Snodra¶
One messaging stack for services that speak Kafka and gRPC — without a commercial messaging platform.
Snodra gives a service commands, queries, events and sagas. Commands and queries go through Snodra Server, a stateless gRPC router. Events go to the event store you already run — Kafka or RabbitMQ Streams — through a transactional outbox, with the message store of your choice (MongoDB, PostgreSQL or MySQL) holding the outbox, the processed markers and the saga state. One protocol, first-class Kotlin and Rust libraries.
You run the server, point your services at it, and stop maintaining a service registry, deadline handling and a control plane of your own.
Map of the documentation¶
-
Install the server, add a client library, and send a first command.
-
CQRS, commands, queries, events, sagas and the guarantees behind them.
-
The server, the Kotlin library, the Rust library and the wire contract.
-
Event stores and message stores: Kafka, RabbitMQ Streams, MongoDB and SQL.
-
Monitoring, testing, tuning and upgrades for a running deployment.
-
Spring Boot, an OAuth2 provider, OpenTelemetry, containers and existing services.
-
The event browser and the conformance runner.
-
Versions, upgrades, and what the protocol version guarantees.
-
Runnable starting points, the licence and the editions of Snodra.
Where to go first¶
| You are… | Start with | Then |
|---|---|---|
| new to CQRS and event-driven services | CQRS | Commands, Queries and Events |
| adopting Snodra in an existing system | Installation | Your first service and Spring Boot |
| operating Snodra | Operations | Monitoring, Tuning and Upgrades |
Try the server¶
docker run --rm -p 7000:7000 -p 7001:7001 \
-e SNODRA_SERVER_OTLP_ENDPOINT= \
snodra/snodra-server:0.5.0
The gRPC listener is on 7000; the dashboard and its API are on http://localhost:7001/ui/. An
empty topology is correct: the server learns ownership from the connections. Installation
covers the client libraries, and the quick start takes the server from there.
What it is not¶
A message broker, an event store, a durable queue, or a workflow engine. The server routes commands and queries to a live handler and relays the answer; events go to your event store and state lives in your database. Core concepts explains what each part owns.
Licence¶
Apache-2.0. See Editions for what is open source and what is sold separately.