The latest TurboVNC 3.0 post-beta pre-release build now sets
VGL_PROBEGLX=0 automatically, so you can get the fix simply by upgrading
your TurboVNC Server RPM.
https://turbovnc.org/DeveloperInfo/PreReleases
Ultimately, I was unable to reproduce the issue outside of Tecplot 360,
so I hypothesize that the issue is related to the fact that Tecplot 360
links with its own version of Mesa. If my hypothesis is correct, then
it doesn't make sense to change VirtualGL's visual probing behavior in
general, because technically, VirtualGL isn't doing anything wrong.
However, it does make sense to disable that behavior in environments for
which it will never be useful, including TurboVNC.
Notes:
-- At the time VirtualGL builds the visual attribute table for the 2D X
server, VGL doesn't actually know which image transport will be used.
(In fact, the image transport can be changed dynamically using the
VirtualGL Configuration dialog.) Thus, unfortunately VGL cannot
intelligently set a default value for VGL_PROBEGLX.
-- The issue can also be worked around, without disabling GLX probing,
by setting __GLX_VENDOR_LIBRARY_NAME=nvidia. This simulates what
VirtualGL would do if it interposed the GLX_EXT_libglvnd extension.
That doesn't seem to be the right approach in general, however, because
nVidia's GLX client library doesn't fully work with an X server running
Mesa. This seems to support my hypothesis that VGL is doing the right
thing and that the issue is application-specific.
DRC
On 4/2/22 9:26 AM, Richard Ems wrote:
Great! Many thanks for looking into this issue and finding a solution!
I will add VGL_PROBEGLX=0 to our environment.
Kind regards
Richard
El 1 abr. 2022, a la(s) 19:53, 'DRC' via VirtualGL User Discussion/Support
<[email protected]> escribió:
The issue is related to the fact that VirtualGL probes the 2D X server to see
if it has any stereo visuals. Because of libglvnd, this causes the system's
Mesa GLX vendor library to be loaded into the 3D application process. This
apparently interacts badly with Tecplot 360, perhaps because Tecplot 360 has
its own Mesa implementation. I have no idea why the issue only occurs in batch
mode, but I do know that the issue is triggered by a specific sequence of GLX
calls, so that probably has something to do with it.
The quick workaround is to set VGL_PROBEGLX=0 in the environment. I will
modify the vncserver script in TurboVNC to do that automatically, since
TurboVNC will never have any stereo visuals to probe. Thus, this won't be a
problem again for you after you upgrade to TurboVNC 3.0. But I also need to
see if I can find a more robust solution for VirtualGL, since this will affect
other X proxies. VGL probably needs to learn how to play more nicely with
libglvnd.
I have asked for an extension of my Tecplot eval license, so hopefully I'll be
able to play with this more next week.
DRC
On 3/29/22 8:15 AM, Richard Ems wrote:
Hi DRC,
There is this "Troubleshooting Linux Crashes" article from January 8, 2021 at
https://kb.tecplot.com/2021/01/08/troubleshooting-linux-crashes/
<https://kb.tecplot.com/2021/01/08/troubleshooting-linux-crashes/>
that you probably already know, but I thought I would send the link here just
in case. It may be helpful for others finding this thread.
Thanks again
Richard
On Fri, 25 Mar 2022 at 17:47, 'DRC' via VirtualGL User Discussion/Support
<[email protected] <mailto:[email protected]>>
wrote:
I tested 510.xx, so apparently that isn't the issue. The GLX back
end is the default, but the EGL back end can be selected with
'vglrun -d egl' or 'vglrun -d {DRI-device}' (where {DRI-device} is a
device under /dev/dri, such as /dev/dri/card0. I should be able to
diagnose the issue using a regular (non-administrative) account, as
long as I have SSH access. I will e-mail you an SSH public key.
DRC
On 3/25/22 3:38 PM, Richard Ems wrote:
NVIDIA-SMI 510.47.03 Driver Version: 510.47.03 CUDA Version:
11.6
How would I select one or the other back end?
And yes, giving you access to the workstation should be fine.
Let me run some more tests myself, but the workstation is on
production, so I have limited timeframes to test.
Richard
On Fri, 25 Mar 2022 at 12:53, 'DRC' via VirtualGL User
Discussion/Support <[email protected]
<mailto:[email protected]>> wrote:
I assume that you are using the GLX back end when the error
occurs, but I couldn't make the error occur with the EGL back
end either. Which nVidia driver version are you using? I can
try to match your driver version as closely as possible in
case that has something to do with the failure.
