rronen
commented
9 years ago Update: also tried with --genomeSAindexNbases 4 as I saw on some forums/lists it might help resolve such issues. No luck.
Closed rronen closed 9 years ago
rronen
commented
9 years ago Update: also tried with --genomeSAindexNbases 4 as I saw on some forums/lists it might help resolve such issues. No luck.
rronen
commented
9 years ago Sorry, just noticed you may have fixed this on 2.4.0i.
2.4.0i 01/14/2015
Fixed a bug with the _STARtmp temporary directory name for the 2-pass runs.
Fixed a bug causing seg-faults for genome generation.
Fixed a bug causing seg-faults for --quantMode TranscriptomeSAMI got the current (static) executable & running no -- will close this as soon as that works (hopefully).
rronen
commented
9 years ago Nope, getting the exact same error with latest executable:
ubuntu@master:~/bcbio_datadir/genomes/Hsapiens/GRCh37/star$ ./STAR --genomeDir /home/ubuntu/bcbio_datadir/genomes/Hsapiens/GRCh37/seq/../star --genomeFastaFiles /home/ubuntu/bcbio_datadir/genomes/Hsapiens/GRCh37/seq/GRCh37.fa --runMode genomeGenerate --sjdbOverhang 99 --sjdbGTFfile /home/ubuntu/bcbio_datadir/genomes/Hsapiens/GRCh37/seq/../rnaseq/ref-transcripts.gtf --genomeSAindexNbases 14 --runThreadN 16 --limitGenomeGenerateRAM 30000000000
Mar 20 18:44:13 ..... Started STAR run
Mar 20 18:44:13 ... Starting to generate Genome files
Mar 20 18:44:24 ... Starting GTF processing
Mar 20 18:44:42 ... Finished GTF processing
Mar 20 18:45:18 ... finished processing splice junctions database ...
Mar 20 18:45:36 ... starting to sort Suffix Array. This may take a long time...
Mar 20 18:45:55 ... sorting Suffix Array chunks and saving them to disk...
Mar 20 18:58:56 ... loading chunks from disk, packing SA...
terminate called after throwing an instance of 'std::bad_alloc'
what(): std::bad_alloc
Aborted (core dumped)
ubuntu@master:~/bcbio_datadir/genomes/Hsapiens/GRCh37/star$ du -h .
4.0K ./fail_11am/_STARtmp
47G ./fail_11am
4.0K ./_STARtmp
94G .
ubuntu@master:~/bcbio_datadir/genomes/Hsapiens/GRCh37/star$
ubuntu@master:~/bcbio_datadir/genomes/Hsapiens/GRCh37/star$ ./STAR --version
STAR_2.4.0j_modifiedAny suggestions would be most welcome.
Thanks, Roy
alexdobin
commented
9 years ago Hi Roy,
how much RAM do you have available? For this genome you need at least 30GB.
Cheers Alex
rronen
commented
9 years ago I think I have exactly that. Do you expect this is causing the issues?
ubuntu@master:/bcbio_work/ubuntu/encrypted/STAR_index$ cat /proc/meminfo
MemTotal: 30610040 kB* this is an AWS/EC2 instance type c3.4xlarge
alexdobin
commented
9 years ago I think this may not be enough, since it's 29GB. It will probably be enough for mapping, but genome generation requires 1-2GB more RAM than mapping. Do you have a server with slightly more RAM?
rronen
commented
9 years ago I see, yes I can manage something larger. Will give it a spin with more RAM shortly. Thanks!
Update: this was indeed the issue. It finished without problems on a larger server.
Hi Alex,
I am preparing STAR for use and getting this failure on genomeGenerate. I've tried twice, the second time adding
--limitGenomeGenerateRAM 30000000000but this doesn't seem to help. It failed both times having written ~47Gb in the genomeDir and seemed to be using ~19Gb or RAM most of the time (though I'm not sure exactly how much it used prior to the crash).Any help resolving this would be much appreciated!
Thanks, Roy
Some extracts from Log: