Reliably stream Postgres table changes to external systems with automatic failover and zero event loss.
Postgres Stream captures changes from your Postgres tables and delivers them to external systems like Kafka, RabbitMQ, Redis, Webhooks, and cloud services. It uses Postgres native logical replication and stores events durably in the database itself.
Note
Postgres Stream is production ready and used in production for months already, processing millions of events per day.
- Single binary - No complex infrastructure or high-availability destinations required
- Postgres-native durability - Events are stored in the database, WAL can be released immediately
- Zero data loss - As long as downtime is less than partition retention (7 days by default)
- Automatic recovery - Handles both sink failures and slot invalidation without operator intervention
Events are inserted into the pgstream.events table and streamed via logical replication to your sink.
Two ways to create events:
- Subscriptions (optional) - Managed triggers
- Manual inserts - Insert directly into
pgstream.eventsfrom your application or database functions
While Postgres Stream provides strong durability guarantees, there are some considerations:
- Small overhead - Additional INSERT into
eventstable on every subscribed operation - Partition management - Monitor partition growth if event volume is very high
- Not for dynamic subscriptions - Each subscription change recreates database triggers
| Sink | Use Case |
|---|---|
| Kafka | High-throughput event streaming |
| NATS | Lightweight pub/sub messaging |
| RabbitMQ | Enterprise message broker |
| Redis Strings | Key-value caching |
| Redis Streams | Append-only event log |
| Webhook | HTTP POST delivery |
| AWS SQS | Managed queue service |
| AWS SNS | Managed pub/sub service |
| AWS Kinesis | Real-time data streaming |
| GCP Pub/Sub | Google Cloud messaging |
| Elasticsearch | Search indexing |
| Meilisearch | Search indexing |
# config.yamlstream:
id: 1pg_connection:
host: localhostport: 5432name: mydbusername: postgrespassword: postgrestls:
enabled: falsebatch:
max_size: 1000max_fill_secs: 5sink:
type: kafkabrokers: localhost:9092topic: events# Run with Docker
docker run -v $(pwd)/config.yaml:/config.yaml \
ghcr.io/psteinroe/postgres-stream:kafka-latestMIT