[villager]

Originally Posted by villager View Post
Have your read my post carefully and done steps 1-3?

Originally Posted by Cipis

Yes. It is not work.

I'm afraid you did not really read carefully the instructions.
It seems you did not set LD_LIBRARY_PATH. This way one will have "Unable to resolve function glXMakeCurrent" errors and DesignModeler would not show anything.


LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/ansys_inc/v162/Framework/bin/Linux64/Mesa
export FLUENT_ARCH="lnamd64"
/ansys_inc/v162/Framework/bin/Linux64/runwb2 -oglmesa


Yes, drop libxp-dev. It is likely it would work without it.

[hadr1an]

Hello, After sucessfully installing ANSYS 17, I get the following error everytime I try to open Design Modeler.

libGL error: unable to load driver: i965_dri.so
libGL error: driver pointer missing
libGL error: failed to load driver: i965
libGL error: unable to load driver: swrast_dri.so
libGL error: failed to load driver: swrast

Has anyone experienced something similar? I think that it is related to the file

/ansys_inc/commonfiles/MainWin/linx64/mw/setup-mwuser

because I had also to correct the RPM function there. I am running Arch Linux on a Lenovo Thinkpad X300 with Intel Graphics G965
Thank You

P.S. In a previous Installation I had found a way to fix this but now I can't figure it out..

[rustorey]

Originally Posted by villager

I'm afraid you did not really read carefully the instructions.
It seems you did not set LD_LIBRARY_PATH. This way one will have "Unable to resolve function " errors and DesignModeler would not show anything.


LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/ansys_inc/v162/Framework/bin/Linux64/Mesa
export FLUENT_ARCH="lnamd64"
/ansys_inc/v162/Framework/bin/Linux64/runwb2 -oglmesa


Yes, drop libxp-dev. It is likely it would work without it.

I thought it might be useful to share my experience installing Ansys 17.1 on Ubuntu 16.04.

I still had the issue with "Unable to resolve function glXMakeCurrent" after going through the recommended fixes too.

I managed to solve it by creating a link to the libGL library as ANSYS DesignModeller is expecting to find libGL.so.1 in /usr/lib/

sudo ln -s /usr/lib/x86_64-linux-gnu/mesa/libGL.so.1 /usr/lib/libGL.so.1

You should use the libGL library specific to your computer. I found mine using:

locate libGL.so

Everything seems to work at the moment

Good luck!

[chaitanyaarige]

Workbench has upgraded to workbench 2.0 after Ansys15.

./runwb2

try this

[flotus1]

If I may add my recent experience installing Ansys 18 on Opensuse Leap 42.1:

After the installation, I was facing errors of this type when starting fluent or other tools with a GUI:

libGL error: No matching fbConfigs or visuals found
libGL error: failed to load driver: swrast

The tips given here definitely helped ANSYS Workbench on "Uncertified" Linux Distros
But this way the Nvidia driver was not used which caused the programs to run extremely slow and choppy.

I figured out the only thing I needed to do instead was adding the library path to the Nividia driver. In my case this was:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/X11R6/lib64

[Shoonya]

Yes, I your solution is correct. I also added the following in ".bashrc" file for myself and it worked.
Mine is local installation (not root). Just to sum up here for others.

1. open bash file
$vim .bashrc

2. add the following lines at the end of file and set the path according to your installation

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/media/shoonya/Pacific/ANSYS171/ansys_inc/v171/Framework/bin/Linux64/Mesa
export FLUENT_ARCH="lnamd64"
/media/shoonya/Pacific/ANSYS171/ansys_inc/v171/Framework/bin/Linux64/runwb2 -oglmesa

3. Source the .bashrc file (or logout)
$. .bashrc

4. run the following command (make changes according to your path of executable "runwb2")

$ /media/shoonya/Pacific/ANSYS171/ansys_inc/v171/Framework/bin/Linux64/runwb2

[ali hatamy]

Originally Posted by Yankee

I've the same errors in ansys_inc/install.err , it seems that there is a small problem in an installation script. On my 32 bit ubuntu this does not prevent Design Modeller to work (but, as said before, Meshing does not).

