Testing on real hardware

Unit tests cover logic against a fake sysfs. Anything involving OMV, drivers, fans or LEDs must also be checked on a NAS. These are the protocols that worked on the reference TS-253D.

Access

• SSH as root with a key: export NAS=root@NAS_HOST. Agents must not type passwords; a person logs in, or keys are set up beforehand.
• The OMV web UI runs on the port set in OMV (the reference box uses SSL only on a non-default port); see ui-validation.md.

Deploy

make test && make docker-deb && make deploy NAS=$NAS

tools/nas-deploy.sh copies the .deb, checks its sha256 and runs apt-get install --reinstall. Then check:

ssh $NAS 'systemctl is-active qnapd; journalctl -u qnapd -n 20 --no-pager; qnapd status; qnapd info'

Every install restarts qnapd (chassis fan briefly to full speed) and OMV’s engined (web UI sessions are logged out).

Safety rules

• Ask the owner before deliberately slowing or stalling fans, killing the daemon, rebooting, toggling reboot=efi, or changing EC power settings. Tell them what they will hear or see.
• While holding fans by hand the daemon’s critical protection is off: use the scripts below, which abort on CPU 85 °C, NVMe/QM2 70 °C, EC 50 °C.
• Never read drivetemp or run smartctl without -n standby on HDDs.
• Stop qnapd before writing PWM by hand (systemctl stop qnapd), and start it again afterwards; it otherwise overwrites your values.
• Close the OMV dashboard while measuring: its widgets load the CPU.

Protocols

LED mapping (needs a person looking at the box)

Announce each step before it starts; 5 seconds per LED is enough.

scp tools/hw/* $NAS:/root/
ssh $NAS '/root/led-test.sh qnap8528::hdd1 5'       # which bay lit?
ssh $NAS '/root/led-test.sh qm2-1:green:slot1 5'    # which slot lit? (rear bracket)
ssh $NAS 'timeout 5 dd if=/dev/sda of=/dev/null bs=1M iflag=direct'   # which bay blinks?

Buttons (needs a person at the box)

ssh $NAS 'python3 /root/button-test.py 20'   # press one button per run

Prints each press and release with the hold time. One button per run avoids guessing which press was which. Never press the power button (ACPI: it shuts the NAS down). The Reset button is a rear pinhole.

Fan PWM → RPM sweep

ssh $NAS 'systemctl stop qnapd; /root/fan-sweep.sh qnap8528 15; /root/fan-sweep.sh qm2 15; systemctl start qnapd'

Prints RPM per PWM going down, then up until the fan restarts. Floors come from the restart PWM plus margin.

Noise check (needs a person at the box)

ssh $NAS 'systemctl stop qnapd; /root/fan-hold.sh 48 64 30; /root/fan-hold.sh 64 64 30; systemctl start qnapd'

Arguments: chassis PWM, QM2 PWM, seconds. Announce each step first.

Failsafe

ssh $NAS 'kill -9 $(pidof -x qnapd)' → chassis 255, QM2 255, status LED blinking red; systemd restarts the daemon after 10 s (the LED then goes green). A clean systemctl stop qnapd → chassis 255, QM2 back to the card (≈1260 RPM on a QM2-2P-244A-A1).

Disk standby

Remove anything else polling disks first (e.g. tempmon sensors that run smartctl without -n standby). Put a disk to sleep by hand rather than waiting for spindown: smartctl -s standby,now /dev/sdX. Check its state without waking it: smartctl -n standby -i /dev/sdX exits 2 in standby. Watch it across three diskintervals: it must stay in standby, qnapd status must skip its temperature, its bay LED must stay solid green.

Also check that idle disks fall asleep on their own: with OMV’s spindown set, no I/O (/sys/block/sdX/stat counters flat) and no smartctl from anything else, the disk must be in standby one spindown time after its last I/O. Every smartctl call on an awake disk, including the -n standby -i check itself, restarts the timer: check once, at the end. qnapd before 8.0.1 failed this by querying idle disks every minute.

If it wakes, find out why before blaming qnapd:

• /sys/block/sdX/stat read/write counters: S.M.A.R.T. queries do not count as I/O, so moving counters mean a real read or write.
• journalctl -u smartmontools: OMV’s smartd with Power mode “Never” (-n never in /etc/smartd.conf) checks sleeping disks every 30 min and wakes them. Set Storage › S.M.A.R.T. › Settings › Power mode to Standby for spindown to work.
• zpool iostat <pool> 10: pool I/O.
• OMV 8 applies APM and spindown without hdparm: the smartctl-hdparm service runs /etc/smartmontools/hdparm.d/* (smartctl --set=apm,N --set=standby,N) at boot and on Apply. Check journalctl -u smartctl-hdparm -b for Standby timer set to …. Exit code 4 with “AAM disable failed” is harmless (the disk has no AAM).

Power loss / boot

After a reboot: qnapd active with NRestarts=0, journal shows control within seconds of boot, ledtrig_timer loaded, LEDs green, pools healthy.

Reporting

Record measured values with model, date and conditions in ../hardware/ts-253d.md (or a new file per model) and tick PLAN.md. Owners of other models use tools/model-test.py instead; see model-testing.md. No IPs, host names or serial numbers in the repo.