Show HN: ReJot – Database replication framework aimed at developers

There is a gap between building internal (REST) APIs and Kafka (event streaming) to share data between services.

Internal APIs start to break down when you have more than a couple services communicating. Their synchronous nature makes them brittle in a distributed system: failures cascade and latency adds up. Companies operating internal APIs at scale often face challenges like managing implicit schemas and versioning. They also need to write significant amounts of code to implement features like circuit breakers and internal load balancing.

Event streaming addresses these issues by using asynchronous communication, but it also introduces significant drawbacks. Kafka is known for its operational complexity and high cost. Engineers must manage outbox tables, outbox processors, and consumers, which makes the system more difficult to understand and maintain.

ReJot is the middle ground solution that re-uses a database system's write-ahead log as an asynchronous communication channel. The WAL is well-suited to double as an outbox, this has been proven by CDC systems like Debezium. ReJot is a lightweight addition to existing infrastructure, and even re-uses existing (relational) database systems to store messages (temporarily) before sending them to the destination/sink databases.

We're developer focused, as opposed to being infrastructure focused. Much like how developers define the database table schemas they use, we enable developers to say how their data should be published to others in the distributed system. This is done through something we call "Public Schemas", they consist of a schema and a (SQL) query. When an item in the underlying table changes, the query is executed to produce an object conforming to the schema. This data is then forwarded through ReJot, ready to be consumed by a different service using a "Consumer Schema". This is again a simple (SQL) query that contains an INSERT statement. All of this is defined from within the codebase of the application, much like how ORMs or query builders work.

In short, ReJot re-uses your database in two ways: by consuming the WAL, and also by using queries to encapsulate and integrate data. This makes ReJot a good middle-ground between the brittleness of synchronous communication and the complexity of event streaming.

Excited to hear what you think!