VSaaS Edge GPU recovery -- diagnostic summary
=============================================
Collected      : Fri Aug 21 09:09:29 PM UTC 2026
Host           : soterix
Bundle         : /home/soterix/soterix_vsaas_edge_nvid
Kernel         : 6.8.0-138-generic
Script         : recover_nexaiq_edge_gpu.sh

FINDINGS
--------
GPU present on PCI bus  : true
GPU fallen off bus      : false
nvidia-smi state        : error
Driver reported by smi  : <none>
Kernel module (loaded)  : 575.51.03
Kernel module (on disk) : 575.51.03
Target driver version   : 575.51.03
DKMS built for kernel   : true
Secure Boot enabled     : false
Xid codes seen          : 119 
GSP firmware fault      : true
Module state corrupted  : false
D-state pids (in driver): <none>
Leaked probe pids       : <none>
Graceful reboot hangs   : false
Reboot required         : false
Cold power cycle needed : true

WHERE TO LOOK
-------------
kernel/xid_history.txt        Xid codes across this and previous boots
kernel/module_refcounts.txt   negative refcnt => module can never be unloaded
system/dstate_stacks.txt      kernel stack of every unkillable task
journal/prev_boot_kernel.txt  what the box logged before the last power cycle
journal/shutdown.txt          why a graceful shutdown did not complete
gpu/SKIPPED_nvidia-smi.txt    present when GPU collectors were skipped, and why
../gpu_recovery_*_trace.log   every guarded command with rc and duration