for Amazon S3 target where AWS DMS task would crash after receiving alter table DDL when GlueCatalogGeneration is enabled.
• Fixed an issue for data validation feature where validation would fail on NUL (0x00) characters.
• Fixed an issue for babelfish endpoint where tables names with mixed case would be suspended.
• Fixed an issue for Amazon Redshift target where data loss would occur when ParallelLoadThreads was >0 under certain conditions.
• Enhanced the data validation feature for Amazon Redshift target to support HandleCollationDiff setting.
• Fixed an issue for Amazon S3 target data validation where validation would fail when there were no other columns than the PK in the table.
• Fixed an issue for data validation feature where the CloudWatch metrics would be missing for validation which took a short amount of time to complete.
• Fixed an issue for data validation feature where the re-validation option was unavailable in certain situations.
• Fixed an issue where the maximum number of events per transaction was limited to 201,326,592 under certain conditions.
• Fixed an issue for MySQL to Amazon S3 migration where first DML executed after "add column" DDL would be missed resulting in data loss.
• Fixed a memory leak issue for batch apply feature which would occur under certain conditions.
• Fixed an issue where AWS DMS task start woudl take a very long time and never complete.
• Fixed an issue for PostgreSQL source where data loss would occur due to unknown events in the replication slot.
• Fixed an issue for S3 as a target, where the resultant CSV file header is off by one column when AddColumnName is true and TimestampColumnName is "".
• Fixed an issue for S3 as a source, where a DMS task in full load was only releasing the used memory after the entire table was loaded to the target database.
• Fixed an issue for S3 as a target, where a table reload operation missed generating CDC files.

Original source Report a problem

New features in AWS DMS 3.5.4

Introduced data masking, allowing transforming sensitive data with options for digit randomization, masking, or hashing at the column level.

Introduced enhanced data validation performance, enabling faster processing for large datasets during full load and CDC migration tasks across select migration paths.

Fixed an issues for PostgreSQL source where degraded migration performance was observed while using filtering. Introduced disableUnicodeSourceFilter ECA to control this behavior.

Introduce support for all transformation rules for the enhanced throughput feature.

Enhanced logging to inform customers about missing target table columns at default logging level. This improvement changes the logging level from VERBOSE to WARNING, making column discrepancy notifications more visible without requiring detailed debug settings.

Added support for migrating SQL Server Binary(16) data types to PostgreSQL UUID format. This enables seamless conversion of binary GUID data to native UUID types, improving data type compatibility between SQL Server and PostgreSQL endpoints.

AWS DMS version 3.5.4 includes the following resolved issues:

Fixed an issue for PostgreSQL source where certain events would not be replicated while using the test_decoding plugin.

Fixed an issue for MySQL to Redshift migrations where a timestamp column would not be defined correctly on the target.

Fixed an issue for Oracle source with binary reader where DMS task would crash after applying Oracle July 2024 PSU.

Fixed an issue for MySQL endpoint where credentials would become corrupted when using secrets manager.

Fixed an issue for Amazon DocumentDB/MongoDB endpoints where some records would be sent to the target twice causing duplicate key exception and AWS DMS task failure.

Fixed an issue for RDS for SQL Server to noSQL migrations where the document structure was incorrect due to incorrect PK handling.

Fixed an issue for Oracle source where data validation would report false positives for null or empty LOBs.

Fixed an issue for PG - Babelfish target where AWS DMS task would fail while replicating tables with PK defined as uniqueidentifier.

Fixed an issue for PostgreSQL source where AWS DMS MAZ failover would cause a fatal AWS DMS task failure.

Fixed an issue where LOB data was not replicated correctly when the column order differed between the source and target.

Fixed an issue for internal DMS database where the AWS DMS task would fail due to internal AWS DMS database concurency issues.

Fixed an issues for internal AWS DMS database where the AWS DMS task would fail due to the lack of presence of certain internal database objects.

Fixed an issue for Oracle source where data validation would return false positives for certain rare types of replicated events.

Fixed an issue for the data validation feature where ceratin unicode data types were not compared properly which resulted in false positives.

Fixed an issue for parquet target where zero-timestamp was replicated as null.

Fixed an issue with MySQL source and target migrating Lob columns. Now DMS uses the column id from the target table instead of source table when deciding to which column we need to write the LOB Data

Fixed an issue with Babelfish target where we have restricted the fractional seconds precision for Datetime and Time types to solve rounding errors.

