Change Data Capture (Planned)
Planned — Q1 2027
CDC is on the committed roadmap for Q1 2027. Requires both Managed Kafka (streaming) and Managed PostgreSQL (databases) to be operational.
Purpose: For platform engineers and data engineers, explains the planned CDC service — scope, supported databases, lifecycle model, and dependencies.
Overview
Change Data Capture captures row-level changes from source databases and publishes them as events to Kafka topics. openCenter CDC uses Debezium connectors deployed on KafkaConnect, managed via KafkaConnector CRDs in Git and reconciled by FluxCD.
This is a narrow, opinionated CDC service — not a general-purpose data integration platform. PostgreSQL is the primary source; MySQL and MongoDB support follows.
Planned Scope
| Capability | Description |
|---|---|
| PostgreSQL capture | pgoutput logical decoding — row-level inserts, updates, deletes |
| MySQL capture | binlog-based capture |
| MongoDB capture | Change streams |
| Kafka sink | Events published to Kafka topics (one topic per table default) |
| Schema handling | Integrates with Schema Registry (Q4 2026) when available |
| Connector lifecycle | Deploy, pause, resume, restart, delete via KafkaConnector CRD |
| Monitoring | Prometheus metrics for connector lag, errors, throughput |
| GitOps managed | Connector definitions in Git, FluxCD reconciliation |
Supported Databases
| Database | Mechanism | Minimum Version | Status |
|---|---|---|---|
| PostgreSQL | pgoutput logical replication | 14+ | Primary target |
| MySQL | binlog (ROW format) | 8.0+ | Secondary |
| MongoDB | Change streams | 6.0+ | Secondary |
Lifecycle Operations
| Operation | Mechanism |
|---|---|
| Deploy connector | Apply KafkaConnector CR via Git commit |
| Pause capture | Set state: paused in KafkaConnector CR |
| Resume capture | Set state: running — resumes from last committed offset |
| Restart (error recovery) | Annotation-triggered restart via operator |
| Schema evolution | Handled by Schema Registry compatibility checks |
| Upgrade Debezium | Connector image version bump in KafkaConnect CR |
Architecture
Dependencies
| Dependency | Required | Notes |
|---|---|---|
| Managed Kafka | Yes | Sink for CDC events; provides KafkaConnect runtime |
| Managed PostgreSQL | Yes (for PG source) | Source database with logical replication enabled |
| Schema Registry | Recommended | Schema evolution without breaking consumers |
| cert-manager | Yes | TLS for connector-to-database and connector-to-Kafka |
| kube-prometheus-stack | Yes | Connector metrics and alerting |
| FluxCD | Yes | GitOps lifecycle for KafkaConnector CRDs |
Constraints
- PostgreSQL source requires
wal_level = logical(configured by Managed PostgreSQL service) - One Debezium connector per source database (fan-out via Kafka topics)
- Large initial snapshots may require dedicated KafkaConnect task allocation
- Topic naming follows convention:
<connector-name>.<schema>.<table>
Further Reading
- Data Services Overview — family overview
- Streaming Blueprint — Kafka and KafkaConnect details
- Schema Registry (Planned) — schema compatibility enforcement
- PostgreSQL (Planned) — source database service
- Portfolio Strategy — sequencing rationale