GitHub Issue：Windows + Node22 下 fetch failed 崩溃的稳定性加固流程

解决“gateway 启动后约 60 秒崩溃”的高危场景：先复现并采集 autoSelectFamily 日志，再加网络与异常处理防护，避免服务循环宕机。

GITHUBDiscovered 2026-02-13Author garfield2005

Prerequisites

Windows host running Node 22.x and OpenClaw gateway with reproducible `TypeError: fetch failed`.
You can capture startup logs for at least 2 minutes and control gateway restart policy.

Steps

Run a clean startup test and confirm crash timing plus `autoSelectFamily` log pattern.
Apply network-stack mitigation in staging (prefer IPv4 / happy-eyeballs fallback / fetch error catching path).
Execute 24h soak test with periodic tool calls and observe whether unhandled rejection disappears.
Promote only after crash-loop risk is gone; keep rollback ready.

Commands

openclaw gateway status

openclaw logs --local-time

openclaw gateway restart

Verify

Gateway remains healthy beyond previous 60-second failure window and no recurring `fetch failed` unhandled rejection appears.

Caveats

Do not apply aggressive network overrides directly in production without staged validation.
Final root cause/fix may depend on upstream Node/undici changes（需验证）。

Source attribution

This tip is aggregated from community/public sources and preserved with attribution.