[maker-devel] maker MPI

Fri Feb 15 10:15:53 MST 2013

It cannot establish the communication ring on LAMMPI, so it is acting like
multiple independent instances of MAKER up to the number of processors
specified to MPI.  So it gives the warning "MAKER already running" when it
detects the other job's file locks.

I'd recommend to just use MPICH2, as I don't have any idea what might
happen with LAMMPI.

--Carson

On 13-02-15 12:10 PM, "Kathryn E. Bushley"
<bushleyk at science.oregonstate.edu> wrote:

>Hello Carson,
>
>Thanks.  We currently have either LAMMPI or MPICH2 up and running on
>our cluster so are trying to make it work in either one of those.
>We're able to get it run in LAMMPI but get this error about the MAKER
>EVALUATOR already running, any idea what that might be?  It did work
>previously in MPICH2 so maybe we need to go that route.
>
>[bushleyk at fungi0 Test]$ /usr/lib64/lam/bin/lamboot
>/home/bpp/bushleyk/MPI2/fungi0_machinefile.txt
>
>LAM 7.1.4/MPI 2 C++/ROMIO - Indiana University
>
>[bushleyk at fungi0 Test]$ lamnodes
>n0      fungi0.cgrb.oregonstate.local:4:origin,this_node
>[bushleyk at fungi0 Test]$ ps -aef | grep lam
>bushleyk 11818     1  0 08:56 ?        00:00:00
>/usr/lib64/lam/bin/lamd -H 192.168.1.136 -P 36112 -n 0 -o 0
>bushleyk 11821 11262  0 08:56 pts/1    00:00:00 grep lam
>sullichr 13795     1  0 Feb14 ?        00:00:00
>/usr/lib64/lam/bin/lamd -H 192.168.1.136 -P 57579 -n 0 -o 0
>[bushleyk at fungi0 Test]$ /usr/lib64/lam/bin/mpirun -np 4
>/local/cluster/maker/bin/mpi_maker
>ERROR: Cannot start process. MAKER/EVALUATOR already running
>with different settings in this same directory.
>
>
>FATAL ERROR
>--------------------------------------------------------------------------
>---
>It seems that [at least] one of the processes that was started with
>mpirun did not invoke MPI_INIT before quitting (it is possible that
>more than one process did not invoke MPI_INIT -- mpirun was only
>notified of the first one, which was on node n0).
>
>mpirun can *only* be used with MPI programs (i.e., programs that
>invoke MPI_INIT and MPI_FINALIZE).  You can use the "lamexec" program
>to run non-MPI programs over the lambooted nodes.
>--------------------------------------------------------------------------
>---
>Received SIGTERM
>Received SIGTERM
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>Best,
>
>Kathryn
>
>
>"To err is human...to really foul up requires the root password."
>
>
>Kathryn E. Bushley
>Postdoctoral Researcher
>2082 Cordley Hall
>Department of Botany and Plant Pathology
>Oregon State University
>Corvallis, OR 97331
>office: (541) 737-5284
>cell: (541) 908-0116
>
>
>
>
>Quoting Carson Holt <carsonhh at gmail.com>:
>
>> LAMMPI is pretty much dead, even their website recommends using OpenMPI
>> instead.
>>
>> For installing MAKER with OpenMPI you need to set a couple of
>> environmental variables before MAKER setup.
>>
>> export OMPI_MCA_mpi_warn_on_fork  0
>> export LD_PRELOAD /software/openmpi-1.4.3/lib/libmpi.so:$LD_PRELOAD
>>
>> These not only need to be set before compilation, but also before any
>>run
>> (so add them to you ~/.bashrc or ~/.bash_profile or any module load
>> scripts thanks).  The LD_PRELOAD statement needs to be set for any
>>program
>> using OpenMPI's shared libraries and not just MAKER, so it's normally a
>> good idea to have that set system wide for all users.  The detail can be
>> Found in the OpenMPI documentation.  Note sometimes system library
>>updates
>> can break OpenMPI's shared libraries while not breaking OpenMPI itself,
>>so
>> you might also need to recompile OpenMPI if it has broken shared
