cnr-ibf-pa / hbp-bsp-issues

Ticketing system for developers/testers and power users of the Brain Simulation Platform of the Human Brain Project
4 stars 0 forks source link

Allocation memory problem on Jureca booster when running BluePyOpt #540

Closed clupascu closed 3 years ago

clupascu commented 4 years ago

Expected behavior

BluePyOpt optimization runs without problems with the Python3 modules from issue #533

Actual Behavior

When runnning the optimization with OFFSPRING_SIZE=128 MAX_NGEN=2 I get BrokenProcessPool: [Errno 12] Cannot allocate memory

Steps to reproduce the problem

The same issue does not happen with Python2. The job runs successfully in less than 7 minutes. The working Python2 script can be found at /p/home/jusers/lupascu1/jureca/test_jaro_morph3_young_control/ The log of the working job is this one /p/home/jusers/lupascu1/jureca/test_jaro_morph3_young_control/logs/ipyparallel_8196473.log The sbatch file to run the job is ipyparallel_jureca.sbatch

The not working Python3 script can be found at /p/home/jusers/lupascu1/jureca/test_jaro_morph_young_control_Py3/ The log of the not working job is this one /p/home/jusers/lupascu1/jureca/test_jaro_morph_young_control_Py3/logs/ipyparallel_8196500.log The sbatch file to run the job is ipyparallel_jureca22042020.sbatch

The Python3 script works perfectly though with OFFSPRING_SIZE=10 MAX_NGEN=2 The log is this one /p/home/jusers/lupascu1/jureca/test_jaro_morph_young_control_Py3/logs/ipyparallel_8196701.log The sbatch file to run the job is ipyparallel_jureca22042020_2.sbatch

clupascu commented 4 years ago

@jorblancoa

clupascu commented 4 years ago

@pramodk @jorblancoa any news on this?

jorblancoa commented 4 years ago

Hi @clupascu

We will run some tests next week and contact support if necessary and we will get back to you as soon as possible.

jorblancoa commented 4 years ago

hi @clupascu

Right now Juelich systems are down, but once they are up and running and I would be able to run some test I will let you know.

pramodk commented 3 years ago

I believe this is no longer relevant ticket.