Background
I’m relatively new to TrueNAS/HexOS but have been running this same hardware as a Windows media server for several years. I recently migrated to HexOS (TrueNAS Scale 25.10.3.1) and shortly after the migration started experiencing random silent reboots with no kernel panic and no log entries at the point of failure.
Hardware
-
Motherboard: ASUS ROG STRIX B450-F Gaming
-
CPU: AMD Ryzen 7 3700X
-
RAM: 16GB DDR4
-
GPU: NVIDIA RTX 2080 Super
-
PSU: EVGA G3 750W, ECO mode OFF
-
OS drive: Samsung PM871 256GB SATA SSD
-
Storage pool: 3× Seagate IronWolf 4TB in RAIDZ1 (
/mnt/HDDs) -
Cache pool: Kingston KC3000 1TB NVMe, single stripe (
/mnt/SSDs) -
Remote KVM: Sipeed NanoKVM, connected via USB to the server
Software
-
HexOS (TrueNAS Scale 25.10.3.1), kernel 6.12.33
-
Docker containers managed via Dockge (Jellyfin, Sonarr, Radarr, qBittorrent behind gluetun VPN, Tdarr transcoding, etc.)
-
NVIDIA drivers active, GPU used for transcoding via NVENC
The Problem
The server reboots silently and repeatedly — up to 7+ times in a single night. last reboot from this morning:
Wed Jun 3 08:10 still running
Wed Jun 3 07:59
Wed Jun 3 04:59
Wed Jun 3 04:08
Wed Jun 3 02:31
Wed Jun 3 02:16
Wed Jun 3 00:49
Wed Jun 3 00:15 - 00:16 (lasted 1 min)
Tue Jun 2 22:57 - 23:12 (lasted 15 mins)
Tue Jun 2 22:48 - 23:12 (lasted 23 mins)
Some boots last only one minute before rebooting again. The reboots appear to happen mostly at idle — overnight with no active transcoding or streaming jobs.
What the logs show
Every time, the kernel journal cuts off cleanly mid-operation with no panic, no watchdog trigger, no OOM killer, no thermal event. The last entries are always routine — Docker bridge networking coming up, containers starting. Then nothing. Example from the most recent prior boot:
Jun 03 08:00:25 kernel: br-54c583a9967c: port 15(veth8e9dde6) entered forwarding state
Jun 03 08:00:25 kernel: eth0: renamed from vethe05b07d
[journal ends]
The watchdog IS configured (kernel.nmi_watchdog=1, kernel.hung_task_panic=1) via Pre-Init script, but has never fired. The kernel simply stops writing as if power was cut.
What we’ve ruled out
ZFS ARC memory exhaustion — ZFS ARC was consuming ~14.5GB of the 16GB RAM, starving the NVIDIA GSP firmware and causing NV_ERR_NO_MEMORY. Fixed by capping ARC at 6GB via Pre-Init script writing to /sys/module/zfs/parameters/zfs_arc_max. This resolved the GPU driver errors but reboots continued.
RAM — Memtest86 completed two full passes overnight, zero errors.
PSU under load — tested with a multimeter at the Molex connector while running a Tdarr GPU transcoding job. +12V rail measured steady at 12.25V. Within spec.
Thermals — sysmon logging temps every 60 seconds. CPU sits at 41-44°C, drives at 34-47°C, no thermal events recorded at any point.
Watchtower — the auto-update container reinitialises on every reboot but never actually ran its 3am scheduled job, confirming it’s not the trigger.
I have been using Claude to try and diagnose as I am new to TrueNAS but there’s been no noticeable change in behaviour