Troubleshooting
Agent Connectivity
Section titled “Agent Connectivity””Connection Refused”
Section titled “”Connection Refused””- Cause: The agent cannot reach the server IP/Port.
- Fix:
- Ensure the server container is running (
docker compose ps). - Check if the server firewall allows port
3143(or yourPANKHA_PORT). - Verify the
server_urlinconfig.jsonuses the correct IP (notlocalhostif on a different machine).
- Ensure the server container is running (
”WebSocket Handshake Failed”
Section titled “”WebSocket Handshake Failed””- Cause: Protocol mismatch or proxy issue.
- Fix: Ensure your URL starts with
ws://. If you are running behind a reverse proxy (like Nginx Proxy Manager), ensure that WebSocket support is enabled.
Agent Disconnects After Backend Restart
Section titled “Agent Disconnects After Backend Restart”- Info: Agents have automatic reconnection with exponential backoff. They will retry indefinitely until the backend is available again. If you need to force a reconnect:
Terminal window # Linux./pankha-agent --restart# Windows (via Tray Icon)Right-click Tray -> Restart Service
Agent Stays Stopped After Every Reboot
Section titled “Agent Stays Stopped After Every Reboot”- Cause: The systemd service was never installed - nothing starts the agent at boot.
- Fix:
sudo ./pankha-agent --install-service(see Linux Agent). Deployment Center installs it automatically; the setup wizard asks - answer yes.
Accounts & Approval
Section titled “Accounts & Approval”Agent Stuck on “Pending Approval”
Section titled “Agent Stuck on “Pending Approval””- Info: This is by design - machines without credentials wait for an admin. Click Approve on its card (see Deployment Center).
- If the button says to stage binaries first: the agent is on an old version and updates as part of the approval - download the latest release in Deployment → Release & Hub cache, then approve again.
Known System Reappeared Asking for Approval
Section titled “Known System Reappeared Asking for Approval”- Cause: The agent lost its stored token - usually an OS reinstall, wiped config, or restored backup.
- Fix: Approve it. A fresh token is issued; the system keeps its history, names, and fan profiles.
Lost the Admin Password
Section titled “Lost the Admin Password”- Fix: Set
PANKHA_AUTH_RESET=truein.envand restart the server - all accounts are removed (your systems and data stay) and the first-run setup screen returns. Remove the variable immediately after, or every restart wipes accounts again. Only the exact valuetruetriggers it.
Server Issues
Section titled “Server Issues”Port Already in Use
Section titled “Port Already in Use”- Error: “Port 3143 is already allocated”
- Fix:
- Change port in
.env:Terminal window PANKHA_PORT=7000 - Restart containers:
Terminal window docker compose down && docker compose up -d - Important: agents connect to this port too - update each agent’s server URL (Linux:
sudo ./pankha-agent --setup; Windows: tray Configure…).
- Change port in
Database Connection Failed
Section titled “Database Connection Failed”- Cause: PostgreSQL container not running or misconfigured.
- Diagnosis:
Terminal window docker compose ps # Check containers runningdocker compose logs pankha-postgres # View PostgreSQL logs - Last resort - reset the database (deletes all history, profiles, and settings): stop the stack (
docker compose down), deletedocker-data/backend/database/postgres_data/, thendocker compose up -d. Note thatdocker compose down -vdoes not reset it - the data lives in that folder, not in a Docker volume (Server Configuration).
Dashboard Slow/Unresponsive
Section titled “Dashboard Slow/Unresponsive”- Cause: Database needs maintenance after months of sensor data.
- Fix: Vacuum the database (use your
POSTGRES_USERandPOSTGRES_DBfrom.env):Also consider lowering Data Retention in Settings.Terminal window docker compose exec pankha-postgres psql -U <user> -d db_pankha -c "VACUUM ANALYZE;"
Hardware Issues
Section titled “Hardware Issues”No Sensors/Fans Detected (Linux)
Section titled “No Sensors/Fans Detected (Linux)”- Cause: Missing kernel modules.
- Fix: Run
sensors-detect(fromlm-sensorspackage) and load the recommended kernel modules. Reboot and retry.
No Sensors/Fans Detected (Windows)
Section titled “No Sensors/Fans Detected (Windows)”- Cause: PawnIO driver missing, permission issue, or unsupported hardware.
- Fix:
- Verify the PawnIO driver is installed - see Windows Agent troubleshooting.
- Check
logs\pankha-agent.login the install directory for specific “Access Denied” errors. - Note that some anti-cheat software blocks hardware access entirely.
Fan Control Not Working (Linux)
Section titled “Fan Control Not Working (Linux)”- Cause: Hardware doesn’t support PWM or requires manual enable.
- Diagnosis:
-
Check for PWM control files:
Terminal window ls -la /sys/class/hwmon/*/pwm*If no
pwm*files exist, your fans may not support speed control. -
Test manual control (this will change fan speed):
Terminal window # Set to ~50% speed (128 out of 255)echo 128 | sudo tee /sys/class/hwmon/hwmon0/pwm1Monitor temperatures after running. Replace paths with your actual hwmon device.