Fixed an issue with MySQL 5.5 source, we have added retry mechanism to prevent task failure when DMS would fail to read binary log events during on going replication (CDC).

Fixed an issue with PostgreSQL source where certain onging replication (CDC) events failed to be parsed correctly when using the test_decoding plugin for Postgres.

Fixed an issue with DocumentDB target with parallel apply setting which was preventing the use of multiple threads while using this feature.

Fixed an issue with Oracle HCC compression DIRECT INSERT with parallel DML hint causing missing and duplicate data.

Fixed an issue with Oracle source, DMS task with binary reader were failing due to Oracle July 2024 CPU

Fixed an issue with Babelfish target where DMS task were failing while replicating tables with Primary Key defined as UUID.

Fixed an issue with TaskRecoveryTableEnabled enabled, where DMS attempts to update the target system table awsdms_txn_state after the target connection is terminated.

Fixed an issue with PostgreSQL source where some transactions would be replicated twice when the TaskrecoveryTableEnabled setting was enabled.

Fixed an issue with S3 source to S3 target where DMS task was not replicating data during full load and on going replication.

Fixed an issue for S3 source where DMS task was seg faulting during on going replication for DMS vesion 3.5.3

Fixed an issue with DB2 soruce with CcsidMapping, CCSID mappin ECA is now properly applied to task when code page is 0 and data is migrated properly

Fixed an issue where DMS migration from Aurora PostgreSQL to Redshift Serverless was seeing issue with Boolean value.

Data validation operation can now accurately processes unbound characters and TEXT data types, ensuring correct validation results.

PostgreSQL source replication maintains connectivity during Multi-AZ failover events, preventing task failures.

Data validation now correctly compares datetime values when using Babelfish as a target, improving cross-platform compatibility.

MySQL source replication now correctly handles mid-table duplicate column additions, preventing task interruptions.

MySQL source replication maintains column sequence integrity when multiple columns are added during CDC operations.

DynamoDB target replication now correctly processes LOB data during CDC, ensuring complete data transfer.

PostgreSQL source data validation now correctly interprets boolean data type mappings, producing accurate comparison results.

Oracle source replication preserves trailing spaces in VARCHAR2(4000) columns when extended data type support is enabled.

SQL Server source replication maintains connectivity during DDL operations on secondary replicas, preventing task interruptions.

AWS Secrets Manager connection strings now support special characters while preserving security protocols.

MongoDB and Amazon DocumentDB replication prevents record duplication that previously triggered key constraint errors.

Oracle source replication accurately processes timestamp values across various session time zone configurations.

Oracle source replication now handles data type conversions more robustly, preventing ORA-01460 errors and associated task failures.

AWS DMS ignores DDL statements for tables not configured in table mappings, preventing unnecessary processing of source database schema changes for unmapped tables.

When migrating data from source SQL Server view to target endpoints, the column length is correctly preserved and no columns are truncated.

AWS DMS now accurately reports validated record metrics when using enhanced data validation.

When resuming tasks after an AWS DMS version upgrade, DMS tasks do not fail to resume due to incompatible internal file formats.

Fixed an issue where AWS DMS now correctly identifies the primary replica in Microsoft SQL Server Always On Availability Groups, resolving previous case sensitivity detection errors.

Fixed an issue where AWS DMS tasks failed when migrating unbound numeric data types to PostgreSQL target endpoints.

AWS DMS now correctly validates CHAR and VARCHAR data during migrations, eliminating false positive reports in validation tasks.

AWS DMS prevents data corruption in Large Object (LOB) data during parellel batch operations when using Amazon Redshift as a target endpoint.

Validation failures are prevented and queries are successfully executed on Amazon S3 target when source column filters are applied to VARCHAR, CHAR, DATE, and DATETIME data types.

When using Amazon S3 targets, consistent data validation states are maintained throughout Change Data Capture (CDC) operations.

Fixed an issue where data validation failed due to string formatting errors during record comparison operations.

Resolved an issue where empty VARCHAR values from IBM DB2 LUW source databases were incorrectly captured as NULL values instead of empty strings during Change Data Capture (CDC) operations.

Fixed memory leak in Oracle source endpoints during LOB lookups when full LOB mode is enabled. This prevents continuous memory growth and out-of-memory failures during replication tasks with LOB data.

