Closed wangyinz closed 1 year ago
Testing on my laptop, neither pumgen nor seissol works.
Now, can you give instructions on how to run it on frontera? What I tried:
idev
.
ulrich@c207-005:~$ vncserver -geometry 1600x1200
You will require a password to access your desktops.
Password: Password must be at least 6 characters - try again Password: Verify: Would you like to enter a view-only password (y/n)? n A view-only password is not used
New 'c207-005.frontera.tacc.utexas.edu:1 (ulrich)' desktop is c207-005.frontera.tacc.utexas.edu:1
Creating default startup script /home1/09160/ulrich/.vnc/xstartup Creating default config /home1/09160/ulrich/.vnc/config Starting applications specified in /home1/09160/ulrich/.vnc/xstartup Log file is /home1/09160/ulrich/.vnc/c207-005.frontera.tacc.utexas.edu:1.log
- connect with vncviewer
![image](https://user-images.githubusercontent.com/13536910/236235773-93e31e79-c9cb-4d09-a177-bb3ca5e7b62e.png)
(Then I get an error).
Weird. The same command runs fine on my laptop:
I cannot find the msh file so was not able to test the pumgen command.
Had a call with Thomas on Slack, and we have both issues addressed.
I've opened a PR to your PR: https://github.com/wangyinz/SeisSolTraining/pull/1
Thanks @wangyinz for your contributions. Most things look good. I took your work and continued working on that for two reasons:
continued by #33
I have changed the base container to the SCOPE base container, which has jupyterlab and Intel MPI built in. This version of container should be able to run locally as well as on hpc systems of any mpich variant of MPIs available (at least true for TACC's system).
I have tested a little bit and it seems working fine, but my test is not thorough, so please give it a try using the build available from docker hub here, or use the following command.
One issue that I discovered while building this is that the Intel MPI does not support Fortran 2008 when using with gfortran. The details are in this issue, https://github.com/SeisSCOPED/container-base/issues/5. However, it seems the fortran binding of HDF5 is not required, so I simply disabled that here.
Also, because the base container has python 3.10, the original fixed version numbers in
requirements.txt
does not work anymore. I switched that to using conda, and it seems doing OK.