Otherwise, I think the only way forward is to get temporary
access to a machine on which the failure can be reproduced.
DRC
On 3/25/22 7:14 AM, Richard Ems wrote:
Can I provide more setup info or do some more testing?
How can I be helpful?
Richard
On Tue, 22 Mar 2022 at 17:30, 'DRC' via VirtualGL User
Discussion/Support <[email protected]
<mailto:[email protected]>> wrote:
OK, that's an important data point. Now I just need to
be able to reproduce the failure. :|
On 3/22/22 2:56 PM, Richard Ems wrote:
Hi DRC,
Just tested and it does not fail just using the local X
server. See attached output.
Thanks,
Richard
On Tue, 22 Mar 2022 at 15:19, 'DRC' via VirtualGL User
Discussion/Support <[email protected]
<mailto:[email protected]>> wrote:
On 3/22/22 12:26 PM, Richard Ems wrote:
Is there any way that you could try running the
Tecplot batch job on the 3D X server without
VirtualGL? If it fails in the same way, then
it isn't our bug.
Do you mean on a local attached display? That won't
be easy, since this graphics workstation is part of
a cluster located far away from me or the user
getting this bug.
I have just contacted Tecplot Support and I'm
waiting for their response and a trial license to
test myself using my own environment.
The 3D X server is on the VirtualGL server, which
should be the same machine on which Tecplot 360 is
installed. Unless the machine has been configured
otherwise, the 3D X server should be listening on
Display :0.0, so try:
xauth merge /etc/opt/VirtualGL/vgl_xauth_key
LD_PRELOAD= DISPLAY=:0.0 ./Tec360_FlowViz_batch.sh
I remember to have to preload libvglfaker.so for
starccm+ to work properly. Anything similar perhaps
needed here for tec360? Just some more guessing.
Worth a try?
Preloading libvglfaker.so is literally what vglrun
does. I am not trying to make Tecplot 360 work
properly. I am trying to make it fail.
Thanks,
Richard
On Tue, 22 Mar 2022 at 13:58, 'DRC' via VirtualGL
User Discussion/Support
<[email protected]
<mailto:[email protected]>> wrote:
I tried it with '$useVGL = 1' as well and was
still unable to reproduce the issue. (Note
that, from the point of view of the
application, setting '$useVGL = 1' in
turbovncserver.conf or starting the TurboVNC
session with '-vgl' is the same as launching
the application with 'vglrun +wm'.)
I also tried with both CentOS 7.9 and CentOS
Stream 8. My CentOS 7.9 machine has a Quadro
K5000 with driver v470.xx. My CentOS 8 machine
has a Quadro P620 with driver v510.xx.
Is there any way that you could try running the
Tecplot batch job on the 3D X server without
VirtualGL? If it fails in the same way, then
it isn't our bug.
DRC
On 3/22/22 11:01 AM, Richard Ems wrote:
Hi DRC,
sorry, I forgot to comment out that module
command, that would only add the Tecplot
installation PATH to the bash PATH variable,
But as I see from your output the tec360
command in Tec360_FlowViz_batch.sh started
fine, so it is already in your PATH, so all good.
Did that tec360 run end without errors?
Is your setup also using $useVGL = 1; in
/etc/turbovncserver.conf ?
Just guessing possible differences.
Thanks for testing,
Richard
On Tue, 22 Mar 2022 at 11:37, 'DRC' via
VirtualGL User Discussion/Support
<[email protected]
<mailto:[email protected]>> wrote:
Please clarify how this is supposed to
work. 'module load' is not a valid Bash
command.
==========
> vglrun ./Tec360_FlowViz_batch.sh
./Tec360_FlowViz_batch.sh: line 3: module:
command not found
Tecplot 2021.2.1.9698 - 21:07:35 Feb 3
2022 [linux64-centos6.10]
Color maps directory :
/usr/local/tecplot/360ex_2021r2/colormaps
Configuration File :
/usr/local/tecplot/360ex_2021r2/tecplot.cfg
Process Temp Dir :
/usr/tmp/tecplot_drc/tpaqmYUE0
Tecplot internal font file:
/usr/local/tecplot/360ex_2021r2/tecplot.fnt
LaTeX Configuration File:
/usr/local/tecplot/360ex_2021r2/tecplot_latex.mcr
Your license expires at the end of
26-Mar-2022.
Update your license, or contact
[email protected]
<mailto:[email protected]> for help.
