#24 caught std::bad_alloc in seatac step

[Dan Bolser]

I'm running ATAC following the instructions here:
https://sourceforge.net/apps/mediawiki/kmer/index.php?title=Getting_Started_with_ATAC

The process seems to die at the 'seatac' step:

perl $ATACDIR/bin/atac.pl \ > -meryldir $GENOMEDIR/meryl \ > -dir $GENOMEDIR \ > -id1 SL2.40 -seq1 $GENOMEDIR/$(basename $TOM) \ > -id2 ST3.10 -seq2 $GENOMEDIR/$(basename $POT)

/home/dbolser/cxgn/atac/bin/meryl -B -C -threads 2 -m 20 -s /tmp/ATAC/meryl/SL2.40.fasta -o /tmp/ATAC/meryl/SL2.40.ms20
/home/dbolser/cxgn/atac/bin/meryl -v -M lessthanorequal 1 -s /tmp/ATAC/meryl/SL2.40.ms20 -o /tmp/ATAC/meryl/SL2.40.ms20.le1
/home/dbolser/cxgn/atac/bin/meryl -B -C -threads 2 -m 20 -s /tmp/ATAC/meryl/ST3.10.fasta -o /tmp/ATAC/meryl/ST3.10.ms20
/home/dbolser/cxgn/atac/bin/meryl -v -M lessthanorequal 1 -s /tmp/ATAC/meryl/ST3.10.ms20 -o /tmp/ATAC/meryl/ST3.10.ms20.le1

Finding the min count between SL2.40.ms20.le1 and ST3.10.ms20.le1.
/home/dbolser/cxgn/atac/bin/meryl -M min -s /tmp/ATAC/meryl/SL2.40.ms20.le1 -s /tmp/ATAC/meryl/ST3.10.ms20.le1 -o /tmp/ATAC/work/min.
SL2.40.ms20.le1.ST3.10.ms20.le1

includeSize is proportional to 226439128.
excludeSize is proportional to 2479295720.
/tmp/ATAC/work/min.SL2.40.ms20.le1.ST3.10.ms20.le1 has 44103697 mers.
/tmp/ATAC/meryl/SL2.40.ms20 has 737124220 mers.
includeSize is exactly 44103697 mers.
excludeSize is exactly 693020523 mers.

/home/dbolser/cxgn/atac/bin/seatac -verbose -mersize 20 -minlength 20 -maxgap 0 -numthreads 4 -table /tmp/ATAC/meryl/SL2.40.fasta -stream /tmp/ATAC/meryl/ST3.10.fasta -only /tmp/ATAC/work/SL2.40vsST3.10.include -output /tmp/ATAC/work/SL2.40vsST3.10-segment-000.matches -filtername /home/dbolser/cxgn/atac/bin/../lib/filter-heavychains.so -filteropts "-1 SL2.40 -2 ST3.10 -S 100 -J 100000" > /tmp/ATAC/work/SL2.40vsST3.10-000.out 2>&1

Failed to run SL2.40vsST3.10-000

less /tmp/ATAC/work/SL2.40vsST3.10-000.out
Building onlyDB from '/tmp/ATAC/work/SL2.40vsST3.10.include'
potential configurations for approximately 781666164 20-mers (posnW=30).
tblBits=16 shifts=24,12 -- size 5.005GB -- work 0.000 ***
...
tblBits=35 shifts=05,02 -- size 123.275GB -- work 0.000
tblBits=36 shifts=04,02 -- size 243.185GB -- work 0.000
tblBits=25 s1=15 s2=7 -- merSize=20 bits + posnWidth=30 bits (est 781666164 mers) FINAL
Allocated bucket size counting space with total size 131072 KB
2.10 Mmers -- 5.47 Mmers/second^M 4.19 Mmers -- 5.46 Mmers/second^M ...

Found 737124220 mers (max position = 781666391)
Allocated 4139125KB for buckets (529808034 64-bit words)
positionDB()-- caught std::bad_alloc in /data/shared/new_home/dbolser/cxgn/kmer/libkmer/positionDB.C at line 433
positionDB()-- _countingBuckets = new u64bit [529808034]

Is this basically a simple 'out of memory' error, or is there some other problem?

Cheers,
Dan.

[Dan Bolser]

Hmm... Using this tool: https://gist.github.com/1298116

I got the following output:
memusg: peak=288368

[Dan Bolser]

Seems this may have been a memory issue, as it works on a box with more memory. However, it could be related to any number of other issues regarding the different boxes. The failed box was a XEN virtual machine.

[Brian Walenz]

Yes, this looks like out of memory. It's attempting to allocate a block of size 4GB. In your initial report, somewhere in the '...' lines will be a 'tblBits=25' with the estimated memory size needed.

e.g.,

tblBits=36 shifts=04,02 -- size 243.185GB -- work 0.000

wants 243GB for tblBits=36.

tblBits=25 used the smallest amount of memory.

[Dan Bolser]

Thanks for the reply brianwalenz, in future this error will have some google results, which I think will be a big help for others with the same problem.

P.S. How come you think it needs 4 GB when the msg says 243 Gb?

Cheers,
Dan.

[Dan Bolser]

Can I close this now?

[Brian Walenz]

Closed.