SQL Server and Azure SQL connector

Use jdbc:sqlserver: for SQL Server and Azure SQL. SQL Server supports source and sink roles and all three replication modes. The bandwidth-throttling option is not applied on the sink path.

Connection string

Format	Example
Host and port	jdbc:sqlserver://${MSSQL_HOST}:${MSSQL_PORT};databaseName=${MSSQL_DATABASE}
Azure SQL	jdbc:sqlserver://${AZURE_SQL_SERVER}.database.windows.net:1433;databaseName=${MSSQL_DATABASE}

Credentials are set with source.user / source.password (or the sink.* equivalents), or with source.auth.* / sink.auth.* for Microsoft Entra modes. The driver class is com.microsoft.sqlserver.jdbc.SQLServerDriver.

Example properties

mode=complete-atomic
jobs=4
fetch.size=10000

source.connect=jdbc:sqlserver://${MSSQL_HOST}:${MSSQL_PORT};databaseName=${MSSQL_DATABASE}
source.user=${MSSQL_USER}
source.password=${MSSQL_PASSWORD}
source.table=${SOURCE_TABLE}

sink.connect=jdbc:sqlserver://${MSSQL_HOST}:${MSSQL_PORT};databaseName=${MSSQL_DATABASE}
sink.user=${MSSQL_USER}
sink.password=${MSSQL_PASSWORD}
sink.table=${SINK_TABLE}

Bulk copy and data types

The sink uses the Microsoft SQLServerBulkCopy API with table locking. The configured fetch.size is the bulk-copy batch size, so it controls how many rows are sent per batch. SQL Server reports XML with a vendor-specific JDBC type, which ReplicaDB maps to SQLXML and writes as XML. Test XML and other vendor types across the exact source and sink pair.

Bulk copy has no mid-transfer cancellation hook. A cancelled run may leave the sink partially written, so use complete-atomic or incremental with staging where a partial write matters.

Microsoft Entra authentication

Use the source.auth.* or sink.auth.* options rather than embedding secrets in a JDBC URL. The supported source.auth.mode and sink.auth.mode values are ActiveDirectoryInteractive, Default, ManagedIdentity, ServicePrincipal, ServicePrincipalCertificate, and Integrated. ActiveDirectoryPassword is deprecated.

ActiveDirectoryInteractive forces jobs=1; use it only where an interactive sign-in is possible. Integrated does not accept a user name. Default credentials, managed identity, service principal, and certificate flows must be supplied by the runtime environment.

Staging and limits

Complete-atomic and incremental sinks use a staging table. The identity that runs ReplicaDB needs permission to create and drop it, or to write to the configured staging table.

• Test Azure firewall, TLS, and identity access with the same host identity that will run ReplicaDB.
• Advice: validate throughput with a representative fetch.size before changing jobs.
• See Oracle, PostgreSQL, and MySQL and MariaDB for other connectors.