[Bioclusters] Re: new on using clusters: problem running mpiblast (2)

[Zhiliang Hu]

Hi Zhao - Thanks for providing more clues to solve my problem!

I have actually gone through all these steps to match up the right number 
of formated sequence database segments, the number of processes to use, 
having the excutables and sequence database on shared nsf file paths 
mounted to all nodes, use the correct paths in ~/.ncbirc, and try out with 
fewest but necessary options, and so on.

While the previous trials were using a mpiblast seemingly compiled 
correctly, when I tried to make sure the compile/run are using the same 
mpi sets, I tried to re-compile as

> export CC=/opt/openmpi.gcc/bin/mpicc
> export CPP=/opt/openmpi.gcc/bin/mpicxx
> ./configure --with-ncbi=/home/share/src/ncbi --prefix=/opt/local
   --with-mpi=/opt/openmpi.gcc

-- this time the compilation failed with errors:
http://www.animalgenome.org/~hu/share/tmp/config.log
http://www.animalgenome.org/~hu/share/tmp/config2.log

The trials were with OpenMPI-1.2.1 and 1.2.3 --are these version 
compatibility problems?  Is there a known MPI version that worked with 
MPIblast 1.4.0?

Zhiliang


On Thu, 6 Sep 2007, Zhao Xu wrote:

> Date: Thu, 6 Sep 2007 09:08:54 +0800
> From: Zhao Xu <xuzh.fdu at gmail.com>
> Reply-To: HPC in Bioinformatics <bioclusters at bioinformatics.org>
> To: bioclusters at bioinformatics.org
> Subject: [Bioclusters] Re: new on using clusters: problem running mpiblast (2)
> 
> Hi Zhiliang,
> Firstly, I suggest that you run mpiBlast with the default parameters, only
> use it with -p -i -d and -o.  If the default parameter works fine, you
> should check the system environments: is there any memory limitation? write
> access? etc.
>
> If the default parameter can not work either, I think you should check:
> 1. How many segements did you format the database? Is the number 12? Make
> sure you run mpiBlast with two more processes.
> 2. Can every nodes read in the input sequences?
> 3. Check your ~/.ncbirc, make sure each node has ~/.ncbirc and each node can
> access the directorys written in ~/.ncbirc.
>
> Regards,
> XU Zhao
> _______________________________________________
> Bioclusters maillist  -  Bioclusters at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters
>