import ChangeLog from ‘../changelog/connector-cdc-sqlserver.md’;

SQL Server CDC

Sql Server CDC source connector

Support SQL Server Version

• server:2019 (Or later version for information only)

Support Those Engines

SeaTunnel Zeta
Flink

Key Features

• [ ] batch
• [x] stream
• [x] exactly-once
• [ ] column projection
• [x] parallelism
• [x] support user-defined split

Description

The Sql Server CDC connector allows for reading snapshot data and incremental data from SqlServer database. This document describes how to setup the Sql Server CDC connector to run SQL queries against SqlServer databases.

:::tip

When discovering table columns via JDBC metadata, SeaTunnel filters metadata rows by the exact schema/table identifier to avoid mixing columns from other tables (some drivers treat schemaPattern/tableNamePattern as SQL LIKE patterns). For case-sensitive databases, make sure the configured identifier case matches the database.

:::

Supported DataSource Info

Datasource	Supported versions	Driver	Url	Maven
SqlServer	server:2019 (Or later version for information only)	com.microsoft.sqlserver.jdbc.SQLServerDriver	jdbc:sqlserver://localhost:1433;databaseName=column_type_test	https://mvnrepository.com/artifact/com.microsoft.sqlserver/mssql-jdbc

Using Dependency

Install Jdbc Driver

For Spark/Flink Engine

1. You need to ensure that the jdbc driver jar package has been placed in directory ${SEATUNNEL_HOME}/plugins/.

For SeaTunnel Zeta Engine

1. You need to ensure that the jdbc driver jar package has been placed in directory ${SEATUNNEL_HOME}/lib/.

Data Type Mapping

SQLserver Data Type	SeaTunnel Data Type
CHAR VARCHAR NCHAR NVARCHAR TEXT NTEXT XML	STRING
BINARY VARBINARY IMAGE	BYTES
INTEGER INT	INT
SMALLINT TINYINT	SMALLINT
BIGINT	BIGINT
FLOAT(1~24) REAL	FLOAT
DOUBLE FLOAT(>24)	DOUBLE
NUMERIC(p,s) DECIMAL(p,s) MONEY SMALLMONEY	DECIMAL(p, s)
TIMESTAMP	BYTES
DATE	DATE
TIME(s)	TIME(s)
DATETIME(s) DATETIME2(s) DATETIMEOFFSET(s) SMALLDATETIME	TIMESTAMP(s)
BOOLEAN BIT	BOOLEAN

