Crypto Bot Monitoring Setup

Crypto Bot Monitoring Setup

A trading or arbitrage bot — this is a stateful process that runs 24/7 with real money. Without monitoring the first failure turns into a loss you discover in the morning. A good monitoring system catches the problem before it becomes expensive.

What We Monitor

Process health: bot is alive, heartbeat comes regularly. The most basic level — /healthz endpoint, checked every minute.

On-chain state: wallet balances, no transactions pending > 10 minutes (stuck nonce), gas price within allowed range.

Exchange state: exchange API responds, WebSocket connected, open orders are current (no orders older than N hours that aren't executing).

P&L metrics: realized and unrealized P&L, daily drawdown. Alert if drawdown > X% of capital.

Monitoring Stack

Prometheus + Grafana — standard for production bots. Bot exports metrics:

import { Counter, Gauge, Registry } from "prom-client";

const register = new Registry();

const tradesTotal = new Counter({
  name: "bot_trades_total",
  help: "Total number of trades executed",
  labelNames: ["side", "symbol"],
  registers: [register],
});

const walletBalance = new Gauge({
  name: "bot_wallet_balance_usd",
  help: "Current wallet balance in USD",
  registers: [register],
});

const lastHeartbeat = new Gauge({
  name: "bot_last_heartbeat_timestamp",
  help: "Unix timestamp of last successful loop iteration",
  registers: [register],
});

// In main bot loop:
lastHeartbeat.setToCurrentTime();

Grafana dashboard with key panels: balances over time, trades per hour, P&L over time, exchange request latency.

Alertmanager for notifications:

# alerting-rules.yml
groups:
  - name: bot
    rules:
      - alert: BotHeartbeatMissed
        expr: time() - bot_last_heartbeat_timestamp > 300
        for: 2m
        annotations:
          summary: "Bot heartbeat missed for 5+ minutes"

      - alert: DrawdownExceeded
        expr: bot_daily_drawdown_pct > 5
        annotations:
          summary: "Daily drawdown exceeded 5%"

      - alert: PendingTransactionStuck
        expr: bot_pending_tx_age_seconds > 600
        annotations:
          summary: "Transaction stuck in pending for 10+ minutes"

Telegram Notifications

For immediate alerts Telegram Bot API is simplest:

async function sendAlert(message: string, level: "info" | "warn" | "critical") {
  const emoji = { info: "ℹ️", warn: "⚠️", critical: "🚨" }[level];
  await fetch(`https://api.telegram.org/bot${BOT_TOKEN}/sendMessage`, {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
      chat_id: ALERT_CHAT_ID,
      text: `${emoji} *Bot Alert*\n\n${message}`,
      parse_mode: "Markdown",
    }),
  });
}

Logging

Structured logs (JSON) with mandatory fields: timestamp, level, event, data. For aggregation — Loki (with Grafana) or Datadog:

import pino from "pino";
const logger = pino({ level: "info" });

logger.info({ event: "trade_executed", symbol: "ETHUSDT", side: "buy", amount: 0.5, price: 3400 });
logger.error({ event: "api_error", exchange: "binance", error: err.message, retryIn: 5000 });

Dead Man's Switch

For critical bots — external watchdog. If bot doesn't "knock" in N minutes — automatically restart or send SMS. Simplest implementation via healthchecks.io or Cronitor: bot sends GET request every 5 minutes, service raises alert on missing ping.

Basic monitoring setup with Prometheus, Grafana, and Telegram alerts: 1 working day. Including alerting rules configuration specific to particular bot.