Oracle connector

Oracle is supported as source and sink for complete, complete-atomic, and incremental replication. Use the jdbc:oracle: scheme with the Oracle thin driver (oracle.jdbc.OracleDriver).

Connection string formats

Keep host, port, SID, service, and credentials in the runtime environment.

Addressing	Format
SID	jdbc:oracle:thin:@${ORAHOST}:${ORAPORT}:${ORASID}
Service name	jdbc:oracle:thin:@//${ORAHOST}:${ORAPORT}/${ORASERVICE}
TNS alias	jdbc:oracle:thin:@${ORATNSALIAS}

A TNS alias requires tnsnames.ora to be reachable through the Oracle client configuration. Advice: service names are the usual choice for pluggable databases.

Example properties

source.connect=jdbc:oracle:thin:@//${ORAHOST}:${ORAPORT}/${ORASERVICE}
source.user=${ORAUSER}
source.password=${ORAPASSWORD}
source.table=SOURCE_SCHEMA.ORDERS
source.connect.parameter.oracle.net.tns_admin=${TNS_ADMIN}
source.connect.parameter.oracle.net.networkCompression=on

sink.connect=jdbc:oracle:thin:@//${ORAHOST}:${ORAPORT}/${ORASERVICE}
sink.user=${ORAUSER}
sink.password=${ORAPASSWORD}
sink.table=SINK_SCHEMA.ORDERS

mode=complete-atomic
jobs=4

oracle.net.tns_admin and oracle.net.networkCompression are Oracle JDBC driver properties. ReplicaDB does not interpret them; it passes them to the driver as source.connect.parameter.* or sink.connect.parameter.* entries. Check the Oracle JDBC documentation for the values your driver version supports.

Parallel reads and Flashback

Table-backed reads with jobs greater than 1 partition rows with ora_hash(rowid, jobs - 1) = ?. A free-form source.query owns its own selection and does not receive this partitioning.

Flashback: ReplicaDB reads the current system change number with SELECT CURRENT_SCN FROM V$DATABASE and applies AS OF SCN to the source read. If the SCN cannot be read, ReplicaDB logs a warning and continues without Flashback.

If a read fails with ORA-01555 (snapshot too old) or another SCN-related error, ReplicaDB logs the error and rethrows the exception. Advice: increase undo retention or shorten the run window, then rerun.

Limits and notes

• LOB handling: BLOB and CLOB columns are streamed to the sink. Advice: test LOB and vendor-specific types with representative values before a large transfer.
• The sink applies bandwidth.throttling on the insert path.
• Complete-atomic and incremental sink runs write to a staging table before replacement or merge. Set sink.staging.schema to place generated staging tables in a dedicated schema. A fixed staging table can be named with sink.staging.table, but it cannot be used with replication.table.* entries.
• General guidance: run ReplicaDB close to the database server to reduce network latency. This is not a ReplicaDB requirement.

See also: PostgreSQL, SQL Server, and DB2.