Redshift target endpoints now correctly map new boolean columns added during CDC when MapBooleanAsBoolean is enabled. This maintains data type consistency with PostgreSQL sources instead of creating varchar columns.

PostgreSQL source endpoints now correctly migrate UUID array data types when inline LOB mode is enabled.

Data validation now correctly reports mismatch details for nullable columns.

Data validation now correctly formats date filter conditions for Oracle target endpoints when using SQL Server sources.

Added warning log message when AWS DMS automatically creates target tables in DO_NOTHING mode. This improves visibility and helps users identify when tables are created despite selecting a mode that implies no automatic table creation.

Fixed SQL Server BIT to PostgreSQL SMALLINT conversion during CDC operations.

Enhanced S3 bucket ownership validation to improve security when using S3 as a migration target or source.

Redshift endpoints now properly reconnect after connection termination and correctly validate credentials during initialization. This prevents "Server name must be supplied" errors and task failures during recovery.

Fixed memory leak in data validation. This prevents out-of-memory failures during long-running validation tasks.

Fixed false positive MISSING_TARGET errors when validating SQL Server to PostgreSQL migrations with CHAR/NCHAR primary keys containing trailing spaces.

Fixed race condition when multiple tasks share S3 endpoints by isolating Athena databases per task.

Original source Report a problem

New features in AWS DMS 3.6.1

AWS DMS automatically fixes data inconsistencies identified through data validation between your source and target databases. The data resync feature is supported for Oracle and SQL Server as a source and Postgre SQL as a target databases.

Introduced support connecting to Amazon RDS and Aurora MariaDB, MySQL, and PostgreSQL endpoints through AWS Identity and Access Management (IAM) database authentication. With this enhancement you can use IAM to centrally manage access to your database resources, instead of managing access individually on each database endpoint.

Introduced support for PostgreSQL version 17. For more information, see: Using a PostgreSQL database as an AWS DMS source Using a PostgreSQL database as a target for AWS Database Migration Service

AWS DMS supports using PostgreSQL read replicas as source endpoints for Change Data Capture (CDC) replication, available with PostgreSQL version 16.x and later, starting with AWS DMS version 3.6.1. This feature allows you to leverage read replicas for CDC tasks. For more information, see Using a PostgreSQL database as an AWS DMS source.

Enhanced logging to inform customers about missing target table columns at default logging level. This improvement changes the logging level from VERBOSE to WARNING, making column discrepancy notifications more visible without requiring detailed debug settings.

Added support for migrating SQL Server Binary(16) data types to PostgreSQL UUID format. This enables seamless conversion of binary GUID data to native UUID types, improving data type compatibility between SQL Server and PostgreSQL endpoints.

AWS DMS version 3.6.1 includes the following resolved issues:

• Table selection rule filters now work correctly with data masking transformation rules. The fix ensures proper filtering when data masking transformations are applied.
• PostgreSQL target endpoints can handle unbound numeric data types through optimized memory allocation. This prevents task failures when migrating tables with large numeric values.
• Redshift target endpoints now maintain LOB data integrity during high-volume parallel migrations. The improved LOB handling prevents data corruption when using parallel apply threads.
• Fixed an issue with expressions in transformation rules that were incorrectly applied.
• Replication tasks now resume successfully when source tables are dropped and recreated during task stoppage. Improved metadata handling and status synchronization for recreated source tables.
• Improved primary replica detection for SQL Server sources in AlwaysOn configurations by handling case sensitivity variations.
• Resolved migration issues with TIME values containing fractions and hours exceeding 24 when loadUsingCSV is disabled.
• Data validation now handles NUL (0x00) characters correctly without failing. Improved validation processing for data containing NUL characters.
• Resolved validation comparison issues with UNIQUEIDENTIFIER columns for SQL Server sources.
• AWS DMS tasks now complete validation successfully when tables are in pre-load phase during CDC replication.
• Oracle source error logs now display the correct SQL statement during timezone query execution. Improved logging accuracy for better troubleshooting capabilities.
• SQL Server driver support is updated from ODBC 17 to ODBC 18.
• Resolved an issue where empty VARCHAR values from IBM DB2 LUW source databases were incorrectly captured as NULL values instead of empty strings during Change Data Capture (CDC) operations.
• Fixed memory leak in Oracle source endpoints during LOB lookups when full LOB mode is enabled. This prevents continuous memory growth and out-of-memory failures during replication tasks with LOB data.
• Redshift target endpoints now correctly map new boolean columns added during CDC when MapBooleanAsBoolean is enabled. This maintains data type consistency with PostgreSQL sources instead of creating varchar columns.
• PostgreSQL source endpoints now correctly migrate UUID array data types when inline LOB mode is enabled.
• Data validation now correctly reports mismatch details for nullable columns.
• Data validation now correctly formats date filter conditions for Oracle target endpoints when using SQL Server sources.
• Added warning log message when AWS DMS automatically creates target tables in DO_NOTHING mode. This improves visibility and helps users identify when tables are created despite selecting a mode that implies no automatic table creation.
• Fixed SQL Server BIT to PostgreSQL SMALLINT conversion during CDC operations.
• Enhanced S3 bucket ownership validation to improve security when using S3 as a migration target or source.
• Redshift endpoints now properly reconnect after connection termination and correctly validate credentials during initialization. This prevents "Server name must be supplied" errors and task failures during recovery.
• Fixed memory leak in data validation. This prevents out-of-memory failures during long-running validation tasks.
• Fixed false positive MISSING_TARGET errors when validating SQL Server to PostgreSQL migrations with CHAR/NCHAR primary keys containing trailing spaces.
• Fixed race condition when multiple tasks share S3 endpoints by isolating Athena databases per task.

