arrakis-server 0.9.0

Server implementation of the Arrakis low-latency timeseries data distribution platform

arrakis-server

Arrakis server

---

Server implementation of the Arrakis low-latency timeseries data distribution platform. Serve historical and live detector data to clients over Apache Arrow Flight, either from a backend you provide or by acting as an information server that routes requests to other Arrakis endpoints.

Resources

• Documentation
• Source Code
• Issue Tracker
• Arrakis Python client

Installation

With pip:

pip install arrakis-server

With Docker:

docker run --net=host -it containers.ligo.org/ngdd/arrakis-server:latest

Where to Start

• Tutorial — New to arrakis-server? Run a mock server and connect a client step by step.
• User Guide — Running the server, configuring backends, scope maps, multi-backend deployments, and production tips.
• Background — How the server works: Flight dispatch, the backend trait, retention-based routing, and the metadata model.
• API Reference — Auto-generated documentation from source code.

Features

• Pluggable backends via Python entry points
• Built-in mock backend for local testing and CI
• Information-server mode that routes requests across multiple endpoints
• Retention-aware request routing, splitting time ranges across live and historical backends
• Time-valued channel metadata for evolving detector configurations
• Publish and partition endpoints for Kafka-backed publishing backends

Quickstart

Run a mock server

The mock backend generates synthetic timeseries for a predefined set of H1 and L1 channels:

arrakis-server mock

Then connect with the Arrakis client CLI:

arrakis find "H1:.*"

Serve custom mock channels

Pass one or more TOML files defining your own channels:

arrakis-server mock my_channels.toml

A minimal channel file:

["common"]
publisher = "MY_PUBLISHER"
sample_rate = 256
data_type = "float32"
stride = 62_500_000
max_latency = 1_000_000_000

["MY:CHANNEL-NAME"]
func = "3*t + cos(t)"

func accepts any SymPy expression in t (GPS seconds) to generate deterministic waveforms.

Run as an information server

An information server holds no data of its own; it routes client requests to other Arrakis endpoints based on a scope map:

arrakis-server --scope-map scope.yaml

The scope map lists each downstream endpoint along with the scopes and retention windows it serves. See the Scope Maps guide for the file format, and Multi-Backend Deployments for composing live and historical backends.

Write a backend

Backends plug in via the arrakis-server-backend entry-point group. A minimal backend implements the ServerBackend protocol and registers in its package's pyproject.toml:

[project.entry-points.arrakis-server-backend]
mybackend = "my_package.backend:MyBackend"

See Writing a Backend for the full interface and a complete example.