Re: Problem with mpi abyss-pe execution

486 views
Skip to first unread message
Message has been deleted

Tony Raymond

unread,
Feb 7, 2013, 1:03:47 PM2/7/13
to <fsanchez1975@gmail.com> <fsanchez1975@gmail.com>, abyss...@googlegroups.com
Hi Friman,

Don't use the mpirun command. abyss-pe will call mpirun when it needs to.

Cheers,
Tony

On 2013-02-05, at 11:11 AM, <fsanch...@gmail.com> <fsanch...@gmail.com> wrote:

Dear all,

I am having a problem executing mpi abyss version. The command I use for the execution is like this:

mpirun /path_of_abyss_binaries/abyss-pe name=my_test np=4 c=10 k=63 in='/path_to_my_fastq_files/file_R1.fastq /path_to_my_fastq_files/file_R2.fastq'

After executing this line I have the following output:

make: Nothing to be done for `/path_to_my_fastq_files/file_R1.fastq /path_to_my_fastq_files/file_R2.fastq'.
make: Nothing to be done for `/path_to_my_fastq_files/file_R1.fastq /path_to_my_fastq_files/file_R2.fastq'.
make: Nothing to be done for `/path_to_my_fastq_files/file_R1.fastq /path_to_my_fastq_files/file_R2.fastq'.
make: Nothing to be done for `/path_to_my_fastq_files/file_R1.fastq /path_to_my_fastq_files/file_R2.fastq'.

It seems like abyss-pe is not taking the second fastq file as another input. 

I have tried things like putting the input in a environtment variable, like this:

export in='/path_to_my_pe_fastq_files/file_R1.fastq /path_to_my_pe_fastq_files/file_R2.fastq'

But it did not work.

I am using abyss-1.3.4 configure like:

cd ../build/abyss-1.3.4

../../sources/abyss-1.3.4/configure --prefix=/path_of_abyss_binaries --with-boost=//path_of_boost_1_49_0/include --enable-maxk=96 --with-mpi=/apps/INTEL/impi/4.1.0.024/intel64

What is the problem?, do you have any suggestion for solving this issue?

Thanks a lot for your help!.
Friman






--
You received this message because you are subscribed to the Google Groups "ABySS" group.
To unsubscribe from this group and stop receiving emails from it, send an email to abyss-users...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

Friman Sánchez

unread,
Feb 8, 2013, 11:07:26 AM2/8/13
to abyss...@googlegroups.com, <fsanchez1975@gmail.com> <fsanchez1975@gmail.com>
Hi Tony,

Thanks a lot for your answer!.

I have tried what you suggested. That is, my execution command now is:

/path_of_abyss_binaries/abyss-pe name=my_test np=4 c=10 k=63 in='/path_to_my_fastq_files/file_R1.fastq /path_to_my_fastq_files/file_R2.fastq'

And the previous problem did not appear again.

How ever I have this other:

mpirun.lsf: LSF_PJL_TYPE is undefined. Exit ...
make: *** [my_test-1.fa] Error 1

So, yes, I am using LSF queue system in the machine. But I do not see what would be the problem.

Now, I did my abyss configuration with:

../../sources/abyss-1.3.4/configure --prefix=/path_of_abyss_binaries/abyss-1.3.4 --with-boost=/path_of_boost_1_49_0/include --enable-maxk=96 --with-mpi=/apps/OPENMPI/1.5.4

And the config.log say:

PATH: /apps/OPENMPI/1.5.4//bin   
...
configure:6297: checking for mpi.h
configure:6297: result: yes
configure:6633: checking for MPI_Init in -lmpi

which is OK.

(Note I have changed to openmpi 1.5.4, which is the default mpi installed in my system).

I have verified that I am using the correct mpi library doing:

which /apps/OPENMPI/1.5.4/bin/mpirun

An also, ldd ABYSS-P                           
        linux-vdso.so.1 =>  (0x00007fff0a527000)
        libmpi.so.1 => /apps/OPENMPI/1.5.4//lib/libmpi.so.1 (0x00002b9295ae9000)
        libdl.so.2 => /lib64/libdl.so.2 (0x00002b9295e83000)
        libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00002b9296087000)
        libm.so.6 => /lib64/libm.so.6 (0x00002b9296391000)
        libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00002b929660b000)
        libc.so.6 => /lib64/libc.so.6 (0x00002b9296821000)
        libpthread.so.0 => /lib64/libpthread.so.0 (0x00002b9296b95000)
        libnsl.so.1 => /lib64/libnsl.so.1 (0x00002b9296db3000)
        libutil.so.1 => /lib64/libutil.so.1 (0x00002b9296fcb000)
        /lib64/ld-linux-x86-64.so.2 (0x00002b92958c8000)


So, I do not know what is the problem..

Do you have any suggestion?..

Thanks again.
Friman Sanchez.

Tony Raymond

unread,
Feb 8, 2013, 7:09:05 PM2/8/13
to Friman Sánchez, abyss...@googlegroups.com
Hi Friman,

It looks like that error is related to your job submission method. Some googling suggests you need to submit with the -a openmpi option of bsub, but  I have no way to test this…

I think your ABySS installation is fine. I may be able to help more if I knew how you were submitting this job to your queue system. Are you just running abyss-pe from the command line?

Thanks,
Tony

Friman Sánchez

unread,
Feb 8, 2013, 7:35:26 PM2/8/13
to abyss...@googlegroups.com, Friman Sánchez
Hi Tony,

I am executing abyss-pe through a script like this:


-------------
#!/bin/bash

# BSUB -J lanza_assemblers.cmd
# BSUB -cwd .
# BSUB -o lanza_assemblers.out
# BSUB -e lanza_assemblers.err
# BSUB -n 2
# BSUB -R"span[ptile=4]"
# BSUB -W 40:00

/home/bsc18/bsc18025/ASSEMBLERS/abyss-1.3.4/bin/abyss-pe name=my_test np=4 c=10 k=63 in='/home/bsc18/bsc18025/inputs/file_R1.fastq /home/bsc18/bsc18025/inputs/file_R2.fastq'
----------------

That is the way the system administrators suggest to send jobs to the machine. Do you see something wrong?

However I will try using the -a openmpi option on bsub.

Thanks a lot,
Friman.

Tony Raymond

unread,
Feb 8, 2013, 8:09:53 PM2/8/13
to Friman Sánchez, abyss...@googlegroups.com
Hi Friman,

Adding #BSUB -a openmpi to the header might work:
http://hpc-dev.cineca.it/content/what-does-message-mpirunlsf-lsfpjltype-undefined-exit-mean

Let me know if it does.

Thanks,
Tony

Friman Sánchez

unread,
Feb 12, 2013, 7:33:10 AM2/12/13
to abyss...@googlegroups.com, Friman Sánchez
Hi Tony, 

I added the BSUB -a openmpi option in the comand and it worked perfectly!.

Thanks a lot for your help.
Friman
Reply all
Reply to author
Forward
0 new messages