Starla RIPE Atlas Software Probe

Starla is an alternative unofficial RIPE Atlas software probe written in Rust. It connects to the RIPE Atlas controller infrastructure, receives measurement requests, executes them, and reports results back to the API.

Features

• All measurement types — Ping, Traceroute, DNS, HTTP, TLS, NTP
• Pure Rust SSH — no OpenSSH dependency, uses russh
• No local ports — all communication flows through the SSH tunnel
• Minimal container image — just the binary + CA certs, multi-arch (amd64/arm64)
• NixOS module — declarative configuration with systemd hardening
• Prometheus metrics — optional observability export
• Persistent result queue — survives restarts via RocksDB

Quick Start

# Container (simplest)
docker run -d --name starla \
  -v starla-state:/state \
  --cap-add NET_RAW \
  ghcr.io/ananthb/starla:latest

# NixOS
services.starla.enable = true;

# Binary
curl -LO https://github.com/ananthb/starla/releases/latest/download/starla-amd64.tar.gz
tar xzf starla-amd64.tar.gz && sudo ./starla/starla

See the full installation guide for all options.

How It Works

On first run, starla generates an SSH key pair and registers with the RIPE Atlas registration servers. After registration, it connects to an assigned controller via SSH, receives measurement schedules over a reverse-tunneled telnet interface, executes measurements, and uploads results through SSH direct-tcpip channels.

Probe state (SSH keys, probe ID, known hosts) is stored in the state directory and persists across restarts. The measurement result queue is backed by RocksDB for crash resilience.

Registration

After first startup, register your probe at atlas.ripe.net/apply/swprobe. You'll need the probe's public key, which is printed in the log output and saved to the state directory as probe_key.pub.