Original source Report a problem

New features in AWS DMS 3.6.0

Introduced two new metadata variables for transformation rules: $AR_M_MODIFIED_SCHEMA and $AR_M_MODIFIED_TABLE_NAME. These variables allow you to create more dynamic and flexible transformations that can adapt to schema or table name changes during your migration process.

Introduced support for replication of Large Object (LOB) columns when using the "No record found for applying an UPDATE: Insert the missing target record" error-handling option. With this enhancement, you can now accurately replicate LOB columns, ensuring complete and accurate data replication.

AWS DMS version 3.6.0 includes the following resolved issues:

Resolved excessive memory consumption when using PostgreSQL as a source. Fixed an issue where the Write-Ahead Log (WAL) slot would continuously grow, leading to degraded performance. This update enhances the stability and efficiency of PostgreSQL migrations, especially for large databases or long-running tasks.

Addressed the issue where wide Character Large Object (CLOB) and Character (CHAR) values containing non-ASCII characters (e.g., special symbols or international characters) were not replicated correctly. This fix ensures accurate replication of large text data, reducing errors, and maintaining data consistency.

Resolved an issue where the Microsoft SQL Server source endpoint would incorrectly report very high latency. This fix provides more accurate performance metrics, allowing you to better monitor and optimize your SQL Server migration tasks.

Fixed an issue with redundant calls to retrieve other tables' metadata, which resulted in degraded performance and latency after a table's metadata had been altered. This optimization improves overall task performance, especially for databases with frequent schema changes.

Fixed an issue where non-standard UPDATE operations caused tasks to terminate unexpectedly without generating error messages. This fix ensures that such operations are processed correctly, avoiding task failures during complex update scenarios.

Fixed an issue in Batch Optimized Apply mode where Large Object (LOB) lookup would fail to find a record when a DELETE operation and an INSERT operation were combined into a single UPDATE operation. This fix improves data consistency and integrity during migrations involving LOB data and complex Data Manipulation Language (DML) operations.

Addressed an issue where tasks involving numerous transformations would crash during startup. This fix ensures stability and reliability for tasks with complex transformation logic, enabling you to confidently execute data processing tasks.

Fixed a failure to capture special format Data Definition Language (DDL) changes during the Change Data Capture (CDC) phase. This fix ensures that all schema changes are properly replicated, maintaining schema consistency between source and target databases throughout the migration process.

Enhanced S3 bucket ownership validation to improve security when using S3 as a migration target or source.

Fixed memory leak in data validation. This prevents out-of-memory failures during long-running validation tasks.

Original source Report a problem

New features in AWS DMS 3.5.4

Introduced data masking, allowing transforming sensitive data with options for digit randomization, masking, or hashing at the column level.

Introduced enhanced data validation performance, enabling faster processing for large datasets during full load and CDC migration tasks across select migration paths.

Fixed an issues for PostgreSQL source where degraded migration performance was observed while using filtering. Introduced disableUnicodeSourceFilter ECA to control this behavior.

Introduce support for all transformation rules for the enhanced throughput feature.

