[maker-devel] maker mpi failed

[Carson Holt]

I don't think you will need to reinstall perl (you may but I don't think
so). MVAPICH2 doesn't work because they have overridden malloc with their
own version which causes weirdness with shared libraries in Perl.  I've
submitted some bug reports to the MVAPICH2 developers and made modifications
to the development version of MAKER, and had some progress, but I am still
not getting reliable behavior in MVAPICH2 when compiling for the
OFA-IB-Nemesis or OFA-IB-CH3 interfaces.  So for now using MPICH2 is your
best choice.  You lose some of the infiniband optimizations but IO rather
than messages passing is MAKER's true bottleneck (so you won't get a
performance hit by using MPICH2).

--Carson



From:  "Dian \"Oscar\" Jiao" <jiao at tacc.utexas.edu>
Date:  Thursday, 4 October, 2012 5:20 PM
To:  Carson Holt <carsonhh at gmail.com>, "maker-devel at yandell-lab.org"
<maker-devel at yandell-lab.org>
Subject:  Re: [maker-devel] maker mpi failed

Hi Carson,
I didn't know maker does not work with MVAPICH2. The cluster I was
installing maker on only has mvapich2. So I guess I will have to install
mpich2 under my directory. However, I installed perl threading based on
mvapich2. Does that mean I need to reinstall perl with mpich2 before
reinstalling maker?
Oscar
-- 
Dian (Oscar) Jiao, Ph.D.,
Research Associate, Life Sciences Computing Group
Texas Advanced Computing Center
University of Texas at Austin
jiao at tacc.utexas.edu | (832) 303-1166

From: Carson Holt <carsonhh at gmail.com>
Date: Thursday, October 4, 2012 3:53 PM
To: Oscar Jiao <jiao at tacc.utexas.edu>, "maker-devel at yandell-lab.org"
<maker-devel at yandell-lab.org>
Subject: Re: [maker-devel] maker mpi failed

You either tried to run with the -dsindex option before the log file existed
or you tried to launch via MPI when not compiled for MPI or not compiled
correctly.  It's as if you called MAKER serially multiple times in a row
which you can do, but if you start too many before one of them gets a chance
to get the log file started a race condition exists.

Which MPI flavor are you using?  MVAPICH2 won't work with MAKER, MPICH2
will, and OpenMPI may or may not (non-robust shared library behavior).

You may have to reinstall MPICH2.

MAKER can try and do this for you using necessary flags to make it easier
To do this -->
cd ./maker/src
./Build mpich2

Then reinstall maker
cd ./maker/src
perl Build.PL
<say yes to mpi question>
./Build clean
./Build install

Then use this executable to launch --> ./maker/exe/mpich2/bin/mpiexec

Note you cannot use other launchers.  You have to use the one you compiled
with.

--Carson



From: "Dian \"Oscar\" Jiao" <jiao at tacc.utexas.edu>
Date: Thursday, 4 October, 2012 4:42 PM
To: "maker-devel at yandell-lab.org" <maker-devel at yandell-lab.org>
Subject: [maker-devel] maker mpi failed

Hi, I was trying to run maker 2.26. It was compiled with MPI. The non-mpi
executable works just fine. But I got the error below while running mpirun
­n maker ŠŠ

What is this Iterator::Fasta::skip_file? Any idea what is going on here?

Oscar

STATUS: Processing and indexing input FASTA files...
STATUS: Setting up database for any GFF3 input...
A data structure will be created for you at:
/work/02045/jiao/modules/Maker-GMOD/maker2.26/data/dpp_contig.maker.output/d
pp_contig_datastore

To access files for individual sequences use the datastore index:
/work/02045/jiao/modules/Maker-GMOD/maker2.26/data/dpp_contig.maker.output/d
pp_contig_master_datastore_index.log

ERROR: Log file does not exist in Iterator::Fasta::skip_file
MPI process (rank: 8) terminated unexpectedly on
c341-213.ls4.tacc.utexas.edu
Exit code -5 signaled from c341-213

_______________________________________________ maker-devel mailing list
maker-devel at box290.bluehost.com
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20121004/a3702f68/attachment-0003.html>