openclaw gateway restart creates orphaned gateway process on macOS (launchd)
Running openclaw gateway restart on macOS with the gateway managed by launchd (ai.openclaw.gateway LaunchAgent) causes a race condition that leaves the system in a thrashing state.
What happens:
The restart command appears to both signal launchd to spawn a new gateway AND directly spawn its own gateway process. Whichever wins the port lock (18789) becomes the listener — but launchd is tracking its own spawn, not the CLI-spawned one. Result: launchd's process fails with "gateway already running / port in use" and retries every ~10 seconds indefinitely.
Symptoms:
• gateway.err.log floods with Gateway failed to start: gateway already running (pid XXXXX); lock timeout after 5000ms every ~10s
• launchctl print gui/$UID/ai.openclaw.gateway shows a different PID than the actual listener on port 18789
• runs counter in launchd climbs continuously
Repro:
1. Install openclaw with launchd service (openclaw gateway install)
2. With gateway running healthy under launchd, run openclaw gateway restart
3. Thrash begins immediately
Workaround:
Kill the orphan (lsof to find the port owner, kill <pid>) — launchd's next spawn will then win the lock and stabilize. Or use launchctl directly:
launchctl stop gui/$UID/ai.openclaw.gateway
launchctl start gui/$UID/ai.openclaw.gateway
Expected behavior: openclaw gateway restart should use launchctl kickstart -k exclusively (kill existing + start fresh atomically) and not directly spawn a gateway process when running under launchd.
Version: 2026.2.19-2 · macOS arm64
What happens:
The restart command appears to both signal launchd to spawn a new gateway AND directly spawn its own gateway process. Whichever wins the port lock (18789) becomes the listener — but launchd is tracking its own spawn, not the CLI-spawned one. Result: launchd's process fails with "gateway already running / port in use" and retries every ~10 seconds indefinitely.
Symptoms:
• gateway.err.log floods with Gateway failed to start: gateway already running (pid XXXXX); lock timeout after 5000ms every ~10s
• launchctl print gui/$UID/ai.openclaw.gateway shows a different PID than the actual listener on port 18789
• runs counter in launchd climbs continuously
Repro:
1. Install openclaw with launchd service (openclaw gateway install)
2. With gateway running healthy under launchd, run openclaw gateway restart
3. Thrash begins immediately
Workaround:
Kill the orphan (lsof to find the port owner, kill <pid>) — launchd's next spawn will then win the lock and stabilize. Or use launchctl directly:
launchctl stop gui/$UID/ai.openclaw.gateway
launchctl start gui/$UID/ai.openclaw.gateway
Expected behavior: openclaw gateway restart should use launchctl kickstart -k exclusively (kill existing + start fresh atomically) and not directly spawn a gateway process when running under launchd.
Version: 2026.2.19-2 · macOS arm64
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
134,827Members
Resources
Similar Threads
Was this page helpful?
