Plex transcoding with hardware acceleration enabled leads to server crash

Hi,

I have a server with the following specs:

  • Pentium G4600 CPU (with Intel 630 GPU)
  • 16GB ECC RAM
  • Supermicro X11SSH-F µATX board with IPMI

Software:

  • Debian sid
  • Linux kernel 4.14.17
  • kbl_dmc_ver1_01.bin, kbl_guc_ver9_14.bin, kbl_guc_ver9_39.bin, kbl_huc_ver02_00_1810.bin in /lib/firmware/i915 (a warning in dmesg told me to add those)
  • Docker 18.02.0-ce with runc 1.0.0-rc5
  • tried both the official Plex and now the LinuxServer.io container
  • Plex 1.12.0.4829

I added /dev/dri as a device and enabled hardware acceleration in the PMS settings.

It uses the GPU successfully, I can check that with intel_gpu_top in the main OS.

Transcoding a 1080p h264 or a 4K HEVC movie for an iOS device works fine.
But when I use Safari on my Mac and start something to transcode, the entire machine freezes and I get this in the serial console (via the IPMI interface of my server board):

Disabling lock debugging due to kernel taint
mce: [Hardware Error]: CPU 0: Machine Check Exception: 5 Bank 4: ba00000011000402
mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffffb548b605> 
{intel_idle+0x95/0x110}
mce: [Hardware Error]: TSC 2033626b37c2 
mce: [Hardware Error]: PROCESSOR 0:906e9 TIME 1520000339 SOCKET 0 APIC 0 microcode 70
mce: [Hardware Error]: Run the above through 'mcelog --ascii'

# ./mcelog --ascii < file 
Hardware event. This is not a software error.
CPU 0 BANK 4 
RIP !INEXACT! 10:ffffffffb548b605
MISC 0 
STATUS ba00000011000402 MCGSTATUS 5
{intel_idle+0x95/0x110}
Hardware event. This is not a software error.
CPU 0 BANK 0 TSC 2033626b37c2 
TIME 1520000339 Fri Mar  2 15:18:59 2018
MCG status:
MCi status:
Machine check not valid
Corrected error
MCA: No Error
STATUS 0 MCGSTATUS 0
CPUID Vendor Intel Family 6 Model 158
SOCKET 0 APIC 0 microcode 70

I don’t know how to proceed debugging this, so I have started here in the Plex forums.
Has anyone an idea what can cause this?

Kind regards
hanzoh

did u attach a monitor to ur server? try unplug it