EGL for rendering without X server errors

Hello there,

I’m trying to use EGL do OpenGL rendering without an X server on a Centos 7 machine that has 2 Tesla K20Xm’s, but when I run the code provided in that blog post I get a couple errors I can’t seem to track down.

(base) [baylyd@gpu2 ~]$ ./egl 
libEGL warning: DRI3: failed to query the version
libEGL warning: DRI2: failed to authenticate

It would help me greatly if anyone knew if this was an issue with the way the driver was installed, or something else.

Another program that I’m trying to use has issues also. I am attempting to use VirtualGL’s 3.0 pre-release with the EGL backend, but I get stuck at

vglrun -d /dev/dri/card1 glxinfo
name of display: localhost:10.0
[VGL] ERROR: in init3D--
[VGL]    222: Could not initialize EGL
(base) [baylyd@gpu2 ~]$ 

So something is not setup correctly, but I’m not sure what. This was the command we ran to install the drivers

sh cuda_11.0.3_450.51.06_linux.run --silent --driver 

Does any of this explain why the EGL errors are showing up?