Oracle Change Data Capture (CDC)
Survived, partially.
✓ certified by the mortician
In life
Oracle Change Data Capture recorded inserts, updates and deletes on source tables into change tables that subscribers consumed through DBMS_CDC_PUBLISH and DBMS_CDC_SUBSCRIBE. It offered synchronous capture via triggers and asynchronous capture built on redo mining and Streams.
Migrations care because downstream extract-transform-load jobs that read change tables must be redesigned; Azure options include Data Factory change data capture, GoldenGate or other log-based tools. The presence of CDC also tells you the source was on 11.2 or earlier, which affects which migration tools can connect to it.
Oracle's 12.1 Upgrade Guide states that Change Data Capture is not included in 12c and has been replaced by GoldenGate. That page does not name the release in which it was deprecated, so the deprecation field is left unset.
Cause of departure
Desupported in Oracle Database 12.1. Oracle names Oracle GoldenGate as the successor.
Obituary
Oracle Change Data Capture, born before 11.2, published row changes through DBMS_CDC_PUBLISH change tables for ETL consumers. It served downstream systems faithfully, preserving the useful details of change.
It was desupported in 12.1, with Oracle GoldenGate named as its replacement. Azure offers successors, though none inherits the entire estate without conditions.
It is survived by Azure Database for PostgreSQL Flexible Server, partly; Azure SQL Managed Instance, partly; Azure Data Factory, only by workaround; and Microsoft Fabric, partly. Their duties range from logical decoding and change tables to watermark pipelines and one-way mirroring into Delta tables—respectable arrangements, but not quite the deceased.
Survived by
Azure Database for PostgreSQL Flexible Server partial effort M
PostgreSQL logical decoding on Flexible Server exposes row changes from the write-ahead log through replication slots and output plugins (pgoutput, wal2json, test_decoding) that consumers such as Debezium read, and native logical replication publishes selected tables to subscribers. This covers Oracle CDC's change-table role for downstream systems, without DDL replication and with the operational duty of dropping unused slots before they fill storage.
- Set wal_level to logical, raise max_worker_processes and restart; grant REPLICATION to the replication role.
- Create publications for the tables that had change tables, or a wal2json slot for a change stream.
- Attach subscribers or a Debezium-style connector to the slot.
- Monitor pg_replication_slots and alert on WAL growth.
Complications
Azure SQL Managed Instance partial effort M
After the schema moves to SQL, SQL Server Change Data Capture on Managed Instance provides the same model Oracle CDC offered: log-based capture into per-table change tables with before/after images, operation codes and LSN ordering, read through table-valued functions. It depends on SQL Server Agent capture and cleanup jobs, retains three days of changes by default and allows at most two capture instances per table.
- Enable CDC on the database and on each formerly published table with sys.sp_cdc_enable_db and sys.sp_cdc_enable_table.
- Rewrite subscribers to call cdc.fn_cdc_get_all_changes_ or _net_changes_ functions between LSN bounds.
- Adjust the cleanup job retention to the consumers' latency.
- Watch log truncation, which waits for the capture job.
Complications
Microsoft Fabric partial effort M
For the classic CDC use case, feeding a warehouse with incremental changes, Fabric mirroring of Oracle does the capture (LogMiner) and the apply (Delta tables in OneLake) as a managed service, replacing both the change tables and the ETL that read them. Consumers query the mirrored tables instead of subscribing to change sets, so downstream jobs that depended on change-set semantics (before and after images, operation codes) must be redesigned around Delta time travel or snapshots.
- Satisfy the Oracle prerequisites (archive log mode, supplemental logging, LogMiner privileges).
- Mirror the tables that had CDC change tables into a Fabric mirrored database.
- Rebuild warehouse loads as reads of the mirrored Delta tables or Direct Lake models.
- Tune the Delta retention setting if consumers need change history.
Complications
Azure Data Factory workaround effort L
Oracle CDC (the DBMS_CDC_PUBLISH change tables, desupported in 12.1) published row changes for ETL consumers, which is exactly Data Factory's purpose, but Data Factory has no native change capture for Oracle sources. The supported patterns are a customer-managed watermark pipeline over a timestamp or sequence column, or landing changes from another capture mechanism (Fabric mirroring, GoldenGate) in storage and picking up new files.
- Choose the delta detection column per table and add one where none exists.
- Implement the watermark control table and pipeline; run on a tumbling-window trigger.
- Where deletes matter, add soft-delete markers or reconcile periodically.
- Consider Fabric mirroring when the consumer is analytics rather than another operational database.
Notices of correction
- oracle Deprecated and Desupported Features for Oracle Database 12c (12.1) · Upgrade Guide 12.1, E41397 · read 2026-09-26
- microsoft Logical replication and logical decoding in Azure Database for PostgreSQL Flexible Server · page updated 2026-07-20 · read 2026-09-26
- microsoft What is change data capture (CDC)? (SQL Server and Azure SQL Managed Instance) · page updated 2026-08-25 · read 2026-09-26
- microsoft Compare SQL Database Engine features: Azure SQL Database vs Azure SQL Managed Instance · page updated 2026-09-09 · read 2026-09-26
- microsoft Mirror Oracle databases in Microsoft Fabric · page updated 2026-05-21 · read 2026-09-26
- microsoft Mirroring in Microsoft Fabric · page updated 2026-08-28 · read 2026-09-26
- microsoft Change data capture in Azure Data Factory and Azure Synapse · page updated 2026-08-05 · read 2026-09-26
- microsoft Copy data to and from Oracle with Azure Data Factory · page updated 2026-08-05 · read 2026-09-26