Tecplot executable :
/usr/local/tecplot/360ex_2021r2/bin/tec360-bin
Tecplot Home Directory :
/usr/local/tecplot/360ex_2021r2
QuickMacro File :
/usr/local/tecplot/360ex_2021r2/tecplot.mcr
Addon Startup File :
/usr/local/tecplot/360ex_2021r2/tecplot.add
AddOn Loaded : Tecplot Subzone Data Loader
AddOn Loaded : SZL Remote Loader
AddOn Loaded : CFD Analyzer
AddOn Loaded : CGNS Loader
AddOn Loaded : DEM Loader
AddOn Loaded : DXF Loader
AddOn Loaded : EnSight Loader
AddOn Loaded : FLOW-3D Loader
AddOn Loaded : Fluent Common Fluid Files
Loader
AddOn Loaded : Fluent Data Loader
AddOn Loaded : General Text Loader
AddOn Loaded : HDF Loader
AddOn Loaded : PLOT3D Loader
AddOn Loaded : PLY Polygon File Loader
AddOn Loaded : Text Spreadsheet Loader
AddOn Loaded : Kiva Loader
AddOn Loaded : Finite Element Analysis Loader
AddOn Loaded : Write Data as Text
AddOn Loaded : VTK Data Loader
AddOn Loaded : FVCOM netCDF Loader
AddOn Loaded : CONVERGE HDF5 File Loader
AddOn Loaded : CONVERGE Output File Loader
AddOn Loaded : Exodus File Loader
AddOn Loaded : Telemac Data Loader
AddOn Loaded : HDF5 Loader
AddOn Loaded : Akima Spline
AddOn Loaded : General Curve Fit
AddOn Loaded : Stineman Interpolation
AddOn Loaded : extendmcr
AddOn Loaded : Advanced Quick Edit Tool
AddOn Loaded : Multi Frame Manager
AddOn Loaded : Create Multiple Frames
AddOn Loaded : Tensor Eigensystem
AddOn Loaded : Key Frame Animation
AddOn Loaded : Aux Data Editor
AddOn Loaded : Time Series Plot
AddOn Loaded : Extract Blanked Zone
AddOn Loaded : Extract Precise Line
AddOn Loaded : Extract Over Time
AddOn Loaded : Link Solution Time
AddOn Loaded : Extend Time MCR
AddOn Loaded : Strand Editor
AddOn Loaded : Measure Distance
==========
DRC
On 3/21/22 4:25 PM, Richard Ems wrote:
Hi DRC,
find attached an example using Tecplot's
Onera example model.
Thanks,
Richard
On Mon, 21 Mar 2022 at 15:28, 'DRC' via
VirtualGL User Discussion/Support
<[email protected]
<mailto:[email protected]>> wrote:
I have installed Tecplot 360 with an
evaluation license. Please specify
the exact steps needed to reproduce
the issue using one of the Tecplot
examples.
DRC
On 3/21/22 7:54 AM, Richard Ems wrote:
Hi DRC, hi all,
First, thanks DRC for this great
software!
We are having some issue with
Tecplot / tec360 and VirtualGL:
Some system specs:
Rocky Linux 8.5
VirtualGL-3.0-20211119.x86_64
turbovnc-2.2.90-20211222.x86_64
NVIDIA Corporation GM204GL [Tesla M60]
Tecplot 360 EX 2021 R2, Version
2021.2.1.9698 (Feb 3 2022)
VirtualGL is configured with both
GLX + EGL back ends.
$useVGL = 1; in
/etc/turbovncserver.conf
While trying to run tec360 on batch
mode we get the attached error output.
Here the last 10 lines:
/net/c1m1/data_1/opt/Tecplot360/360ex_2021r2/bin/../bin/libtpsdkbase.so(TecEngStartup+0x24)
[0x7f93fdad0394]
/net/c1m1/data_1/opt/Tecplot360/360ex_2021r2/bin/../bin/libtpsdkintegrationmanager.so(_ZN7tecplot3sdk11integration15ManagerAbstract5startEv+0x19)
[0x7f93fe92b1b9]
/net/c1m1/data_1/opt/Tecplot360/360ex_2021r2/bin/tec360-bin()
[0x487ef5]
/lib64/libc.so.6(__libc_start_main+0xf3)
[0x7f93fb4c0493]
/net/c1m1/data_1/opt/Tecplot360/360ex_2021r2/bin/tec360-bin()
[0x489d09]
Tecplot received shutdown signal.
