Skip to content
Networking

NetPulse TG: Sub-Second MikroTik ISP Down Watchdog & Telegram Relay

Architecture notes on building a lightweight daemon for MikroTik RouterOS to detect link flaps, monitor BDIX vs International latency, and relay instant Telegram alerts.

Remix into Workplace Studio:

What Was I Trying to Do?

Local ISP uplinks in Bangladesh can experience subtle partial failures: BDIX connectivity might remain healthy while upstream international transit through SEA-ME-WE 4/5 drops, or an optical fiber cut causes continuous link flapping. Standard ping monitors often fail to distinguish between total gateway failure and localized routing drops.

I created NetPulse TG to run as a low-overhead monitoring daemon that polls RouterOS via API and ICMP sockets, classifying incident severity and sending instant Telegram markdown alerts.

The Watchdog Architecture

Dual Probe Cycle (BDIX & Global DNS) ➔ RouterOS REST/API Query ➔ Flap Damping Window (30s) ➔ Telegram Webhook Queue ➔ Metric Log
  1. Dual-Zone Probe Strategy:
    • Probe A: Pings primary local BDIX exchange nodes (e.g. Bangladesh Internet Exchange peering points) at 3-second intervals.
    • Probe B: Pings upstream global resolvers (1.1.1.1, 8.8.8.8) at 5-second intervals.
  2. Anti-Flap Damping:
    • If an interface toggles state multiple times in under 30 seconds, alerts are throttled into a single consolidated incident report to avoid notification floods.
  3. RouterOS Interface Diagnostic:
    • Upon confirmed packet loss, the daemon automatically queries MikroTik via API to capture optical SFP RX power, TX/RX error counters, and current PPPoE session uptime.

Key Highlights & Performance

  • Detection Speed: Detects complete link failures within 1.8 seconds, triggering Telegram notifications in under 2.5 seconds.
  • Resource Footprint: Python monitoring service runs continuously on a headless Linux host using under 22 MB of RAM and 0.4% CPU.
  • Diagnostic Transparency: Alerts include exact RX optical power levels (e.g., -24.5 dBm) so network administrators immediately know whether an issue is an optical fiber bend or an upstream BGP routing withdrawal.

What I Learned

Raw ICMP pings alone can give false alarms during momentary packet burst queues. Combining socket-level timeouts with RouterOS interface statistics provides true operational insight without false outage alarms.

MikroTikRouterOSTelegram BotISP MonitoringPython
Md Sabbirul Islam Khan

Md Sabbirul Islam Khan

Written by THE SABBiR