Modern enterprise systems rarely keep all their data in one place.
Customer information may live in PostgreSQL, orders in MySQL, events in Apache Kafka, analytics in a Data Warehouse, and searchable data in Elasticsearch.
So how do we reliably connect all these systems without developing and maintaining custom Kafka producers and consumers for every integration?
That's where Kafka Connect comes in.
Kafka Connect architecture at a glance
External Source
↓
Source Connector
↓
Kafka Connect
↓
Kafka Topics
↓
Kafka Connect
↓
Sink Connector
↓
External Target
The easiest rule to remember is:
SOURCE = data coming INTO Kafka
SINK = data going OUT OF Kafka
But production Kafka Connect architecture involves much more than just Source and Sink Connectors.
In my latest guide, I cover:
🔹 Kafka Connect architecture
🔹 Source vs Sink Connectors
🔹 Connectors, Workers & Tasks
🔹 Converters
🔹 Single Message Transforms (SMTs)
🔹 Offset management
🔹 Standalone vs Distributed mode
🔹 Internal Kafka Connect topics
🔹 REST API
🔹 Error handling & Dead Letter Queues
🔹 Change Data Capture (CDC)
🔹 Debezium architecture
🔹 Initial snapshots + continuous CDC
🔹 Kafka Connect vs Kafka Streams
🔹 Kafka Connect + CDC + Microservices
🔹 Production scaling, security & monitoring
CDC with Debezium
One particularly powerful architecture is:
Application
↓
PostgreSQL / MySQL
↓
Transaction Log / WAL / Binlog
↓
Debezium
↓
Kafka Connect
↓
Kafka Topics
↓
Microservices / Analytics / Search
This allows database changes to become events that downstream applications can process without repeatedly polling the operational database.
I’ve included detailed diagrams and practical explanations throughout the full guide.
👉 Read the complete English article:
https://shikhanirankari.blogspot.com/2026/09/kafka-connect-architecture-source-sink-cdc.html
🇫🇷 French version:
https://shikhanirankari.blogspot.com/2026/09/architecture-kafka-connect-source-sink-cdc.html