Blast+ Error Code=12
1
0
Entering edit mode
12.4 years ago
Amr ▴ 160

Something really bizarre is going on. I ran 6 tblastn jobs independently using the following command line entry:

tblastn -query blastqueries*14920.fasta -out blastqueries*14920.fasta.out -db refseq*genomic -outfmt 5 -soft*masking True -show_gis -seg yes &

after repeating this 6 times for all my input files, a quick look at the directory showed the output files had been created as expected, with a size of 0 KB.

in the top list, I saw all 6 of my tblastn jobs running happily at 100% cpu usage each.

after about a day, 2 of them are no longer in the list of active processes, and yet all 6 out files are still empty.

No error messages in case you were wondering.

...so any ideas on where the heck the results are? if there are any? aah!

UPDATE

I am running tblastn 2.2.25 against a local copy of the blast db refseq_genomic. computer is a macpro with 2 quad-core intel xenon processors, 16 GB of RAM.

The blast+ that I downloaded was the disk image version, don't know it tht means that it is 32 bit only??

Okay so I now know more about whats going on!

The rest of my runs also died, this time I was around to notice the full error:

set a breakpoint in malloc*error*break to debug
tblastn(32920) malloc: *** mmap(size=1496113152) failed (error code=12)

Someone else has posted the same problem - Problem Running Blast Jobs.

blast • 3.9k views
ADD COMMENT
1
Entering edit mode

check with a baby fasta file and see if it generates the outputs...

ADD REPLY
1
Entering edit mode

your disk might be full! happened to me...

ADD REPLY
1
Entering edit mode

a bus error is a memory error, I've seen that happen when running an executable on a platform that was not designed for: 64bit application on a 32 bit platform

ADD REPLY
0
Entering edit mode

yea it does... i didnt that a million times before running my big job. i should say, each of my runs is querying over 8000 sequences...

ADD REPLY
0
Entering edit mode

yea it does... i did that a million times before running my big job. i should say, each of my runs is querying over 8000 sequences...

ADD REPLY
0
Entering edit mode

nope - disk is not full, but a few minutes ago the terminal spat out a bus error not sure if that means that blast died

ADD REPLY
0
Entering edit mode

Thanks Istvan but thats not it either..

ADD REPLY
0
Entering edit mode

1 up to all of you for giving it a go though!

ADD REPLY
0
Entering edit mode

So you've split your massive job into six sections that take a day (or more) to run. My suggestion is to split it into many more sections, each of which is smaller (on the same scale as what you tested). BLAST parallelizes beautifully on the level of input sequences, so there's no real reason to some massive job all at once. A simple shell script can immediately queue up the next job when one job finishes...

ADD REPLY
0
Entering edit mode

So you've split your massive job into six sections that take a day (or more) to run. My suggestion is to split it into many more sections, each of which is smaller (on the same scale as what you tested). BLAST parallelizes beautifully on the level of input sequences, so there's no real reason to submit a massive job all at once. A simple shell script can immediately queue up the next job when one job finishes...

ADD REPLY
0
Entering edit mode

But 8000 query seqs is not that much. I ran over 80,000 seqs in one file against NR using blastx recently, took a few days (only one segfault on the way ...) but worked after resuming it there.

ADD REPLY
1
Entering edit mode
12.4 years ago
Amr ▴ 160

Okay so im answering my own question but I welcome anyone with some real insight:

There is no way to figure out what went wrong without knowing why the Bus error occurred, which would take some understanding of hardware and also how blast+ accesses memory. (ive read the paper which describes this, alas no luck)

perhaps its a bug

ADD COMMENT
1
Entering edit mode

Include the version of blast+, how you built it if you did, the output of uname -a, and memory size of your server.

ADD REPLY
0
Entering edit mode

I think sending a bug report to NCBI might be the best option.

ADD REPLY
0
Entering edit mode

okay figured it out i think: it was a 32 bit version of blast running on 64 bit platform, searching a gigantic database. not a good idea apparently..

ADD REPLY

Login before adding your answer.

Traffic: 3228 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6