OpenSeer — Distributed HTTP Monitoring

Self-hosted, horizontally scalable HTTP monitoring with a modern web interface. Multi-user support with secure authentication, real-time dashboards, and comprehensive monitoring management. Built with Go backend and Next.js frontend.

Architecture

graph TB
    Users[👥 Users]
    WebUI[🌐 Web UI<br/>Next.js :3000]
    ControlPlane[🎯 Control Plane<br/>Go :8080]
    Database[(🗄️ TimescaleDB<br/>PostgreSQL)]

    subgraph "Workers - Distributed"
        WorkerUS[⚡ Worker US-East<br/>Go - us-east-1]
        WorkerEU[⚡ Worker EU-West<br/>Go - eu-west-1]
        WorkerAP[⚡ Worker AP-South<br/>Go - ap-south-1]
    end

    Users --> WebUI
    WebUI -.->|Session Auth<br/>Connect RPC| ControlPlane
    WorkerUS -->|Token Auth<br/>HTTP Polling| ControlPlane
    WorkerEU -->|Token Auth<br/>HTTP Polling| ControlPlane
    WorkerAP -->|Token Auth<br/>HTTP Polling| ControlPlane
    ControlPlane --> Database

    subgraph "Database Schemas"
        AppSchema[📋 app schema<br/>monitors, jobs, workers, worker_capabilities]
        TSSchema[📊 ts schema<br/>results_raw, results_agg_*]
        AuthSchema[🔐 auth tables<br/>user, session, account, ...]
    end

    Database --- AppSchema
    Database --- TSSchema
    Database --- AuthSchema

Loading

OpenSeer consists of four main components:

• Web Frontend (Next.js) - Modern dashboard with real-time metrics visualization, multi-user support with secure session-based authentication
• Control Plane (Go) - Service managing workers and job scheduling via HTTP APIs
• Workers (Go) - Distributed agents executing HTTP checks across geographic regions, using HTTP polling with token-based authentication
• Database (PostgreSQL + TimescaleDB) - Time-series storage with automatic aggregation (1-minute, 1-hour, and 1-day intervals), P50/P95/P99 latency tracking, and uptime statistics

Worker Communication

Workers communicate with the Control Plane using a simple HTTP polling model:

1. Enrollment: Workers register using a cluster token and receive an API token (ostk_...)
2. Job Polling: Workers poll GetJobs endpoint to fetch available jobs for their region
3. Result Submission: After executing checks, workers submit results via SubmitResult
4. Lease Renewal: For long-running checks, workers can extend job leases via RenewLease

This polling-based architecture provides:

• Simplicity: No persistent connections to maintain
• Scalability: Stateless control plane, easy horizontal scaling
• Resilience: Workers automatically reconnect on network issues

For detailed architecture documentation, see:

• Control Plane Architecture
• Worker Architecture
• Multi-Cloud Production Guidance

Quick Start

For a fresh environment:

task migrate
task up

task migrate runs backend and web auth migrations. task up starts the app profile (TimescaleDB, control plane, worker, and web).

Then access:

• Web UI: http://localhost:3000
• API: http://localhost:8080

To stop services:

task down

To stop services and remove volumes:

task clean

Task Commands

task up        # Start all app services
task down      # Stop app services
task clean     # Stop app services and remove volumes
task migrate   # Run backend + web migrations
task --list    # Show all available tasks

Contributing

Contributions are welcome! Please see the detailed architecture documentation in the cmd/ directories for implementation details.

License

MIT