Hi, I'm stuck on trying to index genome references for gatk. From what i see it happens (no space left) when the script try to create dictionnary (even it being small..). Anyway, using java, i tried the usual arguments, without success and now i don't know what to do/where to look anymore. For other picard tools, i just overwrote some arguments through job_conf.xml so for a start i added this one to the destination. After that, i've looked in the python script and tried to add to it the same arguments (previous attempt had been unsuccesful). Any ideas about "how to define a directory other than /tmp" ? Our cluster is made with small tmp dirs on all its nodes. The actual attempts are : ------------------ data_manager_gatk_picard_index_builder.py line 40 : args = [ 'java', '-Xmx5120m', '-Djava.io.tmpdir=/local', '-jar', jar, 'R=%s' % gatk_sorted_fasta_filename, 'O=%s.dict' % sequence_id ] ------------------ job_conf.xml : <destination id="java_heavy" runner="drmaa"> <param id="nativeSpecification">-p kosmos --mem-per-cpu=5120 --ntasks-per-node=4 --nodes=1</param> <env id="_JAVA_OPTIONS">-Xmx5120m -Djava.io.tmpdir=/local</env> </destination> ------------------ Regards, Laurent Fournier. --- Laurent FOURNIER IT Specialist University of Liège, GIGA +1 laurent.fournier@ulg.ac.be +32 4 366 34 53 +32 497 87 96 46 www.giga.ulg.ac.be