products

Oracle GoldenGate: An Introduction to Data Replication

Oracle GoldenGate captures changes from a source database and delivers them to a target so you can keep selected data synchronized. Its main building blocks are Extract for capture, trail files for storing changes, and Replicat for applying them. In Microservices Architecture, Distribution and Receiver services move those trails between deployments.

This overview covers Oracle GoldenGate 26ai, specifically the self-managed Microservices Architecture platform. It explains the replication flow and the decisions you need to make before implementation; it is not an installation procedure for a particular operating system or database version.

What Oracle GoldenGate does

Change data capture, or CDC, identifies changes made to source data. For continuous replication, GoldenGate captures changes from database recovery or transaction logs; the capture method depends on the source database. For Oracle Database, those logs include redo. GoldenGate also supports initial loads, which establish a starting copy of the data on the target.

The distinction matters: an initial load supplies existing records, while ongoing replication keeps up with subsequent transactions. Turning on capture does not, by itself, establish that a target already contains a consistent copy of the source.

Oracle describes these capabilities in its GoldenGate 26ai overview. Supported sources, targets, data types, and DDL operations depend on the database, product, and release combination. Do not assume that every capability applies to every endpoint.

How the replication flow works

A simple unidirectional deployment moves selected changes from one source to one target. The following sequence describes the flow, not a set of setup commands:

  1. Extract captures source changes. During ongoing synchronization, Extract reads the supported source logs and captures transaction changes. The committed changes form the data that GoldenGate delivers downstream.
  2. Extract writes a trail. A trail is a series of files containing captured changes. It separates capture from delivery, allowing the stages to make progress independently.
  3. A distribution path transports the trail. The Distribution Service reads trail data and sends it to a receiving deployment. The Receiver Service accepts the incoming stream.
  4. Replicat applies changes. Replicat reads the target-side trail and applies mapped operations to the target database. The appropriate Replicat mode and database interface depend on the endpoint.

Oracle’s data replication components documentation describes Extract, Replicat, distribution paths, trails, and checkpoints. This flow is a starting model: a real topology can have multiple processes, targets, or intermediate deployments.

Why trails and checkpoints matter

Trails provide storage between stages rather than requiring capture and apply to run in lockstep. If delivery falls behind, the retained trail data can support later processing. Retention and available disk space therefore belong in the design, not just in housekeeping.

Checkpoints record processing positions used for restart and recovery. They are not disposable progress markers. Do not directly edit trail files, checkpoint files, or checkpoint tables. Oracle directs administrators to use supported management interfaces and warns against changing checkpoint objects without Oracle Support guidance.

Which microservices you manage

GoldenGate Microservices Architecture separates replication administration, transport, and monitoring. Oracle documents the responsibilities in its microservices components reference:

  • Service Manager manages deployments and their associated services on a host.
  • Administration Service manages Extract and Replicat processes, configuration, status, reports, and related replication objects.
  • Distribution Service transports trails to destinations through configured paths. Treat transport and data transformation as separate responsibilities; do not assume that a distribution path transforms the data.
  • Receiver Service receives incoming trail data at the destination deployment.
  • Performance Metrics Service provides monitoring information about the deployment and replication components.

The services expose web and REST-based interfaces. Admin Client communicates with these services; it is not a reason to apply commands from an older architecture unchanged. In particular, classic data-pump diagrams should not be treated as the component layout of a Microservices deployment.

When GoldenGate is useful

Common uses include maintaining reporting copies, feeding analytical systems, and supporting database migrations. For reporting, the goal is to make selected operational data available elsewhere instead of directing every report at the production source. Our MySQL business intelligence article provides a separate example of reporting and replication workloads; it is not a GoldenGate setup guide.

For migration, GoldenGate can keep target data current while the source remains in use. That does not make cutover automatic. You still need an initial-load strategy, a way to verify synchronization, an application switchover plan, and a recovery decision if validation fails.

GoldenGate is a data-movement platform, whereas Oracle Exadata is a platform for running Oracle Database workloads. Evaluating the database infrastructure and designing the replication pipeline are different tasks.

What to decide before implementation

Before selecting an installation guide or configuring a process, assemble the following prerequisites for your design:

  • The exact GoldenGate release, source and target database versions, operating systems, and processor architectures.
  • The tables and operations to replicate, including any DDL requirements and data-type restrictions.
  • The source logging and database privileges required by the chosen capture method.
  • An initial-load method and a documented transition to ongoing synchronization.
  • Secure service access, network paths, credential management, and responsibility for certificate maintenance.
  • Capacity for captured trails, temporary transaction data, and the recovery retention you need.
  • Ownership of target writes, error handling, validation, and migration cutover or rollback.

Check the GoldenGate certification information for the actual combination. Oracle’s certification and system requirements guidance explains why a general product description is not enough to establish support. Review licensing for the product and deployment you intend to use rather than assuming that a database license covers GoldenGate.

How to evaluate whether replication is healthy

A running process is only one signal. Review capture, transport, and apply progress separately; examine lag, checkpoints, reports, and errors through the supported administration and monitoring interfaces. Validate target data against the intended source state and mappings rather than treating a successful start as proof of synchronization.

If lag increases, determine which stage is falling behind before changing configuration. Check resource pressure and retained trail capacity. Oracle’s operating system requirements explains how open transactions, memory pressure, temporary storage, and trail retention affect resource needs.

If apply fails, inspect the Replicat report and relevant error details before deciding on recovery. Do not skip transactions, alter checkpoints, or remove trail files merely to clear an error. Preserve the diagnostic evidence and recoverable data, and follow the documented procedure for your database and replication mode.

Begin with one clearly defined source-to-target flow and a measurable data-freshness requirement. Expand the topology only after you have verified data correctness, restart behavior, and the operational responsibilities for that flow.