-
Some systems require enabling PWM mode first:
Terminal window cat /sys/class/hwmon/hwmon0/pwm1_enable# Values: 0=disabled, 1=manual, 2=auto(BIOS)# To enable manual mode:echo 1 | sudo tee /sys/class/hwmon/hwmon0/pwm1_enableRebooting restores BIOS defaults.
-
Agent Shows “Disconnected” but Backend is Running
Section titled “Agent Shows “Disconnected” but Backend is Running”- Fix:
- Check backend logs:
docker compose logs pankha-app --tail=100 - Verify agent can reach backend:
curl http://<server-ip>:3143/health - Check agent config.json has correct
server_url - Restart agent:
./pankha-agent --restart
- Check backend logs:
Calibration & Fan Health
Section titled “Calibration & Fan Health”What do the fan badges mean?
Section titled “What do the fan badges mean?”| Badge | Meaning |
|---|---|
| Calibrating | A calibration run is in progress - manual control is locked until it completes |
| Stalled | The fan was commanded to spin but reports 0 RPM - stuck, disconnected, or needing recalibration |
| Attention / Check fan | The health engine flagged this fan - click the info icon on the fan row for the details |
Calibration failed or seems stuck
Section titled “Calibration failed or seems stuck”- Failed: calibration stops early on purpose if temperatures rise too far during the sweep - that’s the safety working. Click the calibrate icon to retry when the system is idler.
- Stuck on “Calibrating”: a run sweeps the fan through its full range including brief stops, so give it a few minutes. The calibrate icon’s tooltip always states the current calibration state.
- A fan that never calibrates may not report its speed at all (no tachometer) - such fans can still be controlled, just not measured. See Fan Calibration & Health.
Virtual sensor shows no value
Section titled “Virtual sensor shows no value”- Cause: its member sensors aren’t reporting - usually the agent is offline, or members were hidden/removed since the sensor was built.
- Fix: check the system is online, then open the sensor’s edit view (pencil icon) and re-check its members (Dashboard).
IPMI Agents
Section titled “IPMI Agents”Error badge right after deployment
Section titled “Error badge right after deployment”- Cause: usually a BMC profile mismatch - commands meant for another vendor make the BMC answer
Invalid command(0xc1), and sensor discovery fails. - Fix: no need to touch the server - assign the correct profile from the card’s BMC section; the agent fetches and hot-reloads it, and the badge clears when telemetry resumes (IPMI Agent).
Could not open device / no /dev/ipmi0
Section titled “Could not open device / no /dev/ipmi0”- Fix: load the kernel IPMI modules:
modprobe ipmi_devintf ipmi_si. Also confirmipmitoolis installed - the install script does not install it.
Usage Questions
Section titled “Usage Questions”Can I control GPU fans?
Section titled “Can I control GPU fans?”- Linux: NVIDIA - yes, via the NVIDIA driver; the GPU appears as an extra sensor and controllable fan automatically. Other GPUs work when they expose standard
sysfsPWM (common withamdgpu). - Windows: NVIDIA - yes, via NvAPI. Other GPUs are monitor-only in most cases.
Why do my fans spin up on boot?
Section titled “Why do my fans spin up on boot?”- Reason: Before the agent starts, the BIOS/UEFI controls the fans.
- Fix: Configure a silent curve in your BIOS as a fallback, so the fans stay quiet until the OS/Agent loads and takes over.
CLI Quick Reference
Section titled “CLI Quick Reference”Linux Agent
Section titled “Linux Agent”# Setup & Service./pankha-agent -e # Run interactive setup wizard./pankha-agent -I # Install systemd service./pankha-agent -U # Uninstall systemd service
# Daemon Control./pankha-agent -s # Start daemon./pankha-agent -x # Stop daemon./pankha-agent -r # Restart daemon
# Status & Logs./pankha-agent -i # Show status./pankha-agent -l # View logs (live)./pankha-agent --log-level INFO # Set log level
# Config & Debug./pankha-agent -c # Show config./pankha-agent --check # Run health check./pankha-agent --test # Test mode (hardware discovery)The IPMI agent shares this CLI, plus --profile <path> and --dry-run.
Windows Agent (run as Administrator)
Section titled “Windows Agent (run as Administrator)”pankha-agent.exe --start # Start servicepankha-agent.exe --stop # Stop servicepankha-agent.exe --restart # Restart servicepankha-agent.exe --status # Show statuspankha-agent.exe --setup # Interactive setup wizardpankha-agent.exe --config-show # Show configpankha-agent.exe --logs follow # View logs (live)pankha-agent.exe --logs 50 # View last 50 log linespankha-agent.exe --test # Hardware discovery testNext Steps
Section titled “Next Steps”- Settings: the Diagnostics tab exports hardware reports to attach to bug reports.
- Still stuck? Open an issue - the Diagnostics tab’s Report + Diagnostics button pre-fills one for you.