It's impossible to say without knowing more.  It could be a bug in
TigerVNC, or the window manager, or VirtualGL, or the nVidia drivers.
Unfortunately, there's no way to know unless you can somehow narrow down
the possibilities for me:

-- Are you running the window manager using VirtualGL, or are you using
TigerVNC's built-in (software) OpenGL implementation to run the WM?  If
you're using the latter, then this isn't a VirtualGL bug, and you should
report it to Cendio.

-- Can you ascertain whether any particular application triggers the
bug, or does it seem to occur irrespective of the applications that have
been run?  I'm asking this to attempt to determine whether the window
manager itself is generating the corrupt pixels or whether they are
somehow triggered by running a 3D application.  If you are running the
WM in VirtualGL, then trying to run a 3D application within the WM, then
I could envision there being a possible conflict between the two.  Note
that if you're running the WM in VirtualGL, you shouldn't use vglrun to
launch 3D applications (because VirtualGL is already preloaded into
everything launched from within the WM.

-- Can you ascertain whether the issue occurs with TurboVNC as well?
Use TurboVNC 2.1 and the -3dwm switch to load the window manager in
VirtualGL.  If that configuration also fails, then it would point to
this being a bug in either VirtualGL or the nVidia drivers.

Also, we generally recommend using MATE rather than GNOME 3.  It's much
more VNC-friendly.


On 9/15/16 1:48 PM, Carsten Rose wrote:
> Dear all
> 
> I'm new to virtualGL and try to use it in a Cendio/Thinlinc environment. 
> The first tests (VirtualGL 2.5) have been _quite_ impressive by using 
> Gnome3 or Compiz/Unity (Ubuntu 16.04, thinlinc 4.6). Thanks a lot for 
> this superb software - in fact, I couldn't believe that there exist a 
> free solution for sharing GPU(s) among several desktop sessions. Fantastic.
> 
> For our test, we use a NVIDIA Corporation GT218 [GeForce 210]. We 
> installed two of them, but are using only one at the moment.
> 
> After a few days some desktop sessions starts to show garbled desktops, 
> typically for less than a second. 'garbled' means a lot of pixel (not 
> all) are wrong.
> Moving the mouse or a window fixes the broken display - but only until 
> it happens again (sometimes after 10 seconds, sometimes after hours). 
> This appears neither regular nor in every desktop session (one of three 
> daily users never saw this effect). Killing the session and re-login 
> doesn't change anything.
> We see no error messages indicating a problem.
> The nvidia monitoring shows:
> 
> $ nvidia-smi
> Thu Sep 15 20:39:59 2016
> +------------------------------------------------------+ 
> 
> | NVIDIA-SMI 340.96     Driver Version: 340.96         | 
> 
> |-------------------------------+----------------------+----------------------+
> | GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile 
> Uncorr. ECC |
> | Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util 
> Compute M. |
> |===============================+======================+======================|
> |   0  GeForce 210         Off  | 0000:81:00.0     N/A | 
>   N/A |
> | N/A   36C    P8    N/A /  N/A |    935MiB /  1023MiB |     N/A 
> Default |
> +-------------------------------+----------------------+----------------------+
> |   1  GeForce 210         Off  | 0000:82:00.0     N/A | 
>   N/A |
> | N/A   35C    P8    N/A /  N/A |      2MiB /  1023MiB |     N/A 
> Default |
> +-------------------------------+----------------------+----------------------+
>  
> 
> +-----------------------------------------------------------------------------+
> | Compute processes:                                               GPU 
> Memory |
> |  GPU       PID  Process name                                     Usage 
>       |
> |=============================================================================|
> |    0            Not Supported 
>       |
> |    1            Not Supported 
>       |
> +-----------------------------------------------------------------------------+
> 
> 
> Any ideas or hints about the reason or how to do more investigation?
> 
> Thanks a lot for your time.
> 
> CU
> Carsten
> 

------------------------------------------------------------------------------
_______________________________________________
VirtualGL-Users mailing list
VirtualGL-Users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/virtualgl-users

Reply via email to