Restart Issue

[swetarkad23]

The alamo documentation suggests the restart is done using the following command:
<executable_name> input.file restart=/address/of/the/output/directory/upto/final/written/data
This is not working with an error being showed:
amrex::Error::0::Couldn't open file: //00008cell/Header !!!
I checked the code and the Restart function in Integrator.cpp which suggested a 'restart_cell' input in the input file, I did that as well, by providing the address of the final written data in the input file. But, it fails at the first iteration only.
So, if anyone can pour some idea about how exactly the restart of a simulation is done in present state of the alamo code, that will be really helpful.
thank you.

[bsrunnels]

Thanks for the post @swetarkad23. The restart has some glitches and we are working on them currently. @etsandall , would you be able to take a look and possibly apply a fix to development?

@swetarkad23 Can you please provide a complete example so we can reproduce your issue? https://solidsgroup.github.io/alamo/docs/Questions.html

[etsandall]

@swetarkad23 Can you confirm one thing for me?

After adding the restart=/path/to/###cell, are the paths for plot_file and restart the same? If a directory already exists when starting a run, alamo will rename the old path to /path/to/plot_file.old.####. Thus the restart file that you are trying to read from technically no longer exists. You can try explicitly renaming your previous run, say to something like mv /path/to/plot_file /path/to/plot_file_run1 and then set restart=/path/to/plot_file_run1/###cell

[swetarkad23]

@etsandall I faced this earlier. Now, when I restart a case, I name the output directory different name. So alamo reads the restart file from existing output directory from previous run and starts writing the new restart results in newly named another output directory. Let's say if it ends at 1s and I restart the case from 1s again, then alamo writes the new output directory. But right at the beginning of the first time step it fails after showing "STEP xxx starts ..." with a 'Segfault' error.

[etsandall]

@swetarkad23 Okay, thanks for confirming that. Can you please provide the info requested by @bsrunnels per https://solidsgroup.github.io/alamo/docs/Questions.html? This will help me reproduce and hopefully isolate the error.

[swetarkad23]

@bsrunnels @etsandall
thank you for your kind replies. I'm attaching a zip-file with this comment. The zip-file contains a word file named as 'Alamo_install_info_Swetarka.docx'. Please follow the content of this word file which answers the informations you needed according to the link, provided above. Also there is a 'Case' folder, which contains an example of Lid-Driven-Cavity run. And the input file is made for restart. History of the runs are provided as well. In nutshell, the zip file is self explanatory.
The same is mailed as well. If any further information is required, please let me know.

ISSUE1_RestartIssue.zip

[etsandall]

@swetarkad23 Thank you for sending the requested information. I was able to replicate your issue, but unfortunately, the scimitarx branches are very much under development and haven't been reviewed for merging into development. Perhaps @PradeepKSeshadri can help address this issue? Otherwise, if @bsrunnels agrees, I will just close this issue as not planned for the time being.

[bsrunnels]

We have verified that the restart capability in Integrator::Integrator works as expected, and have added a regression test to check this capability for Integrator::Hydro (#212). However, it is up to each implementation to ensure that restart works for it, and this may not be the case for the scimitarx branches. Please work with @PradeepKSeshadri and the scimitarx team to determine what is necessary to activate restart capability for that family of integrators.

[PradeepKSeshadri]

Hi @bsrunnels ! Yes I am working on this and updating ScimitarX branches in sync with latest alamo development branch.