OpenBLAS errors when using `gemini load` under SGE cluster

[sndrtj]

Hi there,

I'm having trouble loading in VCF files under an SGE cluster. Consider the following script:

gemini load -v the_vcf.vcf --cores 8 -t VEP the_database.db

When I run this with qsub -S /bin/bash -cwd -pe <THE_PE> 8 the_script.sh I invariably get the following error, after which the job dies:

OpenBLAS: pthread_creat error in blas_thread_init function. Error code:11

When I use the buit-in SGE scheduler with the following:

gemini load -v the_vcf.vcf --cores 8 -t VEP --queue the_queue.q --scheduler sge the_database.db

... I get the same error in some of the jobs, though not all. The scheduler gives a bunch of Attempting to load config file ipcontroller_config.py in path /home/<user>/.ipython/profile_<some_profile> messages. In the job error files for the jobs which did not give the OpenBLAS error I can see errors like:

2016-01-12 11:18:32.639 [IPEngineApp] Using existing profile dir: u'/home/<user>/.ipython/profile_b722ea62-b915-11e5-9dbe-f8bc12452266'
2016-01-12 11:18:32.865 [IPEngineApp] WARNING | url_file u'/home/<user>/.ipython/profile_b722ea62-b915-11e5-9dbe-f8bc12452266/security/ipcontroller-1d287613-ded1-4101-a706-11a900f086fe-engine.json' not found
2016-01-12 11:18:32.866 [IPEngineApp] WARNING | Waiting up to 960.0 seconds for it to arrive.

The jobs are scheduled, but do not appear to be running. No database file appears.