[Wien] MPI error

Wed May 5 05:24:19 CEST 2021

Three additional comments:

1)  If you are running the slurm.job script as Non-Interactive [1,2], 
you might need a "source /etc/profile.d/ummodules.csh" line like that at 
[3].

[1] https://slurm.schedmd.com/faq.html#sbatch_srun
[2] https://wiki.umiacs.umd.edu/umiacs/index.php/SLURM/JobSubmission
[3] https://wiki.umiacs.umd.edu/umiacs/index.php/Modules

2) To eliminate any possible conflicts between Intel compilers (ifort, 
icc) and GNU compilers (gfortran, gcc) such as those mentioned in posts 
[4,5].  I suggest compiling of WIEN2k using the same C compiler and 
Fortran compiler that Open MPI was compiled with.

The commands [6] below might help you check that the Linux environment 
is using the intended Open MPI and mpi parallel compilers [7]:

username at computername:~/Desktop$ mpirun -V
mpirun (Open MPI) 4.1.1

Report bugs to http://www.open-mpi.org/community/help/
username at computername:~/Desktop$ mpicc --showme:version
mpicc: Open MPI 4.1.1 (Language: C)
username at computername:~/Desktop$ mpifort --showme:version
mpifort: Open MPI 4.1.1 (Language: Fortran)
username at computername:~/Desktop$ mpicc --version
gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

username at computername:~/Desktop$ mpifort --version
GNU Fortran (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

[4] 
https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg08315.html
[5] 
https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg17052.html
[6] 
https://stackoverflow.com/questions/10056898/how-do-you-check-the-version-of-openmpi
[7] https://www.open-mpi.org/faq/?category=mpi-apps#general-build

3) Those cluster administrators are usually more savvy than I am with 
installation and optimization of software (using compiler documentation, 
e.g. [8,9]) on a high performance computing (hpc) supercomputer 
[10,11].  They would know your situation better. For example, they could 
login to their administrator account on the cluster to install WIEN2k 
only in your user account directory (/home/users/mollabashi), and they 
would know how to set the appropriate access permissions [12].  
Alternatively, if your not using a personal laptop but a computer at the 
organization to remotely connect to the cluster then they might use 
remote desktop access [13] to help you with the installation within only 
your account.  Or they might use another method.

[8] 
https://software.intel.com/content/www/us/en/develop/articles/download-documentation-intel-compiler-current-and-previous.html
[9] https://gcc.gnu.org/onlinedocs/
[10] 
https://www.usgs.gov/core-science-systems/sas/arc/about/what-high-performance-computing
[11] https://en.wikipedia.org/wiki/Supercomputer
[12] 
https://www.oreilly.com/library/view/running-linux-third/156592469X/ch04s14.html
[13] https://en.wikipedia.org/wiki/Desktop_sharing

On 5/4/2021 3:40 PM, Laurence Marks wrote:
> For certain, 
> "/opt/exp_soft/local/generic/openmpi/4.1.0_gcc620/bin/mpiexec 
> /home/users/mollabashi/codes/v21.1/run_lapw -p" is completely wrong. 
> You do not, repear do not use mpirun or mpiexec to start run_lapw. It 
> has to be started by simply "run_lapw -p ..." by itself.
>
> I suggest that you create a very simple job which has the commands:
>
> which mpirun
> which lapw1_mpi
> echo $WIENROOT
> ldd $WIENROOT/lapw1_mpi
> ldd $WIENROOT/lapw1
> echo env
> echo $PATH
>
> Run this interactively as well as in a batch job and compare. You will 
> find that there are something which are not present when you are 
> launching your slurm job that are present interactively. You need to 
> repair these with relevant PATH/LD_LIBRARY_PATH etc
>
> Your problems are not Wien2k problems, they are due to incorrect 
> modules/script/environment or similar. Have you asked your sysadmin 
> for help? I am certain that someone local who is experienced with 
> standard linux can tell you very quickly what to do.
>
> N.B., there is an error in your path setting.
>
> On Tue, May 4, 2021 at 3:38 PM leila mollabashi 
> <le.mollabashi at gmail.com <mailto:le.mollabashi at gmail.com>> wrote:
>
>     Dear all WIEN2k users,
>     Thank you for your guides.
>     >take care on the correct location ...
>     It is the /usr/share/Modules/init
>     After adding the “source /usr/share/Modules/init/tcsh” line in to
>     the script the same error appeared:
>     mpirun: command not found
>
>     In fact, with and without “source /usr/share/Modules/init/tcsh” it
>     is written in slurm.out file that “ module load complete ”.
>
>     I noticed that “export” is also the bash command so I used these
>     commands to path the openmpi and fftw:
>     setenv LD_LIBRARY_PATH
>     {$LD_LIBRARY_PATH}:/opt/exp_soft/local/generic/openmpi/4.1.0_gcc620/lib:/opt/exp_soft/local/generic/fftw/3.3.8_gcc620/lib
>     set path = ($path
>     /opt/exp_soft/local/generic/openmpi/4.1.0_gcc620/bin:/opt/exp_soft/local/generic/fftw/3.3.8_gcc620/bin)
>     But result is the same:
>     bash: mpirun: command not found
>
>     By using this line in the script:
>     /opt/exp_soft/local/generic/openmpi/4.1.0_gcc620/bin/mpiexec
>     /home/users/mollabashi/codes/v21.1/run_lapw -p
>     The calculation stopped with the following error:
>     mpirun does not support recursive calls
>
>     > I wonder that you have only single modules…
>     There are different versions of ifort and mkl: ifort/15.0.0,
>     ifort/15.0.3, ifort/17.0.1, ifort/19.1.3.304(default) mkl/11.2,
>     mkl/11.2.3.187
>     <https://urldefense.com/v3/__http://11.2.3.187__;!!Dq0X2DkFhyF93HkjWTBQKhk!BD88XA2ujgG8Gel0NZaSOKaSrtPN7kq75O9hkG-dnIZQJRbsnQE-ArEqbFqA6XVu7qQFcg$> mkl/2017.1.132,
>     mkl/2019.2.187, mkl/2020.0.4(default). I used the defaults
>     > you may also wish to make a single module file to be loaded…
>     That is a good idea.
>     > On our cluster we have different W2k modules ….
>     As you know WIEN2k is not a free code and the users of the cluster
>     that I am using are not registered WIEN2k users. Thus, according
>     to my moral commitment to the WIEN2k developers, I cannot ask the
>     administrator to install it on the cluster. I should install it on
>     my user account.
>
>     Sincerely yours,
>     Leila
>     >PS.: maybe one should mention this tcsh "problem" in the
>     slurm.job example on the FAQ page by adding (or similar)…
>     That is a good idea. Thank you for your suggestion.
>
> -- 
> Professor Laurence Marks
> Department of Materials Science and Engineering
> Northwestern University
> www.numis.northwestern.edu <http://www.numis.northwestern.edu/>
> "Research is to see what everybody else has seen, and to think what 
> nobody else has thought" Albert Szent-Györgyi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20210504/535d6c1e/attachment-0001.htm>