I followed the step0 suggestion and mapped my raw fastq to hla_reference. Then I used bam2fq to convert bam to fished_fastq.
I have a pair-end sample so it is fish1.fastq (25.5MB) and fish2.fastq(25.1MB).
Despite razers3 is memory-consuming, even it read these two fastq into RAM, it should only be 50MB. I don't know why all the
32Gb RAM was fully occupied with additional several Gb of SWAP space occupied as well.
It stuck during the process "Creating dataframe" after input reads.
Using STAR to do the same thing did not work. But previously I mapped reads of another sample with hisat2 and subset chr6 bam. It ran fast and did not cause such memory issue.
I know that I should increase my RAM. I just don't understand why small input still requires such large RAM.
I followed the step0 suggestion and mapped my raw fastq to hla_reference. Then I used bam2fq to convert bam to fished_fastq. I have a pair-end sample so it is fish1.fastq (25.5MB) and fish2.fastq(25.1MB).
Despite razers3 is memory-consuming, even it read these two fastq into RAM, it should only be 50MB. I don't know why all the 32Gb RAM was fully occupied with additional several Gb of SWAP space occupied as well.
It stuck during the process "Creating dataframe" after input reads.
Using STAR to do the same thing did not work. But previously I mapped reads of another sample with hisat2 and subset chr6 bam. It ran fast and did not cause such memory issue.
I know that I should increase my RAM. I just don't understand why small input still requires such large RAM.