openclaw - 💡(How to fix) Fix [Bug]: Gateway crash-loop on startup when bonjour/mDNS plugin hits "CIAO PROBING CANCELED" [7 comments, 7 participants]

Bug type Crash (process/app exits or hangs)

Beta release blocker No

Summary Gateway crash-loops on startup when the bonjour/mDNS plugin hits CIAO PROBING CANCELLED, and disabling bonjour stops the failure.

Steps to reproduce Install OpenClaw on a Linux host with the bonjour plugin enabled (default) Start the gateway via systemd: openclaw gateway start Observe the gateway reach "ready" state, then exit ~9 seconds after startup Check journalctl: journalctl --user -u openclaw-gateway.service -n 50 --no-pager Observe repeated cycle: bonjour stuck in mDNS probing state → CIAO PROBING CANCELLED unhandled rejection → exit code 1 → systemd restart → repeat Expected behavior Gateway starts and remains running. If mDNS probing fails or is interrupted, the bonjour plugin should handle the error gracefully without crashing the gateway process.

Actual behavior On startup, the gateway exited and entered a systemd restart loop after the bonjour/mDNS plugin threw CIAO PROBING CANCELLED during local network advertisement/probing; disabling plugins.entries.bonjour.enabled stopped the crash loop and the gateway then stayed up.

OpenClaw version 2026.4.24 (cbcfdf6)

Operating system Ubuntu 24.04.3 LTS (Noble Numbat) — kernel 6.8.0-88-generic x86_64

Install method npm global (fnm-managed Node 22.22.2)

Model N/A

Provider / routing chain N/A

Additional provider/model setup details No response

Logs, screenshots, and evidence Copy journalctl output showing the crash loop: Apr 26 14:58:15 <hostname> node[1465876]: 2026-04-26T14:58:15.084-06:00 [plugins] bonjour: restarting advertiser (service stuck in probing for 8718ms (gateway fqdn=<hostname> (OpenClaw)._openclaw-gw._tcp.local. host=openclaw.local. port=18789 state=probing)) Apr 26 14:58:15 <hostname> node[1465876]: 2026-04-26T14:58:15.092-06:00 [plugins] bonjour: advertised gateway fqdn=<hostname> (OpenClaw)._openclaw-gw._tcp.local. host=openclaw.local. port=18789 state=unannounced Apr 26 14:58:15 <hostname> node[1465876]: 2026-04-26T14:58:15.094-06:00 [openclaw] Unhandled promise rejection: CIAO PROBING CANCELLED Apr 26 14:58:15 <hostname> node[1465876]: 2026-04-26T14:58:15.097-06:00 [openclaw] wrote stability bundle: /home/<user>/.openclaw/logs/stability/openclaw-stability-2026-04-26T20-58-15-096Z-1465876-unhandled_rejection.json Apr 26 14:58:15 <hostname> systemd[610541]: openclaw-gateway.service: Main process exited, code=exited, status=1/FAILURE Apr 26 14:58:15 <hostname> systemd[610541]: openclaw-gateway.service: Failed with result 'exit-code'. Apr 26 14:58:20 <hostname> systemd[610541]: openclaw-gateway.service: Scheduled restart job, restart counter is at 2.

Leaked mDNS sockets after multiple crash cycles (ss -tulpn | grep 5353): udp UNCONN 0 0 0.0.0.0:5353 0.0.0.0:* users:(("openclaw-gatewa",pid=1469750,fd=43)) udp UNCONN 0 0 0.0.0.0:5353 0.0.0.0:* users:(("openclaw-gatewa",pid=1469750,fd=42)) ... (12 sockets total, all owned by the same gateway pid)

Fix that resolved the issue: openclaw config set plugins.entries.bonjour.enabled false openclaw gateway stop && openclaw gateway start Gateway remained stable after bonjour was disabled. Impact and severity Affected: Any OpenClaw gateway host where the bonjour plugin is enabled (default) and mDNS probing fails or is cancelled — confirmed on Ubuntu 24.04.3 LTS Severity: Critical — gateway cannot stay running; all agent and channel functionality is unavailable Frequency: Always — every startup attempt failed identically, confirmed across multiple restart cycles Consequence: Gateway is completely non-functional until bonjour is manually disabled; additionally, each crash cycle leaks mDNS UDP sockets (12 observed on port 5353 after several restarts) Workaround: openclaw config set plugins.entries.bonjour.enabled false, then restart the gateway

Additional information No response