For the "blank graphics screen" problem in Design Modeller on my 64 bit distro, I found a workaround : add the following two directories in LD_LIBRARY_PATH before calling runwb2:
ansys_inc/v121/Framework/bin/Mesa
ansys_inc/v121/Framework/bin
In fact the "mono" executable called by runwb2 (an horrible windows-like programming environment, but freeware, so don't shout at it !), should pre-load openGL libraries (more exactly their "mesa" variant) because of an environment variable LD_PRELOAD set by runwb2. Seems that it doesn't work... Maybe it's not transmitted to windows executables running DM or Meshing? But with LD_LIBRARY_PATH, for me it works.

I use the following script called wb2, placed in /usr/local/bin (which is in my path), to run ansys, in place of runwb2:

#!/bin/bash
RUNWB2CMD=`which runwb2`
RUNWB2DIR=`dirname $RUNWB2CMD`
if test -z "$LD_LIBRARY_PATH"; then
    LD_LIBRARY_PATH=$RUNWB2DIR/Mesa:$RUNWB2DIR
else
   LD_LIBRARY_PATH=$RUNWB2DIR/Mesa:$RUNWB2DIR:$LD_LIBRARY_PATH
fi
export LD_LIBRARY_PATH
export LANG=en_US.UTF8
runwb2 $@

Then, just type wb2 in place of runwb2, and for me it works!

hi. i have installed 18.2 64bit on lubuntu 16.04 64bit
for my path /usr/ansys_inc/v182/Framework/bin/Linux64/runwb2 how can i use your code.
i dont know what you mean of `which runwb2` & etc

[jgross]

Hi everyone,

Sorry to hijack this post, but this seems to be the best place to post questions regarding getting CFX to work on uncertified distros. As a first note, my distro is Ubuntu 16.04 and my Ansys version is 18.2. I have used Ansys 18.2 on UBuntu 16.04: Installation Guide in order to install Ansys on my desktop. Everything in general works perfectly fine, including running from workbench or standalone.

However, I run into an issue when I attempt to perform parallel calculations using the Intel MPI method (both local and distributed). Namely, the issue seems to be when writing results. The calculations are performed perfectly fine as I can view the residuals in the Solver Manager and they behave exactly as they do in serial. The issue is when it comes to writing the results file, the solver hangs indefinitely. Checking CPU usage shows that all designated processors are working at 100%, but nothing is being done. My guess is that the slave nodes are continuously sending signals to the master node (hence why all processors are working 100%), but either the master node is not listening or isn't there.

I know this problem is to do with the set up as using the tutorials as a case study gives the exact same problem. Furthermore, it is not an issue with using the GUI as it behaves exactly the same way if I run it from command line or not. Unfortunately, I cannot use a different parallel method as the only available one is the IBM method, but when I attempt to use it, the solver complains about a missing library (libmport.so), which a quick Google turns up no results.

Has anyone experienced something similar or have any advice of how I may fix this problem? Thank you in advance.

Regards,
James

[huey1080]

Hi James,
How are you starting cfx (batch or gui) ?
Can you provide the command line?
Is there anything in the output when it stalls ?
Are you running from a shared directory since the slave nodes needs to write data back at the end of the computation?
Is the user running the simulation is the owner of your working directory ?

[jgross]

Hi Quentin,

First of all, thank you for taking the time to respond to my post.

How are you starting cfx (batch or gui) ?
Can you provide the command line?

In general, I am using the GUI for CFX standalone (i.e. using the cfx5 command). However I have also attempted to run in batch by using

cfx5solve -def <mymodel.def> -par-local -partition 4

where I have defined the parallel method in CFX-pre before writing the def file and the partition scheme has been left as default.

Is there anything in the output when it stalls ?

The output file does not print any thing once it attempts to begin to write the file. I have attached the out file as a txt file here. Furthermore, I turned on verbosity for the Intel MPI method, and inspecting the text from the terminal shows there is no output once the calculations begin and certainly none from when it attempts to write the results. I have also attached the output from the terminal as a txt file.

Are you running from a shared directory since the slave nodes needs to write data back at the end of the computation?
Is the user running the simulation is the owner of your working directory ?

I am writing from a directory contained in my home folder of my user account. The folder is accessible by all users. I will say I have not attempted to run from command line using sudo, but I didn't think this completely necessary as the folder it attempts to write to is user accessible.


Regards,
James

[huey1080]

Okay you run on a single machine, I did not get that earlier, forget what I said about the shared directory...
You computation crashes with code 9, that's why you don't have any output. Don't know the cause though. It seems more of an issue with your case than an installation issue, check your mesh and case.

