Syncra — Network Monitoring Suite

Syncra is a distributed monitoring system with a Core ↔ Probe design. Probes auto-register, stream metrics in real time, and can be orchestrated remotely without agents or manual config drift.

Overview

Syncra helps teams see their infrastructure the way it actually behaves—live. It discovers running probes automatically, normalizes their heartbeats and metrics, and renders a coherent timeline you can trust during incidents.

• Auto-discovery: Probes self-register to a Redis-backed registry. No manual node lists.
• Zero-drift orchestration: Probes receive signed tasks from Core (checks, collectors, diagnostics).
• Live metrics: Streams metrics via server-sent events and persists deltas efficiently.
• Typed API: End-to-end types via tRPC for commands, events, and results.

Problems it solves

• Missing nodes or stale inventories during incidents
• Fragmented metrics, logs, and checks without a shared context
• Complex configuration surfaces that drift from reality

Architecture

1. Probe boot

   • Probe generates an ephemeral identity and announces itself to the registry.
   • Core verifies signature and attaches intent (what to measure, how often).

2. Streaming loop

   • Probe emits heartbeat + metrics with backpressure-aware batching.
   • Core ingests, normalizes, and enriches with topology.

3. Remote execution

   • Core dispatches signed tasks (e.g., TCP check, HTTP SLO probe, traceroute).
   • Probe executes in an isolated sandbox; results are streamed back.

// Simplified event model (tRPC contracts)
export type Heartbeat = {
  probeId: string
  version: string
  ts: number
  region?: string
}

export type Metric = {
  probeId: string
  kind: "cpu" | "mem" | "latency" | "uptime"
  value: number
  unit: "%" | "ms" | "s"
  ts: number
}

Live Topology

The registry maintains a consistent view of all active probes. When probes drop, the UI reflects it within a single interval. This removes guesswork when correlating incidents across regions.

Design note: The UI favors low-noise density. We bias for clarity over charts-for-charts’ sake.

Highlights

• One-command probe bootstrap with signed enrollment
• Real-time stream rendering with graceful degradation on slow links
• Pluggable collectors (HTTP, TCP, ICMP, DNS, custom)
• Simple, typed commands with predictable failure modes

Demo

# Run a probe locally (demo)
SYNC_CORE_URL=https://core.syncra.local \
SYNC_REGION=eu-central-1 \
syncra-probe run

Results

• Mean triage time: −30% after consolidating checks + metrics into timelines
• MTTR: meaningfully reduced by removing stale inventories
• Confidence: on-call engineers get an authoritative, live state

What’s next

• Alert policies with noise budgets
• eBPF-based collectors for low-overhead syscall sampling
• Synthetic SLO monitors with canary routing

Notes

This project was built with Go, Next.js, Tailwind, and tRPC. Images are optimized and lazy-loaded. The page design emphasizes legibility, motion restraint, and intentional color accents.