I've been rendering on a Linux VM for a while, using my own Blender build, built headless. Now I want to use a GPU VM, and I'm having trouble building my own CUDA build, so I tried the nightly (which isn't headless of course). When I start it up on my VM, I get an error "Unable to open a display" even though I'm running with --background.

root@fa2823c96639:/app# blender --debug --background -f 1 simple-test-scene.blend
Switching to fully guarded memory allocator.
Blender 2.91.0 Alpha
Build: 2020-08-23 23:11:51 Linux Release
argv[0] = blender
argv[1] = --debug
argv[2] = --background
argv[3] = -f
argv[4] = 1
argv[5] = simple-test-scene.blend
Blender 2.91.0 Alpha (hash e4932d1167f4 built 2020-08-23 23:11:51)
read file
  Version 280 sub 39 date unknown hash unknown
found bundled python: /app/blender/blender-2.91.0-e4932d1167f4-linux64/2.91/python
Blender startup at 2020-08-24 14:08:56.580353: 4 cores
Unable to open a display
Aborted (core dumped)

The blendfile I'm using is set up for Cycles with GPU Compute.

What am I doing wrong? Alternatively, is there a nightly headless blender build generated somewhere?

  • $\begingroup$ Are you connecting to the VM through SSH? $\endgroup$
    – Robert Gützkow
    Aug 24, 2020 at 14:27
  • $\begingroup$ Yes, though in production it'll be driven by an express app. $\endgroup$
    – GaryO
    Aug 24, 2020 at 14:30
  • $\begingroup$ This Q&A could help you. $\endgroup$
    – Robert Gützkow
    Aug 24, 2020 at 15:26
  • $\begingroup$ I did look at that (super complicated) example. But both of his "solutions" involve an X server -- either forwarding to one, or running a local one. I'm interested in running it truly headless. Maybe the full build just can't work that way? $\endgroup$
    – GaryO
    Aug 24, 2020 at 16:45
  • $\begingroup$ OK, looks like I'll have to figure out how to build nightlies/headless with CUDA. $\endgroup$
    – GaryO
    Aug 24, 2020 at 17:14


You must log in to answer this question.

Browse other questions tagged .