FATAL ERROR: number of bytes expected from the BAM bin does not agree with the actual size on disk:

[Malcolm Cook]

Hi,  

STAR quit due to  /tmp  filling up during BAM sorting with a dozen or so files.

Is there some way to avoid having to run my job again and just pick up with where BAM sorting left off?

I'll either use another /scratch on my network (which will be slower) or extend /tmp (which I've just cleaned up).

Below is my full command and full trace, if it helps you to help me recover from this....

Thanks,

Malcolm

nohup STAR \

     --runThreadN 16 \

             --twopassMode Basic \

             --twopass1readsN -1 \

     --outFileNamePrefix germline/STAR/out/ \

             --genomeDir germline/STAR/genome_${ReadLength}  \

     --genomeLoad  NoSharedMemory \

     --runMode alignReads \

     --limitBAMsortRAM 20000000000 \

     --readFilesIn ${paths} \

     --readFilesCommand zcat \

     --outReadsUnmapped Fastx \

     --outSAMtype BAM SortedByCoordinate \

     --outSAMreadID Number \

     --outWigType bedGraph \

     --outWigStrand Unstranded \

     --outWigNorm RPM \

     --outFilterMultimapNmax 1 \

     --outFilterMultimapScoreRange 1 \

     --outSAMprimaryFlag OneBestScore \

             --limitSjdbInsertNsj 2000000 \

             --limitOutSJcollapsed 2000000 \

     --outSJfilterCountUniqueMin 5 2 2 2 \

     --outSJfilterCountTotalMin 999999 999999 999999 999999 \

     --alignSJoverhangMin 10 \

     --alignSJDBoverhangMin

     --outTmpDir $(mktemp -d -u) &

Nov 30 11:44:06 ..... Started STAR run

Nov 30 11:44:06 ..... Loading genome

Nov 30 11:44:21 ..... Started 1st pass mapping

Nov 30 11:45:19 ..... Started STAR run

Nov 30 11:45:19 ..... Loading genome

Nov 30 11:45:34 ..... Started 1st pass mapping

Nov 30 13:44:57 ..... Finished 1st pass mapping

Nov 30 13:45:01 ..... Inserting junctions into the genome indices

Nov 30 13:49:05 ..... Started mapping

Nov 30 16:46:31 ..... Started sorting BAM

EXITING because of FATAL ERROR: number of bytes expected from the BAM bin does not agree with the actual size on disk: 3028299982   1154658304   14

Nov 30 16:46:32 ...... FATAL ERROR, exiting

[Alexander Dobin]

Hi Malcolm,

you cannot restart STAR unfortunately... In principle, you can re-start only the 2-nd pass, by supplying the SJ.out.tab from the 1st pass:

--sjdbFileChrStartEnd _STARpass1/SJ.out.tab and omitting --twopassMode Basic.

/tmp partition usually does not have a lot of space allocated, and it's filled with tmp files from other process.

Why not use the /home partition?

Cheers

Alex