Messages

1

General Questions and Answers / Re: Checkpoint handler and remote jobs

« on: July 29, 2018, 03:25 »

-   Create a script file specifying a local path for the checkpoint file (i.e. u’C:\User\[...]\Bi_nw\checkpointraw.hdf5’)
-   Go into the script editor and change it to an absolute path on the server (i.e. u’/W/sb255620/VNL/checkpointraw.hdf5’)

You should use relative (not absolute) paths to the directory where the checkpoint file is supposed to be created, e.g.,  u'../tmp/checkpointraw.hdf5’ in the 'tmp' directory is created in the working directory.

-   Create a blank “checkpoint.hdf5” file at the specified location on the server, and make it writable for all users (otherwise atk will throw an error when I start the calculation)

It is a bug, but you have found a work-around, i.e., one should first create a blank checkpoint file in the checkpoint file directory. Thank you for reporting this bug; we will fix it in the next release, so that it will be possible to set u'checkpoint.hdf5' in the Scripter to have the checkpoint file directly in the working directory with all the other (py, log, hdf5) files.

When I do this, the calculations finish, but once I download the result file on my local machine and try to read the band structure, I get the error shown in the picture below.
Apparently the results in the final hdf5 file are dependent on the checkpoint file.

That is another bug related to the use of an absolute path to the checkpoint file directory. Please use relative paths as a work-around. We will fix this issue in the next release.

The first 2 tips especially the relative paths one helped me solve my own issue. Glad I opened this particular topic