OpenClaw Telegram: 2,000+ EADDRINUSE Errors in Your Logs — And Your Bot is Working Fine

Published April 18, 2026 · Bug #68153 · Affects: 2026.3.1+ · Setup: OpenClaw with Telegram channel enabled

Your OpenClaw gateway is running. Your Telegram bot responds to messages perfectly. But your logs look like this — every 30 minutes:

[telegram:default] health-monitor: restarting (reason: stuck)
[default] starting provider (@your_bot)
canvas host mounted at http://0.0.0.0:18789
Gateway is binding to a non-loopback address
Gateway failed to start: another gateway instance is already listening on ws://0.0.0.0:18789
Port 18789 is already in use.
- pid 1234 openclaw: openclaw-gateway (*:18789)

~67 lines of errors per cycle, repeating indefinitely. One user reported 2,204 EADDRINUSE errors in 27 hours across 33 restart cycles — while their bot worked perfectly the entire time.

What's Happening

OpenClaw's channel health-monitor watches for "stuck" channels. Its logic: if no events have arrived in 30 minutes, the channel must be unhealthy — restart it.

This makes sense for WebSocket-based channels like Slack, where a stale socket genuinely means something broke. For Telegram's long-polling, it doesn't — quiet periods with no incoming messages are completely normal. Nobody messaged you. That's not a bug.

When the health-monitor triggers a restart anyway, the Telegram channel restart path attempts to rebind the main WebSocket port (18789). That port is already owned by the running gateway — so you get EADDRINUSE. The channel never actually restarts. The loop repeats.

This is the same class of bug that affected MS Teams channels in issues #22169, #24522, and #25527 — fixed for Teams, but the same path exists in the Telegram restart code.

The Impact

• Functional: Zero. Your bot works fine throughout.
• Operational: Logs become useless noise. Monitoring alerts fire constantly. Hard to spot real errors.
• Performance: Minor CPU overhead from repeated failed bind attempts.

Workarounds (Until Patched)

Option 1: Increase the stale event threshold

Raise the health-monitor's "stuck" threshold so it doesn't trigger during normal quiet periods. Add to your openclaw.json:

{
  "gateway": {
    "health": {
      "staleEventThresholdMs": 7200000
    }
  }
}

This sets it to 2 hours. For a personal bot with infrequent messages, you might go even higher (4–8 hours).

Option 2: Disable health-monitor for Telegram only

If your Telegram channel is stable and you don't want the overhead:

{
  "channels": {
    "discord": {
      "health": { "enabled": false }
    }
  }
}

Replace "discord" with your channel key (e.g. "telegram"). Check your config for the exact key name.

Option 3: Disable health-monitor globally

{
  "gateway": {
    "health": {
      "enabled": false
    }
  }
}

Trade-off: you lose automatic recovery if a channel genuinely gets stuck. Acceptable for stable single-channel setups.

Option 4: Filter logs while waiting for the patch

openclaw gateway logs | grep -v "EADDRINUSE\|already in use\|health-monitor: restarting"

Not a fix, but makes logs readable in the meantime.

Root Cause (Technical)

In channel-health-monitor.ts (~line 1595), isChannelHealthy() uses a single stale-event threshold for all channel types. Long-polling channels (Telegram) should be excluded from the "no events = unhealthy" check, or the threshold should be channel-type-aware. The fix is tracked in issue #68153.

Is Your Bot Actually Broken?

Quick check:

openclaw gateway status

If it shows running and your bot responds to test messages, the errors are cosmetic. You're hitting this bug, not a real outage.

Summary

Need help configuring OpenClaw for Telegram or diagnosing gateway issues? ClawReady offers setup and audit services — we'll get your config clean and your logs quiet.