openclaw - 💡(How to fix) Fix [Bug]: Telegram polling fails with "DNS-resolved IP unreachable" on Windows — node fetch works fine [1 comments, 2 participants]

# Telegram polling fails with "DNS-resolved IP unreachable" on Windows — node fetch works fine

## Environment

- OpenClaw: 2026.5.2 (8b2a6e5)
- OS: Windows 11 Pro (10.0.22621)
- Node.js: 24.14.1
- Gateway mode: local, bind=lan, port=19000
- Channel: Telegram (long-polling)

## Problem

The gateway starts normally and the Telegram provider initializes (`[telegram] [default] starting provider (@my_bot)`), but polling never succeeds. The log shows this repeatedly:


[telegram/network] fetch fallback: DNS-resolved IP unreachable; trying alternative Telegram API IP (codes=none, reason=request-timeout)


The bot never receives messages. After several cycles, polling stalls are detected and the transport is rebuilt, but the problem persists.

## What works

Direct fetch from Node.js in the same environment succeeds instantly:


node -e "fetch('https://api.telegram.org/bot<token>/getMe').then(r=>r.json()).then(d=>console.log(JSON.stringify(d)))"
# Returns {"ok":true,"result":{"id":...,"is_bot":true,...}}


`curl` and `nslookup` also work. DNS resolves to both IPv4 (149.154.166.110) and IPv6 (2001:67c:4e8:f004::9). No outbound firewall rules block node.exe. Only Windows Defender is installed.

## What I tried

1. Setting `NODE_OPTIONS=--dns-result-order=ipv4first` as a Machine-level environment variable — no effect.
2. Adding `set "NODE_OPTIONS=--dns-result-order=ipv4first"` directly in `gateway.cmd` — no effect.
3. Running the gateway manually in foreground (`node ... gateway --port 19000`) instead of as a Scheduled Task — same error.
4. Killing all node.exe processes and restarting cleanly — same error.
5. Upgrading from 2026.4.29 to 2026.5.2 — same error (was present before upgrade too).

## Observations

- The timeout happens specifically on the Telegram `getUpdates` long-poll request, not on short requests.
- The gateway's internal fetch mechanism seems to behave differently from Node's global `fetch()`.
- `event_loop_utilization` frequently hits 0.93+ during polling attempts, suggesting the process gets bogged down.
- On one occasion after a restart, the bot briefly responded with "Something went wrong while processing your request" (meaning the Telegram connection worked momentarily but the Anthropic API call failed). Subsequent restarts went back to the DNS timeout.

## Expected behavior

The gateway should be able to reach `api.telegram.org` for long-polling, since Node.js on the same machine can do so without issues.

## Logs (excerpt)


12:29:44 [telegram] [default] starting provider (@my_bot)
12:29:47 [telegram] menu text exceeded the conservative 5700-character payload budget; shortening descriptions to keep 65 commands visible.
12:30:53 [telegram/network] fetch fallback: DNS-resolved IP unreachable; trying alternative Telegram API IP (codes=none, reason=request-timeout)