변경 데이터 캡처(CDC)

Imagine needing to know every time a record in your database is updated, inserted, or deleted, in real-time. This is where Change Data Capture (CDC) comes in!

CDC is a software design pattern used to track and capture changes made to data in a database. It focuses on identifying and capturing only the data that has changed, rather than performing full scans.

Combining CDC with Kafka unlocks powerful capabilities for real-time data processing and integration:

• Real-time Analytics: Update dashboards and reports instantly.
• Data Synchronization: Keep multiple databases or data stores consistent.
• Event Sourcing: Reconstruct the full history of changes for auditing or debugging.
• Microservices: Enable services to react to changes in other services' data without direct database access.

The most common and efficient CDC method, especially with Kafka, is log-based CDC. Databases like PostgreSQL, MySQL, and SQL Server maintain a transaction log (or write-ahead log - WAL).

This log records every change made to the database. Log-based CDC tools read these logs directly, without impacting the database's performance, to extract changes.

A typical CDC architecture with Kafka involves:

1. Source Database: The database where changes originate.
2. CDC Connector/Tool: Reads the database's transaction log.
3. Kafka Connect: A framework for connecting Kafka with other systems.
4. Kafka Topic: Where the captured change events are published.
5. Consumers: Applications that read and process the change events from Kafka.

Debezium is a popular open-source distributed platform for Change Data Capture. It provides a set of Kafka Connect connectors that monitor specific database systems.

When changes occur in your database, Debezium streams these changes as events to Kafka topics. It supports various databases like PostgreSQL, MySQL, MongoDB, and SQL Server.

To set up Debezium, you'd typically configure a connector via Kafka Connect's REST API. Here's a simplified example of a Debezium PostgreSQL connector configuration:

This config tells Debezium to monitor the mydb database on localhost:5432 and send changes from the public.customers table to Kafka topics prefixed with dbserver1.

When Debezium captures a change, it publishes an event to Kafka. This event usually contains structured information:

• before: The state of the record before the change (for updates/deletes).
• after: The state of the record after the change (for inserts/updates).
• op: The operation type (c for create, u for update, d for delete, r for read/snapshot).
• source: Metadata about the database, table, and transaction.

These events are often serialized as JSON or Avro.

A Kafka consumer application can read these CDC events and react to them. For instance, you might update a search index, invalidate a cache, or trigger another microservice.

Here's a basic Java consumer example illustrating how you might parse a Debezium event:

CDC with Kafka is incredibly versatile. Some common use cases include:

• Data Warehousing: Populate data warehouses with only changed data for efficient ETL (Extract, Transform, Load).
• Cache Invalidation: Automatically clear or update caches when underlying data changes.
• Audit Logs: Maintain a complete, immutable history of all data changes for compliance.
• Search Indexing: Keep search indexes (e.g., Elasticsearch) up-to-date with real-time database changes.

Which of the following is a primary benefit of using log-based Change Data Capture (CDC) with Kafka?

We've explored how Change Data Capture (CDC) is a powerful pattern for streaming database changes in real-time to Kafka. Using tools like Debezium, you can efficiently capture inserts, updates, and deletes from your databases.

This enables a wide array of real-time use cases, from data synchronization and analytics to event sourcing and microservice communication. Understanding CDC is key to building responsive, data-driven applications with Kafka.