Outboxx

PostgreSQL Change Data Capture in Zig

Lightweight tool that streams WAL changes to Kafka. Built in Zig for minimal resource consumption.

🚀 Development Status: Core CDC pipeline implemented with streaming replication. Under active optimization, approaching alpha release.

What is Outboxx?

Outboxx captures PostgreSQL database changes in real-time and streams them to Kafka topics. Inspired by Debezium but designed for simplicity and low resource usage.

Key Features:

• PostgreSQL streaming replication (pgoutput) ✅
• Multi-table CDC streams ✅
• Kafka producer integration ✅
• TOML-based configuration ✅
• Memory-safe Zig implementation ✅

Current vs Planned

Component	Status
PostgreSQL Streaming Replication	✅ Working
Message Processing	✅ Working
Kafka Producer	✅ Working
TOML Configuration	✅ Working
Multi-stream Support	✅ Working
Schema Registry	📋 Planned
Table/Column Filtering	📋 Planned
Production Features	📋 Planned

Inspired by Debezium

Outboxx is heavily inspired by Debezium, the industry standard for Change Data Capture. Debezium is an excellent, battle-tested solution with a rich feature set. However, being a JVM-based application, it comes with resource overhead that may not be suitable for all environments:

Aspect	Outboxx	Debezium
Runtime	Native binary	JVM (Kafka Connect)
Memory Usage	<10 MB	~200-500 MB
Startup Time	<1s	10-30s
Configuration	Simple TOML	Complex JSON/Properties
Deployment	Single binary	Kafka Connect cluster

Choose Outboxx when:

• Memory is constrained (containers, edge computing, cost optimization)
• Simple deployment without Kafka Connect infrastructure
• Learning Zig or experimenting with lightweight CDC solutions
• Native binary and minimal dependencies preferred

Stick with Debezium when:

• Production-critical workloads requiring mature, battle-tested solution
• Complex transformations and extensive connector ecosystem needed
• Kafka Connect infrastructure already in place
• Enterprise support and maximum feature coverage required

Outboxx aims to bring Debezium's excellent CDC concepts to resource-constrained environments where every MB of RAM matters.

Configuration Example

[metadata]
version = "v0"

[source]
type = "postgres"

[source.postgres]
host = "localhost"
port = 5432
database = "mydb"
user = "postgres"
password_env = "POSTGRES_PASSWORD"
slot_name = "outboxx_slot"
publication_name = "outboxx_publication"

[sink]
type = "kafka"

[sink.kafka]
brokers = ["localhost:9092"]

# Multiple streams for different tables
[[streams]]
name = "users-stream"

[streams.source]
resource = "users"
operations = ["insert", "update", "delete"]

[streams.flow]
format = "json"

[streams.sink]
destination = "user_changes"

[[streams]]
name = "orders-stream"

[streams.source]
resource = "orders"
operations = ["insert", "update"]

[streams.flow]
format = "json"

[streams.sink]
destination = "order_changes"

For complete configuration examples and architectural documentation, see docs/examples/config.toml.

Quick Start

1. Prerequisites

PostgreSQL Requirements: Version 14 or later (tested and supported versions)

Configure PostgreSQL (postgresql.conf):

wal_level = logical              # Required for logical replication (CDC)

Restart PostgreSQL after configuration changes.

Create User (run as PostgreSQL superuser):

-- Create user with REPLICATION attribute (required for replication slots and WAL access)
CREATE USER outboxx_user WITH REPLICATION PASSWORD 'secure_password';

-- Grant database access
GRANT CONNECT ON DATABASE my_database TO outboxx_user;

-- Grant schema access (CREATE needed for creating publications)
GRANT USAGE, CREATE ON SCHEMA public TO outboxx_user;

-- Grant table access (SELECT needed for logical replication)
GRANT SELECT ON TABLE my_table TO outboxx_user;

-- Enable REPLICA IDENTITY FULL (required for capturing complete row data in UPDATE/DELETE)
ALTER TABLE my_table REPLICA IDENTITY FULL;

Note: Outboxx automatically creates replication slot and publication on startup. Currently supports public schema only.

Kafka Setup:

• Topics are auto-created by default (auto.create.topics.enable=true)
• For production: pre-create topics with desired partitions and replication factor
• Outboxx will fail-fast if topic doesn't exist and auto-create is disabled

2. Running Outboxx

# Build the application
make build

# Run with your configuration
export POSTGRES_PASSWORD="your_password"
./zig-out/bin/outboxx --config config.toml

3. Production Deployment

⚠️ Outboxx requires a process supervisor (systemd, Kubernetes, Docker restart policy, supervisord) as it uses fail-fast error handling. PostgreSQL replication slots preserve state across restarts.

Contributing

This is a learning project for Zig programming. See dev/README.md for development setup.

Architecture Documentation

For complete configuration examples and design vision, see docs/examples/.

License

MIT License - See LICENSE file for details.