[BiO BB] mpiblast error.

[Joe Landman]

Hi Sravan:

   I just rebuilt the RPM for LAM.  Have a look at 
http://downloads.scalableinformatics.com/downloads/mpiblast/mpiblast-1.3.0-2LAM.x86_64.rpm 
  though I haven't had a chance to test it in any great depth.

   Looks like new ncbi bits are out, so I will build some RPMs of this 
soon as well.


Joe

Joe Landman wrote:
> Hi Sravan:
> 
>   Since you are using the RPMs we built, maybe I can help.
> 
> P Sravan Kumar. PAsst wrote:
> 
>> We need assistance from the users of mpiblast regarding the execution on
>> AMD64 grid environment.
>>
>> 1.System Configuration:
>> ---------------------
>> Operating System:
>> Red Hat Enterprise Linux 3.0 for AMD 64 Architecture
>> Sun Fire v20z machine
>> Bio Clusture Grid Engine
>> 2. We have downloaded the NCBI-2.2.10-1.x86_64.rpm and
>> mpiblast-1.3.0-1.x86_64.rpm from scalableinformatics.com. When we
>> installed NCBI: all the binaries like blastall are loaded onto /usr/bin.
>> and mpiblast: all the binaries are loaded onto /usr/local/bin ex:
>> mpiblast, mpiformatdb.
> 
> 
> Good.  These binaries are built against mpich-1.2.6.  I have not done a 
> build against LAM yet, though I would like to.  This is relevant a 
> little later on in the error message portion.
> 
>> 3. We have created one common blastdb in /apps/Databases/blastdb/ using
>> mpiformatdb -i scaffolds --nfrags=250 N 250 -p F. ( 250 fragments have
>> been created).
> 
> 
> Good.
> 
>> 4. Lamboot and lamnodes are also working fine as we checked by running
>> clustalmpi.sh through mpirun command.
> 
> 
> Since mpiblast (from the RPM) is not using LAM, this is going to be an 
> issue when you try to run it.
> 
>>
>> 5. But we are unable to run mpiblast through mpirun script as it gives 
>> the
>> follwing error:
>>   ( I am attaching a file containing the script and the error message ).
> 
> 
> execution: ./mpiblast.sh on the masternode (silk-node 7 n0)
> Wed May 18 14:29:55 IST 2005
> Sorry, mpiBLAST must be run on 3 or more nodes
> [0] MPI Abort by user Aborting program !
> [0] Aborting program!
> 
> Yes.  Makes sense (oddly enough), as this version of the mpiblast RPM 
> was built against mpich-1.2.6.  I would be happy to build against LAM (I 
> have 7.0.3, 7.0.6, and 7.1.1 installed on my build machine), though it 
> might take a little while.
> 
> Unfortunately, MPI != MPI.  I cannot take a LAM compiled binary and run 
> it on an MPICH system, nor can I take an MPICH compiled binary and run 
> it on an LAM system.  I would like us to get there, but this requires an 
> MPI ABI, and a few other bits (agreement on how MPI processes are 
> started, etc).
> 
>> Our query:
>> ---------
>> 1. Why is this happening?
> 
> 
> See above.
> 
>> 2. Is there an NCBI.TAR.GZ pacakge that supports AMD64(X86_64) for RH
>> Enterprise WS 3.0?
> 
> 
> ???
> 
>> 3. While installing NCBI rpms we were able to find only binary files. We
>> did not find the data files of NCBI ex: blosum, pam etc in any path in
>>  Is there any specific path they may be located?
> 
> 
> I did not include them in the RPM.  Would you like a separate RPM with 
> the data files?
> 
>>     Please assist us in proper integration and running of mpiblast
>> tool. We appreciate your early suggestions.
> 
> 
> Joe
> 
>>
>> Regards.
>> Sravan.
>>
>>
>>
>>
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Bioinformatics.Org general forum  -  
>> BiO_Bulletin_Board at bioinformatics.org
>> https://bioinformatics.org/mailman/listinfo/bio_bulletin_board
> 
> 

-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://www.scalableinformatics.com
phone: +1 734 786 8423
fax  : +1 734 786 8452
cell : +1 734 612 4615