>>libraries.
>>
>> Thanks,
>> Carson
>>
>>
>> On 13-02-04 6:48 PM, "Kathryn E. Bushley"
>> <bushleyk at science.oregonstate.edu> wrote:
>>
>>> Greetings,
>>>
>>> We are reinstalling maker after a system upgrade and having some
>>> issues with getting MPI version running.  One question, will maker
>>> only run in MPI-CH2 or will it also run in LAM-MPI?  When trying to
>>> run in MPI-CH2, we get this error:
>>>
>>> /local/cluster/mpich2-1.2.1p1/bin/mpdboot -f
>>> /home/bpp/bushleyk/MPI2/fungi0_machinefile.txt
>>> cd /pseudospace2/bushleyk/MAKER/Testrun;
>>> /local/cluster/mpich2-1.2.1p1/bin/mpiexec -machinefile
>>> /home/bpp/bushleyk/MPI2/fungi0_machinefile.txt -n 4
>>> '/local/cluster/MAKER_2.10/bin/mpi_maker'
>>> /usr/bin/perl: symbol lookup error:
>>> 
>>>/local/cluster/MAKER_2.10/bin/../perl/lib/auto/Parallel/MPIcar/MPIcar.so
>>>:
>>>
>>> undefined symbol: Perl_Tstack_sp_ptr
>>> /usr/bin/perl: symbol lookup error:
>>> 
>>>/local/cluster/MAKER_2.10/bin/../perl/lib/auto/Parallel/MPIcar/MPIcar.so
>>>:
>>>
>>> undefined symbol: Perl_Tstack_sp_ptr
>>> /usr/bin/perl: symbol lookup error:
>>> 
>>>/local/cluster/MAKER_2.10/bin/../perl/lib/auto/Parallel/MPIcar/MPIcar.so
>>>:
>>>
>>> undefined symbol: Perl_Tstack_sp_ptr
>>> /usr/bin/perl: symbol lookup error:
>>> 
>>>/local/cluster/MAKER_2.10/bin/../perl/lib/auto/Parallel/MPIcar/MPIcar.so
>>>:
>>>
>>> undefined symbol: Perl_Tstack_sp_ptr
>>> make: *** [maker] Error 127
>>>
>>> which maybe means it's not finding the right perl?
>>>
>>>
>>> When trying to run in LAM-MPI we get this error, was wondering what
>>> the MAKER/EVALUATOR is?
>>>
>>> ERROR: Cannot start process. MAKER/EVALUATOR already running
>>> with different settings in this same directory.
>>>
>>>
>>> FATAL ERROR
>>> running  repeat masker.
>>> #--------- command -------------#
>>> Widget::RepeatMasker:
>>> /local/cluster/spatafora/RepeatMasker/RepeatMasker
>>> 
>>>/pseudospace2/bushleyk/MAKER/Testrun/dpp_contig.maker.output/dpp_contig_
>>>da
>>> 
>>>tastore/05/1F/contig-dpp-500-500//theVoid.contig-dpp-500-500/contig-dpp-
>>>50
>>> 0-500.0.all.rb -species all -dir
>>> 
>>>/pseudospace2/bushleyk/MAKER/Testrun/dpp_contig.maker.output/dpp_contig_
>>>da
>>> tastore/05/1F/contig-dpp-500-500//theVoid.contig-dpp-500-500 -pa
>>> 1
>>> #-------------------------------#
>>> 
>>>------------------------------------------------------------------------
>>>--
>>> ---
>>> It seems that [at least] one of the processes that was started with
>>> mpirun did not invoke MPI_INIT before quitting (it is possible that
>>> more than one process did not invoke MPI_INIT -- mpirun was only
>>> notified of the first one, which was on node n0).
>>>
>>> mpirun can *only* be used with MPI programs (i.e., programs that
>>> invoke MPI_INIT and MPI_FINALIZE).  You can use the "lamexec" program
>>> to run non-MPI programs over the lambooted nodes.
>>> 
>>>------------------------------------------------------------------------
>>>--
>>> ---
>>> Received SIGTERM
>>> make: *** [maker_new] Error 252
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Best,
>>>
>>> Kathryn
>>>
>>>
>>> "To err is human...to really foul up requires the root password."
>>>
>>>
>>> Kathryn E. Bushley
>>> Postdoctoral Researcher
>>> 2082 Cordley Hall
>>> Department of Botany and Plant Pathology
>>> Oregon State University
>>> Corvallis, OR 97331
>>> office: (541) 737-5284
>>> cell: (541) 908-0116
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> maker-devel mailing list
>>> maker-devel at box290.bluehost.com
>>> http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
>>
>>
>>
>
>