You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, all temporary file management, in particular that required to have safe creation of files (mv at the end is more transactional) has to be done by the users manually in the code.
In most workflows a simple .tmp suffix is used which turned out to be unsafe in situations in which Roddy is called twice on the same data but cannot recognize that the workflow is already running (e.g. because of the bug that bjobs temporarily does not report all running jobs).
Solution:
Write all files into a temporary folder.
Move the files to their final destinations as soon as the job successfully finished (e.g. in the wrapper).
Implementation idea:
Create a temporary folder with unique name, e.g. in the roddy execution store (but consider how CWL runners like cwlexec or toil solve this issue (they use a directory in the top-level)
Within the temp-folder all files are created with the same subdirectory structure as in the final folder.
After the script finished, let the wrapper move the content of the temp-folder to the top of the output directory.
The text was updated successfully, but these errors were encountered:
Currently, all temporary file management, in particular that required to have safe creation of files (mv at the end is more transactional) has to be done by the users manually in the code.
In most workflows a simple .tmp suffix is used which turned out to be unsafe in situations in which Roddy is called twice on the same data but cannot recognize that the workflow is already running (e.g. because of the bug that bjobs temporarily does not report all running jobs).
Solution:
Implementation idea:
The text was updated successfully, but these errors were encountered: