System:

• Motherboard: ASUS ROG STRIX B550-F GAMING, BIOS version 2806 (10/27/2022)

• CPU: AMD (k10temp detected)

• RAM: ECC

• PSU: Corsair RM550x (RMx Series), 550W, 80+ Gold, purchased 01/23/2023

• Unraid version: Kernel 6.18.38-Unraid

• Array: 3 data disks (2x Seagate 10TB, 1x Toshiba 18TB), no parity drive (parity not present)

• Cache: Samsung 970 EVO Plus 2TB NVMe

• USB devices: boot stick (SanDisk Ultra), ConBee II (Zigbee, connected via powered USB extension cable per manufacturer recommendation), Raspberry Pi (RaspiMatic/HomeMatic), wireless USB keyboard

Problem:
Since this morning, repeated hard system crashes – 5 so far with automatic reboot afterward, 1 complete freeze with no automatic reboot (system still showed the login prompt but was unresponsive to any input, including the USB keyboard – no reaction, no Num Lock LED).

Troubleshooting steps taken so far:

1. Enabled local syslog mirroring (Settings → Syslog Server → Mirror to boot drive) – didn't help, since a hard crash gives no time to flush, and some crashes happened even before the OS fully started.

2. Set up remote syslog via UDP to an external machine (Syslog Watcher), plus an artificial heartbeat every 15 sec via logger to pin down the exact crash timestamp.

3. Log analysis across multiple incidents: in every case the log stops abruptly with no warning whatsoever – no kernel panic, no oom-killer, no WHEA/MCE/hardware error, no disk I/O errors. The last log lines are always completely normal operation (Docker network events, NFS, etc.).

4. Controlled minimal-load test: array started, Docker AND both VMs fully stopped/disabled, no background processes (iotop confirmed no meaningful I/O load). Still crashed later — even under essentially zero load.

5. Possibly related observation: the ConBee II (on usb 3-2, cdc_acm/ttyACM0) repeatedly and irregularly disconnects/reconnects, independent of the server crashes, despite being connected via a powered USB extension cable per manufacturer guidelines.

6. USB stick backup: first copy attempt (Windows Explorer, drag & drop) hit a read error, resulting in an incomplete copy (~3MB difference). Second attempt using robocopy /MIR /R:3 /W:2 completed without errors (415/415 files, 0 failures) — suggests a one-off loose contact rather than a physically damaged stick.

7. Security check performed: login history (/var/log/syslog) shows only successful logins from the local LAN, no suspicious/external IPs. An unfamiliar-looking plugin ("dwpython") turned out to have been installed since March 2025, not a new addition. Ports 80/443 are reachable from outside (intended for Nginx Proxy Manager with mTLS), port 22 (SSH) is closed from outside.

Most recent incident (today, ~16:07) – different pattern:
Unlike the previous 5 crashes (silent abort + automatic reboot), this one was a complete freeze with no reboot: CPU fan still spinning, hard drives audibly/physically still running, but the motherboard status LED was off, the screen showed a frozen login prompt, and there was no response to keyboard input (not even the Num Lock LED).

Suspects considered so far, none confirmed:

• PSU aging/failure, despite being relatively young (2.5 years old)

• Unstable USB/board power delivery (supported by the ConBee symptom)

• RAM failure (though ECC should be logging this — no EDAC/MCE messages seen in the log so far)

• No IPMI/BMC available (consumer board), so no independent hardware log source for power events

Question for the community: What further diagnostic steps would you recommend before I try swapping the PSU as a test?

please post diagnostics after crash, syslog server is continually wrote to log file so it doesnt need to flush. you need to make sure it's setup and logging to the share you selected and if you mirror to flash it will be wrote immediately to the flash device.

I've verified that syslog mirroring to the flash drive is correctly configured and working — I checked /boot/logs/syslog-previous directly via SSH after the latest incident, and it writes continuously right up to the crash (no buffering/flush delay).

However, it confirms the same pattern as before: the log simply stops mid-normal-operation with zero error, warning, or panic entries beforehand — even under this most recent freeze (system unresponsive, screen frozen at login prompt, no reaction to keyboard input, but CPU fan and drives still physically running).

Diagnostics ZIP attached, generated immediately after this latest crash.

jarvis-diagnostics-20260830-0926.zip

Start here: https://forums.unraid.net/topic/46802-faq-for-unraid-v6/page/2/#findComment-819173