25.10.4 hard reboot when selecting/importing Apps pool

TrueNAS General
, , ,
1

Hi all,

I would like to ask for help with a TrueNAS SCALE 25.10.4 hard reboot / boot loop issue.

At this point I am not sure whether this is a 25.10.x Apps/Docker issue, Apps pool content or metadata issue, GPU driver issue, or a hardware issue. The clearest trigger so far is that when I select/import the existing Apps pool again, the system may hard reboot.

System information:

  • TrueNAS SCALE 25.10.4-Goldeye
  • Previous 25.04.1 boot environment is still available
  • CPU: AMD Ryzen 7 8845HS with Radeon 780M iGPU
  • GPU: NVIDIA GeForce RTX 5060 Ti
  • Boot device: 240GB SATA SSD
  • Data pool: 8 x 12TB HDD RAIDZ, currently degraded with one failed/offline disk, but still within redundancy
  • Apps pool: 2 x 500GB NVMe SSD mirror
  • Both NVMe SSDs in the Apps pool passed SMART checks and look normal

Timeline:

  1. The system was originally running 25.04.x without issues.
  2. I later moved the NAS from a smaller case to a larger case and improved airflow. No issues after that.
  3. I then installed an NVIDIA RTX 5060 Ti. No immediate issues.
  4. Because 25.04.x did not properly support the RTX 5060 Ti / NVIDIA driver, I upgraded to 25.10.4.
  5. After the upgrade, everything initially looked normal, and the NVIDIA driver appeared to work.
  6. I later installed and tested several local AI related Apps / containers, such as Ollama, Open WebUI, LocalAI, and Hermes Agent.
  7. After trying to let a local model process a large amount of data, the system hard rebooted.
  8. Since then, 25.10.4 reaches the TrueNAS console menu or Web UI for only a few seconds, then hard reboots again, forming a boot loop.
  9. If I select 25.04.1 from GRUB, the system can boot, but most Apps show as crashed.
  10. If I unset the Apps pool while in 25.04.1, the system can run stably.
  11. However, as soon as I choose/import the original Apps pool again, the system hard reboots.
  12. I later removed Ollama / Open WebUI and deleted some container images, but selecting the Apps pool again can still trigger a hard reboot.

Current observations:

  • 25.04.1 is stable when the Apps pool is unset / not configured.
  • 25.10.4 hard reboots shortly after the Apps pool is enabled.
  • The Apps pool is a 2 x 500GB NVMe SSD mirror, and both SSDs passed SMART checks.
  • The data pool is degraded, but it does not appear to be the direct trigger.
  • I have seen amdgpu-related messages in boot logs / dmesg, such as ATPX Hybrid Graphics and amdgpu timeout messages.
  • I also tested booting with nomodeset modprobe.blacklist=amdgpu from GRUB and disabled the iGPU in BIOS. In that state, I confirmed that amdgpu was not loaded.
  • Even with amdgpu not loaded, selecting the Apps pool again still caused a hard reboot.
  • The NVIDIA RTX 5060 Ti was one reason for upgrading to 25.10, but I have not seen a clear NVIDIA driver error so far. The visible driver-related errors I have seen were mainly amdgpu-related.

I found a few forum threads that look somewhat similar:

  • TrueNAS SCALE System hard reboots when importing “apps” pool (NVMe mirror), pool otherwise healthy
  • Corrupt file puts system into hard crash boot loop via docker
  • 25.10.2.1 - Goldeye Docker NVIDIA Driver issues
  • After 25.10 upgrade, “Critical Failed to start docker for Applications: Docker service could not be started”

My question:

Is this kind of hard reboot after selecting/importing an Apps pool a known issue on 25.10.x?

At this point I am trying to determine which direction is more likely:

  1. A TrueNAS SCALE 25.10.x Apps / Docker / ZFS import bug
  2. Docker metadata, app volumes, or some file content inside the Apps pool triggering a kernel-level crash
  3. amdgpu / hybrid graphics related issue
  4. NVIDIA runtime / GPU app metadata issue
  5. Or still a hardware issue that happens to be triggered by Apps pool activity

Because this is a hard reboot, not just an app crash, and there is almost no clear crash log, I am hesitant to repeatedly test writable imports of the Apps pool.

Has anyone seen a similar issue, or does anyone have suggestions on which direction I should investigate first?