Skip to main content
Change data capture (CDC) sources read inserts, updates, and deletes from operational databases or their transaction logs and emit them as a stream Planasonix can process. You choose capture mode, initial snapshot behavior, and tables to include, then connect downstream transforms and destinations.

​
Database CDC

Planasonix supports CDC for common relational engines. Exact prerequisites (binary logging, supplemental logging, permissions) appear in each connector’s setup wizard.
Use logical decoding (for example, pgoutput or plugin-based slots) with replication permissions. Monitor replication slot lag on the database to avoid disk growth from idle consumers.
Enable row-based binary logging and grant replication client privileges. Large wide tables benefit from careful primary key design to keep event size predictable.
Use change tracking or CDC features as supported by the connector; ensure agents can read from the capture instance without blocking OLTP workloads.
Configure LogMiner or vendor-recommended equivalents per connector docs; archive log retention must exceed your worst-case consumer downtime.
Capture change streams on collections you authorize; consider resume tokens and schema evolution when documents are polymorphic.
Always test initial snapshots on a replica before pointing at production primaries if your DBA policy requires it.

​
Event buses and cloud streaming

In addition to database CDC, you can ingest from high-throughput buses and managed streaming services:
PlatformTypical use
Apache KafkaDurable topics from microservices or CDC Debezium connectors
Amazon KinesisAWS-native streaming with shard-based scaling
Apache PulsarMulti-tenant messaging with geo replication
Google Cloud Pub/SubAsynchronous work queues and GCP-native events
Azure Event HubsKafka-compatible endpoints on Azure
NATSLightweight messaging for edge and internal services
Vendor webhooksHTTPS push endpoints normalized into an internal stream
Configure bootstrap servers, SASL/SSL, consumer groups, and topic patterns. Set partition-aware parallelism only when your downstream processing is associative per key.

​
Operational checklist

1

Rightsize retention

Ensure broker and database retention cover maintenance windows plus worst-case recovery time.
2

Schema strategy

Define how additive schema changes propagate; use parsers and data contracts where you enforce compatibility.
3

Backpressure

Monitor consumer lag; scale processors or destinations before lag triggers SLA breaches.

Stream processing

Windows and transforms on live data.

Observability

Metrics and alerts for streaming health.