compiler crash when running multiple processes within the same directory
hello,
we started to use FEniCS on our scientific cluster recently. at the moment we do not run any parallel jobs, but we just want to do some parameter studies.
therefore multiple single-threaded simulations are started simultaneously on a certain node. Starting only a single simulation works well, but there seem to be some race conditions if multiple simulations are started.
what is the default method to prevent these kind of problems?
I tried to install locking via:
pip install flufl.lock
This should fix the NFS locking issue, right? or am i missing some configuration commands?
Unfortunately this does not solve the problem. It seems that some files are created in the actual working directory (which is shared between all processes). Is there are way to use seperate directories (like INSTANT_CACHE_DIR) also for these temporal files?
The last solution that I see is to duplicate the complete working directory for each process!?
i am not sure if this is a bug, or something is missing in my setup thanks for any advice greetings Florian