Source Options

Name	Type	Required	Default	Description
username	String	Yes	-	Name of the database to use when connecting to the database server.
password	String	Yes	-	Password to use when connecting to the database server.
database-names	List	Yes	-	Database name of the database to monitor.
table-names	List	Yes	-	Table name is a combination of schema name and table name (databaseName.schemaName.tableName).
table-names-config	List	No	-	Table config list. for example: [{“table”: “db1.schema1.table1”,“primaryKeys”: [“key1”],“snapshotSplitColumn”: “key2”}]. The snapshotSplitColumn option must be configured with a unique key. If a non-unique column is provided, the configuration is ignored and SeaTunnel automatically selects an appropriate split column internally.
url	String	Yes	-	URL has to be with database, like “jdbc:sqlserver://localhost:1433;databaseName=test”.
startup.mode	Enum	No	INITIAL	Optional startup mode for SqlServer CDC consumer, valid enumerations are “initial”, “earliest”, “latest”, “timestamp” and “specific”.
startup.timestamp	Long	No	-	Start from the specified epoch timestamp (in milliseconds). This timestamp is converted with server-time-zone when startup.mode = timestamp. Note, This option is required when the “startup.mode” option used 'timestamp'.
startup.specific-offset.file	String	No	-	Start from the specified binlog file name. Note, This option is required when the “startup.mode” option used 'specific'.
startup.specific-offset.pos	Long	No	-	Start from the specified binlog file position. Note, This option is required when the “startup.mode” option used 'specific'.
stop.mode	Enum	No	NEVER	Optional stop mode for SqlServer CDC consumer, valid enumerations are “never”.
stop.timestamp	Long	No	-	Stop from the specified epoch timestamp (in milliseconds). Note, This option is required when the “stop.mode” option used 'timestamp'.
stop.specific-offset.file	String	No	-	Stop from the specified binlog file name. Note, This option is required when the “stop.mode” option used 'specific'.
stop.specific-offset.pos	Long	No	-	Stop from the specified binlog file position. Note, This option is required when the “stop.mode” option used 'specific'.
incremental.parallelism	Integer	No	1	The number of parallel readers in the incremental phase.
snapshot.split.size	Integer	No	8096	The split size (number of rows) of table snapshot, captured tables are split into multiple splits when read the snapshotof table.
snapshot.fetch.size	Integer	No	1024	The maximum fetch size for per poll when read table snapshot.
server-time-zone	String	No	UTC	The session time zone in database server. This value is also used when converting startup.timestamp to LSN. Set it explicitly when database time zone and JVM time zone are different.
connect.timeout	Duration	No	30s	The maximum time that the connector should wait after trying to connect to the database server before timing out.
connect.max-retries	Integer	No	3	The max retry times that the connector should retry to build database server connection.
connection.pool.size	Integer	No	20	The connection pool size.
chunk-key.even-distribution.factor.upper-bound	Double	No	100	The upper bound of the chunk key distribution factor. This factor is used to determine whether the table data is evenly distributed. If the distribution factor is calculated to be less than or equal to this upper bound (i.e., (MAX(id) - MIN(id) + 1) / row count), the table chunks would be optimized for even distribution. Otherwise, if the distribution factor is greater, the table will be considered as unevenly distributed and the sampling-based sharding strategy will be used if the estimated shard count exceeds the value specified by sample-sharding.threshold. The default value is 100.0.
chunk-key.even-distribution.factor.lower-bound	Double	No	0.05	The lower bound of the chunk key distribution factor. This factor is used to determine whether the table data is evenly distributed. If the distribution factor is calculated to be greater than or equal to this lower bound (i.e., (MAX(id) - MIN(id) + 1) / row count), the table chunks would be optimized for even distribution. Otherwise, if the distribution factor is less, the table will be considered as unevenly distributed and the sampling-based sharding strategy will be used if the estimated shard count exceeds the value specified by sample-sharding.threshold. The default value is 0.05.
sample-sharding.threshold	int	No	1000	This configuration specifies the threshold of estimated shard count to trigger the sample sharding strategy. When the distribution factor is outside the bounds specified by chunk-key.even-distribution.factor.upper-bound and chunk-key.even-distribution.factor.lower-bound, and the estimated shard count (calculated as approximate row count / chunk size) exceeds this threshold, the sample sharding strategy will be used. This can help to handle large datasets more efficiently. The default value is 1000 shards.
inverse-sampling.rate	int	No	1000	The inverse of the sampling rate used in the sample sharding strategy. For example, if this value is set to 1000, it means a 1/1000 sampling rate is applied during the sampling process. This option provides flexibility in controlling the granularity of the sampling, thus affecting the final number of shards. It's especially useful when dealing with very large datasets where a lower sampling rate is preferred. The default value is 1000.
split.allow-sampling	Boolean	No	true	Whether to allow sampling-based sharding strategy. When set to false, the system will fall back to unevenly-sized chunk splitting (iterative query approach) regardless of the shard count. The default value is true.
exactly_once	Boolean	No	false	Enable exactly once semantic.
debezium.*	config	No	-	Pass-through Debezium's properties to Debezium Embedded Engine which is used to capture data changes from SqlServer server. See more about the Debezium's SqlServer Connector properties
format	Enum	No	DEFAULT	Optional output format for SqlServer CDC, valid enumerations are “DEFAULT”、“COMPATIBLE_DEBEZIUM_JSON”.
schema-changes.enabled	Boolean	No	false	Schema evolution is disabled by default. Now we only support add column、drop column、rename column and modify column.
schema-changes.include	List	No	-	Only the listed schema change event types are sent downstream (when schema-changes.enabled = true). Empty means all are eligible. See Schema change event filtering.
schema-changes.exclude	List	No	-	Schema change event types listed here are NOT sent downstream. Applied after schema-changes.include; exclude wins on conflict. See Schema change event filtering.
common-options		no	-	Source plugin common parameters, please refer to Source Common Options for details.

Enable Sql Server CDC

1. Check whether the CDC Agent is enabled

EXEC xp_servicecontrol N‘querystate’, N‘SQLServerAGENT’;
If the result is running, prove that it is enabled. Otherwise, you need to manually enable it

2. Enable the CDC Agent

/opt/mssql/bin/mssql-conf setup

3. The result is as follows

1. Evaluation (free, no production use rights, 180-day limit)
2. Developer (free, no production use rights)
3. Express (free)
4. Web (PAID)
5. Standard (PAID)
6. Enterprise (PAID)
7. Enterprise Core (PAID)
8. I bought a license through a retail sales channel and have a product key to enter.

4. Set the CDC at the library level Set the library level below to enable CDC. At this level, all tables under the libraries of the enabled CDC automatically enable CDC

USE TestDB; -- Replace with the actual database name
EXEC sys.sp_cdc_enable_db;
SELECT name, is_tracked_by_cdc FROM sys.tables WHERE name = ‘table’; -- table Replace with the name of the table you want to check

Task Example

initiali read Simple

This is a stream mode cdc initializes read table data will be read incrementally after successful read The following sql DDL is for reference only

env {
  # You can set engine configuration here
  parallelism = 1
  job.mode = "STREAMING"
  checkpoint.interval = 5000
}

source {
  # This is a example source plugin **only for test and demonstrate the feature source plugin**
  SqlServer-CDC {
    plugin_output = "customers"
    username = "sa"
    password = "Y.sa123456"
    startup.mode="initial"
    database-names = ["column_type_test"]
    table-names = ["column_type_test.dbo.full_types"]
    url = "jdbc:sqlserver://localhost:1433;databaseName=column_type_test"
  }
}

transform {
}

sink {
  console {
    plugin_input = "customers"
  }
}

increment read Simple

This is an incremental read that reads the changed data for printing

env {
  # You can set engine configuration here
  parallelism = 1
  job.mode = "STREAMING"
  checkpoint.interval = 5000
}

source {
  # This is a example source plugin **only for test and demonstrate the feature source plugin**
  SqlServer-CDC {
   # Set up accurate one read
    exactly_once=true 
    plugin_output = "customers"
    username = "sa"
    password = "Y.sa123456"
    startup.mode="latest"
    database-names = ["column_type_test"]
    table-names = ["column_type_test.dbo.full_types"]
    url = "jdbc:sqlserver://localhost:1433;databaseName=column_type_test"
  }
}

transform {
}

sink {
  console {
    plugin_input = "customers"
  }
}

Support custom primary key for table

env {
  parallelism = 1
  job.mode = "STREAMING"
  checkpoint.interval = 5000
}

source {
  SqlServer-CDC {
    url = "jdbc:sqlserver://localhost:1433;databaseName=column_type_test"
    username = "sa"
    password = "Y.sa123456"
    database-names = ["column_type_test"]
    
    table-names = ["column_type_test.dbo.simple_types", "column_type_test.dbo.full_types"]
    table-names-config = [
      {
        table = "column_type_test.dbo.full_types"
        primaryKeys = ["id"]
      }
    ]
  }
}

sink {
  console {
  }
}

Schema change event filtering

When schema-changes.enabled = true, you can further control which schema change event types are propagated downstream using schema-changes.include / schema-changes.exclude.

Use these SeaTunnel-owned canonical names:

Canonical name	Operation
add.column	add a column
drop.column	drop a column
modify.column	change a column's type/attributes, name unchanged
change.column	rename a column, optionally re-type
update.columns	group alias for all four column-level changes above

Precedence is deterministic:

1. if schema-changes.include is set, only included event types are eligible;
2. schema-changes.exclude is then applied;
3. exclude wins when a type appears in both lists.

source {
  SqlServer-CDC {
    # ...
    schema-changes.enabled = true
    schema-changes.include = ["add.column", "drop.column"]
    schema-changes.exclude = ["change.column"]
  }
}

**Data handling when drop.column is excluded. For a retained NOT NULL column the NULL write is rejected by the sink, so excluding drop.column for a NOT NULL column that the source has stopped supplying will fail at the sink.

Changelog