Folks
I am having trouble getting a CuffDiff run to successfully complete. It runs
24-28 hours then gets killed by my LSF scheduler with an exit code of 137
which translates to 9 or job memory issues.
The accounting info is below
Accounting information about this job:
Share group charged </hazards>
CPU_T WAIT TURNAROUND STATUS HOG_FACTOR MEM SWAP
736582.94 5 107561 exit 6.8480 23214M 26089M
The job is getting killed because its asking for 26G of swap which is not available.
I have tried a series of runs halving the --max-bundle-frags from the default 1000000 (--max-bundle-frags 250000 is the most recent failure ) . I am looking at 19 conditions at once because the researchers asked for an all vs all comparison. The size of the 19 conditions ranges from 60-128 million reads(single end actually). Both Tophat141/Bowtie0.12.9 and Tophat2.0.7/Bowtie0.12.9 align the files just fine. No errors with CuffMerge
I am using CuffLinks 2.0.2 and using -p 16. My compute nodes have 24G ram and 2G swap.
Are there other options I could evaluate to try to get CuffDiff to ask for less memory?
I am having trouble getting a CuffDiff run to successfully complete. It runs
24-28 hours then gets killed by my LSF scheduler with an exit code of 137
which translates to 9 or job memory issues.
The accounting info is below
Accounting information about this job:
Share group charged </hazards>
CPU_T WAIT TURNAROUND STATUS HOG_FACTOR MEM SWAP
736582.94 5 107561 exit 6.8480 23214M 26089M
The job is getting killed because its asking for 26G of swap which is not available.
I have tried a series of runs halving the --max-bundle-frags from the default 1000000 (--max-bundle-frags 250000 is the most recent failure ) . I am looking at 19 conditions at once because the researchers asked for an all vs all comparison. The size of the 19 conditions ranges from 60-128 million reads(single end actually). Both Tophat141/Bowtie0.12.9 and Tophat2.0.7/Bowtie0.12.9 align the files just fine. No errors with CuffMerge
I am using CuffLinks 2.0.2 and using -p 16. My compute nodes have 24G ram and 2G swap.
Are there other options I could evaluate to try to get CuffDiff to ask for less memory?
