Runner Provision Fails

Symptoms

• Runner status stuck in provisioning state
• Dashboard shows “Failed to start runner” error
• Container never appears in Docker

Likely Causes

1. Docker daemon unreachable — Docker socket not mounted or daemon stopped
2. Image pull failure — Container registry unreachable or image tag missing
3. Resource limits — Host out of memory, disk, or CPU
4. Port conflicts — VNC port 5900 or gateway port 18789 already bound
5. Environment variable errors — Invalid or missing required env vars

Triage Steps

1. Check Docker daemon

docker info
docker ps -a | grep openclaw

2. Check image availability

docker images | grep curate-me
docker pull ghcr.io/curate-me-ai/openclaw-base:latest

3. Check resource usage

docker stats --no-stream
df -h  # Disk space
free -m  # Memory

4. Check runner logs

./scripts/errors by-source gateway | grep "runner_provision"
docker logs <container_id> 2>&1 | tail -50

Resolution

Docker daemon restart

sudo systemctl restart docker

Clear stuck runners

# Mark stuck provisioning runners as failed
curl -X POST https://api.curate-me.ai/gateway/admin/runners/cleanup \
  -H "Authorization: Bearer $JWT"

Resource cleanup

docker system prune -f  # Remove unused containers/images

Escalation

If the Docker daemon is consistently failing, check host kernel logs (dmesg) for OOM kills.