NVIDIA DGX Spark stuck on NVIDIA logo, recovery shows error

[Happy-Celine]

My NVIDIA DGX Spark is unable to boot into the system. After powering on, it stays stuck on the NVIDIA logo and never proceeds further. When I try recovery mode, it shows an error.

Has anyone experienced this issue or have any suggestions on how to fix it?

[aniculescu]

After seeing the NVIDIA logo, are you able to enter BIOS? Press and hold the ESC or DEL key while system is booting.
If you can enter BIOS, I would recommend trying system recovery System Recovery — DGX Spark User Guide

[Happy-Celine]

Thank you for your response. The error message in the second image above was displayed during the reboot process after I performed a reset using the latest recovery tool from the official website. The Spark can still boot into the BIOS normally.

[aniculescu]

Thanks for the info. Can you also perform an NVMe Self-Test through the BIOS and share the result?

[Happy-Celine]

Sure! The first image below shows the NVMe SSD diagnostic test in the BIOS, which passed. Afterward, I rebooted the Spark, and it booted into the OS normally. Strangely, though, after a few minutes, the screen started getting spammed with errors again (see the second image below).

[Neill]

Thanks @Happy-Celine. The screenshots show systemd-journald failing to write to /var/log/journal with Input/output error. That means the OS loses the ability to write to its disk a few minutes after boot, which points to storage, even though the BIOS NVMe test passed.

Could you share:

1. Did the recovery tool report that it completed successfully, or did it show any errors during the reimage?

2. The output of the commands below, run before the error spam starts (or from Ctrl+Alt+F3, or over SSH):

   sudo dmesg | grep -iE "nvme|ext4|i/o error|blk_update|critical"
   sudo nvme smart-log /dev/nvme0
   

3. Is this a new unit, or did it work before? Did anything happen just before the first failure, such as an update or a power loss?