#Subagent with Openai Codex Harness is freezing

1 messages · Page 1 of 1 (latest)

glacial pollen
#

I have a "coder" agent that I have assigned to use the codex runtime, but it is freezing, and then when I go to the CLI, openclaw ... is also frozen. Seems something is stuck. If I cannot run openclaw commands, how do I troubleshoot this?

peak elbowBOT
#

If openclaw ... itself is hanging, treat the CLI as part of the symptom and bypass it first.

Start with read-only OS-level checks:

# don’t let CLI commands hang forever
timeout 10s openclaw --version
timeout 10s openclaw status

# if those hang, read the gateway file log directly
ls -lt /tmp/openclaw/openclaw-*.log | head
tail -n 200 /tmp/openclaw/openclaw-*.log

# see what is stuck/running
pgrep -af 'openclaw|codex|node'
ps -eo pid,ppid,stat,etime,pcpu,pmem,command | grep -E 'openclaw|codex|node' | grep -v grep
#

If you installed the gateway as a service, also check service logs without using the OpenClaw CLI:

Linux/systemd:

systemctl --user status openclaw-gateway --no-pager
journalctl --user -u openclaw-gateway -n 200 --no-pager

macOS/launchd:

launchctl print gui/$UID | grep -i openclaw
ls -lt ~/Library/Logs | grep -i openclaw

If the gateway is wedged, the practical recovery step is to restart the gateway process/service from the OS side, then immediately inspect /tmp/openclaw/openclaw-*.log for the first error after restart. If you see a stuck codex child process, note its PID before killing it so the log/process tree is useful.

For the report, please include: