Crate krabka_connect_worker 

krabka-connect-worker

krabka-connect-worker runs one PostgreSQL logical-decoding source connector per process and writes its change records to Kafka. It uses an idempotent producer with acks=all and persists the acknowledged PostgreSQL LSN in the compacted __krabka_connect_offsets topic, keyed by connector and source identity. At startup it registers the shared CDC key and value Protobuf schemas through the configured Schema Registry and frames records with the allocated IDs. The key also hashes the PostgreSQL URL and slot, so changing the source cannot reuse an incompatible checkpoint from an earlier deployment.

Auto-created data and checkpoint topics use the configured replication factor; the managed operator derives it from the cluster size, capped at three.

Delivery is at least once: a crash after Kafka acknowledges data but before the checkpoint record is durable can replay records, but the worker never advances the PostgreSQL slot before both barriers complete.

The HTTP listener exposes /live, /ready, and /metrics.

Re-exports

pub use config::WorkerConfig;

pub use kafka::CHECKPOINT_TOPIC;

pub use kafka::KafkaCheckpointStore;

pub use kafka::KafkaSink;

Modules

config

Command-line and environment configuration for one connector worker.

kafka

Kafka sink and compacted-topic checkpoint store.

Functions

run

Run one Postgres-to-Kafka connector until Ctrl-C and drain it gracefully.