Signal #11
Please send the output above to
Tecplot support: [email protected]
<mailto:[email protected]>
/net/c1m1/data_1/opt/Tecplot360/360ex_2021r2/bin/tec360:
line 349: 276303 Aborted (core
dumped) "$0-bin" "$@"
Any ideas what is going wrong?
When running with tec360 option
"--osmesa" all runs fine, but it
takes 17 min to run.
--osmesa
Bypass the OpenGL library and
use the LLVM OSMesa library for
off-screen
image rendering. This option
does not require an X server connection.
Best regards,
Richard
-- You received this message because you are subscribed
to
the Google Groups "VirtualGL User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails
from it, send an email to
[email protected]
<mailto:[email protected]>.
To view this discussion on the web visit
https://groups.google.com/d/msgid/virtualgl-users/14b5014b-6277-0440-e9c5-11f7e517419a%40virtualgl.org
<https://groups.google.com/d/msgid/virtualgl-users/14b5014b-6277-0440-e9c5-11f7e517419a%40virtualgl.org?utm_medium=email&utm_source=footer>.
-- You received this message because you are subscribed to the
Google Groups "VirtualGL User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails from
it, send an email to
[email protected]
<mailto:[email protected]>.
To view this discussion on the web visit
https://groups.google.com/d/msgid/virtualgl-users/CAEqczzzNxRQN5N5sFQr08coZV67O5WkNCTBpGa9kE_d5x%2BfmbA%40mail.gmail.com
<https://groups.google.com/d/msgid/virtualgl-users/CAEqczzzNxRQN5N5sFQr08coZV67O5WkNCTBpGa9kE_d5x%2BfmbA%40mail.gmail.com?utm_medium=email&utm_source=footer>.
-- You received this message because you are subscribed to the
Google Groups "VirtualGL User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails from
it, send an email to
[email protected]
<mailto:[email protected]>.
To view this discussion on the web visit
https://groups.google.com/d/msgid/virtualgl-users/0a7a6eb3-5afc-3a8c-9b4e-a1212ecb17ab%40virtualgl.org
<https://groups.google.com/d/msgid/virtualgl-users/0a7a6eb3-5afc-3a8c-9b4e-a1212ecb17ab%40virtualgl.org?utm_medium=email&utm_source=footer>.
-- You received this message because you are subscribed to the Google
Groups "VirtualGL User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails from it,
send an email to [email protected]
<mailto:[email protected]>.
To view this discussion on the web visit
https://groups.google.com/d/msgid/virtualgl-users/CAEqczzw%2B3Na122FXLwxCtyYYRYK_NOJ67X7AAGsjniLABpJuUg%40mail.gmail.com
<https://groups.google.com/d/msgid/virtualgl-users/CAEqczzw%2B3Na122FXLwxCtyYYRYK_NOJ67X7AAGsjniLABpJuUg%40mail.gmail.com?utm_medium=email&utm_source=footer>.
-- You received this message because you are subscribed to the Google
Groups "VirtualGL User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails from it,
send an email to [email protected]
<mailto:[email protected]>.
To view this discussion on the web visit
https://groups.google.com/d/msgid/virtualgl-users/1a86354e-6214-97fc-c643-d4d325323085%40virtualgl.org
<https://groups.google.com/d/msgid/virtualgl-users/1a86354e-6214-97fc-c643-d4d325323085%40virtualgl.org?utm_medium=email&utm_source=footer>.
--
You received this message because you are subscribed to the Google Groups "VirtualGL
User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to
[email protected]
<mailto:[email protected]>.
To view this discussion on the web visit
https://groups.google.com/d/msgid/virtualgl-users/CAEqczzwyD1DQkdSKvF-gDUSo%2B4xvZDaVvwZ2A2XQR5sVrr0AoA%40mail.gmail.com
<https://groups.google.com/d/msgid/virtualgl-users/CAEqczzwyD1DQkdSKvF-gDUSo%2B4xvZDaVvwZ2A2XQR5sVrr0AoA%40mail.gmail.com?utm_medium=email&utm_source=footer>.
--
You received this message because you are subscribed to the Google Groups "VirtualGL
User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/virtualgl-users/29357079-1148-4d2d-d299-16ff80719f62%40virtualgl.org.
--
You received this message because you are subscribed to the Google Groups "VirtualGL
User Discussion/Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/virtualgl-users/e1837dd4-bed0-4773-39b7-26f6cc1d922a%40virtualgl.org.