[jgross]

Oh I'm sorry, I didn't fully explain the out file. There is no error message when I leave the solver to its own devices. The code 9 error message only occurs when I manually kill the processors using the command

kill -9 <PID>

It is most definitely not a case specific problem, as I have tested this on a number of case files including tutorials and I always get the same behaviour.

Regards,
James

[huey1080]

Okay I understand.
Can you run it serial ?

[jgross]

Yes, everything works just fine in serial.

[huey1080]

Try to run with root, see if it's a permission issue.

[jgross]

Hi Quentin,

I just attempted to run with root using

sudo cfx5solve -def PipeValve.def -par-local -partition 4

However when I did so, the command was not found

sudo: cfx5solve: command not found

However it runs just fine when running it from my user account. I attempted to edit my superuser bash script to contain the paths necessary to run the cfx5solve command, but it didn't work. I suppose this is because all of necessary paths and soft links created were done so in my home directory, as per the installation guide (Ansys 18.2 on UBuntu 16.04: Installation Guide).

Either way, I'm not convinced it is a problem with permissions, as I would certainly receive a permission denied error or a request for a password. Furthermore the above command (run from my user account) works just fine in serial. However, that still begs the question, what is happening for the process to hang indefinitely with all designated processors working at full capacity until I manually kill them? Any ideas as to what could be causing this?

Regards,
James

[huey1080]

If you want to run with sudo, try the full path instead of just the exec.
Another idea would be to lower the security see if that change everything, enter:
sudo setenforce 0
and run again with your normal user.

[jgross]

Hi Quentin,


Sorry for the really delayed response. I've been really busy recently and didn't get a chance to try your suggestions.


I just tried the parallel run using

sudo /usr/ansys_inc/v182/CFX/bin/cfx5solve -def PipeValve.def -par-local -partition 4

It worked!! So it was a write permission issue after all!


Thank you again for your help!!


Regards,
James

[jgross]

Hello again,

Sorry if it seems I am spamming this thread, but it seems this is the best thread to post questions regarding Ansys set up for Ubuntu or other "uncertified" distros.

I am trying to create a script for ICEM to import a STEP file and create an unstructured mesh of the geometry. In general, it is fine if I first import the geometry into DesignModeller first. However, DesignModeller is not as easy to script as ICEM (because it doesn't have a log a script). If I attempt to import the STEP directly in to ICEM, I get the following error:

/user/ansys_inc/v182/icemcfd/linux64_amd/dif/step/steptotin: error while loading shared libraries: libX11.so.6: cannot open shared object file: No such file or directory

Using

locate libX11.so.6

displays the following:

/usr/ansys_inc/v182/Electronics/Linux64/defer/libX11.so.6
/usr/ansys_inc/v182/Electronics/Linux64/defer/libX11.so.6.2
/usr/ansys_inc/v182/Electronics/Linux64/mainwin540/Linux64/mw/lib-amd64_linux/X11/libX11.so.6
/usr/ansys_inc/v182/Electronics/Linux64/mainwin540/Linux64/mw/lib-amd64_linux/X11/libX11.so.6.3.0
/usr/ansys_inc/v182/commonfiles/MainWin/linx64/mw/lib-amd64_linux/X11SLES/libX11.so.6
/usr/ansys_inc/v182/commonfiles/MainWin/linx64/mw/lib-amd64_linux/X11SLES/libX11.so.6.3.0
/usr/ansys_inc/v182/polyflow/polyflow18.2.0/lnamd64/libs/add_libs/libX11.so.6
/usr/ansys_inc/v182/polyflow/polyflow18.2.0/lnamd64/libs32/libX11.so.6
/usr/lib/x86_64-linux-gnu/libX11.so.6
/usr/lib/x86_64-linux-gnu/libX11.so.6.3.0

So I definitely have it, but it does not recognise it.

I remember that part of the installation process required me to create a soft link using this command:

sudo ln -sf /usr/lib/x86_64-linux-gnu/libX11.so.6 /usr/lib/libX11.so

Unfortunately, Googling the issue does not seem to give me much information about it. I tried to view the

/user/ansys_inc/v182/icemcfd/linux64_amd/dif/step/steptotin

file, but of course, it's a binary file, so not much help there either.

Does anyone know what could be the cause of this?

Kind regards,
James

[jgross]

Hello again,


As a quick follow up to the previous post, importing STL and other formats works just fine. The issue only arises for STEP and IGES formats.



Does anybody know why this may be happening or have any advice as to what I can do to check what may be the issue?


Regards,
James