AWS DMS version 3.5.4 includes the following resolved issues:

• Fixed an issue for PostgreSQL source where certain events would not be replicated while using the test_decoding plugin.
• Fixed an issue for MySQL to Redshift migrations where a timestamp column would not be defined correctly on the target.
• Fixed an issue for Oracle source with binary reader where DMS task would crash after applying Oracle July 2024 PSU.
• Fixed an issue for MySQL endpoint where credentials would become corrupted when using secrets manager.
• Fixed an issue for Amazon DocumentDB/MongoDB endpoints where some records would be sent to the target twice causing duplicate key exception and AWS DMS task failure.
• Fixed an issue for RDS for SQL Server to noSQL migrations where the document structure was incorrect due to incorrect PK handling.
• Fixed an issue for Oracle source where data validation would report false positives for null or empty LOBs.
• Fixed an issue for PG - Babelfish target where AWS DMS task would fail while replicating tables with PK defined as uniqueidentifier.
• Fixed an issue for PostgreSQL source where AWS DMS MAZ failover would cause a fatal AWS DMS task failure.
• Fixed an issue where LOB data was not replicated correctly when the column order differed between the source and target.
• Fixed an issue for internal DMS database where the AWS DMS task would fail due to internal AWS DMS database concurency issues.
• Fixed an issues for internal AWS DMS database where the AWS DMS task would fail due to the lack of presence of certain internal database objects.
• Fixed an issue for Oracle source where data validation would return false positives for certain rare types of replicated events.
• Fixed an issue for the data validation feature where ceratin unicode data types were not compared properly which resulted in false positives.
• Fixed an issue for parquet target where zero-timestamp was replicated as null.
• Fixed an issue for Babelfish target where the GeoSpatial data type was not supported.
• Fixed an issue for Amazon S3 target where new column additions were not handeled properly when the before image setting was enabled.
• Enhanced S3 directory cleanup process prevents task interruptions during migrations.
• Data validation operation now optimizes transitions between validation phases, reducing unnecessary delays.
• Data validation now accurately processes unbound characters and TEXT data types, ensuring correct validation results.
• PostgreSQL source replication maintains connectivity during Multi-AZ failover events, preventing task failures.
• Data validation now correctly compares datetime values when using Babelfish as a target.
• MySQL source replication now correctly handles mid-table duplicate column additions, preventing task interruptions.
• MySQL source replication maintains column sequence integrity when multiple columns are added during CDC operations.
• DynamoDB target replication now correctly processes LOB data during CDC, ensuring complete data transfer.
• PostgreSQL source data validation now correctly interprets boolean data type mappings, producing accurate comparison results.
• Oracle source replication preserves trailing spaces in VARCHAR2(4000) columns when extended data type support is enabled.
• SQL Server source replication maintains connectivity during DDL operations on secondary replicas, preventing task interruptions.
• AWS Secrets Manager connection strings now support special characters while preserving security protocols.
• MongoDB and Amazon DocumentDB replication prevents record duplication that previously triggered key constraint errors.
• Oracle source replication accurately processes timestamp values across various session time zone configurations.
• Oracle source replication now handles data type conversions more robustly, preventing ORA-01460 errors and associated task failures.
• AWS DMS ignores DDL statements for tables not configured in table mappings, preventing unnecessary processing of source database schema changes for unmapped tables.
• When migrating data from source SQL Server view to target endpoints, the column length is correctly preserved and no columns are truncated.
• AWS DMS now accurately reports validated record metrics when using enhanced data validation.
• When resuming tasks after an AWS DMS version upgrade, DMS tasks do not fail to resume due to incompatible internal file formats.
• Fixed an issue where AWS DMS now correctly identifies the primary replica in Microsoft SQL Server Always On Availability Groups, resolving previous case sensitivity detection errors.
• Fixed an issue where AWS DMS tasks failed when migrating unbound numeric data types to PostgreSQL target endpoints.
• AWS DMS now correctly validates CHAR and VARCHAR data during migrations, eliminating false positive reports in validation tasks.
• AWS DMS prevents data corruption in Large Object (LOB) data during parellel batch operations when using Amazon Redshift as a target endpoint.
• Validation failures are prevented and queries are successfully executed on Amazon S3 target when source column filters are applied to VARCHAR, CHAR, DATE, and DATETIME data types.
• When using Amazon S3 targets, consistent data validation states are maintained throughout Change Data Capture (CDC) operations.
• Fixed an issue where data validation failed due to string formatting errors during record comparison operations.

Original source Report a problem