Hi,
I've been trying to run MarkDuplicates on a shared cluster and I keep getting the job killed because of swap usage. I'm trying to figure out a way to use MarkDuplicates without just throwing tonnes of ram at it. To get more ram, I need to request more cores which means lower priority in the queuing system.
I've been using the command:
java -Xmx40g -Djavaio.tmpdir=/home/gowens/tmp -jar $picard MarkDuplicates I=${name}.bam O=${name}.dup.bam M=${name}.dupmetrics.txt CREATE_INDEX=TRUE TMP_DIR=/home/gowens/tmp MAX_RECORDS_IN_RAM=5000000;
I've tried changing MAX_RECORDS_IN_RAM up or down but I'm not sure if thats helping or hurting the problem. For reference, the BAM files I'm marking are 2G to 23G in size and the issue is with the ones on the larger size.