Hi, I have developed a parallel version of a cellular automata but when the programm ends the logical process continue to remain active as the programm didn’t ends. Why?
Please share an MWE.
Sorry, I find the solution, I must use the @sync because I thought that the programm end but it didn’t ends. Thanks.
@dsHitman I have built and managed HPC systems for a long time. This is a common problem - Matlab I am looking at you!
I have not come across it with Julia and I would like to help.
There are several ways to solve this problem:
a) make sure the Julia job exits properly - of course
b) a post job script is run for every job which cleans up orphaned processesw
c) running in cgroups - which is actually what containers are - you run in a cgroup and cancel the cgroup at the end of the job. All processes are terminated.
Could you please tell us exactly how you are running - do you have a batch system such as PBS or Slurm?
Also do the following when a job is running and also when a job has finished,
where username is your username
ssh into a compute node
ps -eaf --forest | grep username
Thanks for the answer but, as I write before, I find the solution using the @sync macro. Thanks again for the interest.