[Bioclusters] Re: new on using clusters: problem running mpiblast (3)

[Zhiliang Hu]

I hope this is a better post to the list :)

To put a long story short, Joe Landman of ScalableInformatics kindly 
coached me to setup a mpiblast.conf file which contains 3 lines:
------------------------------
/raid/pub/ncbi/blast/mpidb
/tmp
/home/src/ncbi/data
------------------------------

My .ncbirc contains:
-------------------------
[NCBI]
Data="/home/src/ncbi/data"

[BLAST]
BLASTDB="/raid/pub/ncbi/blast/mpidb"
BLASTMAT="/home/src/ncbi/data"

[mpiBLAST]
Shared="/raid/pub/ncbi/blast/mpidb"
Local="/tmp"
-----------------------

I ran MPIblast using OpenMPI v.1.2.3 as in:

bash> /opt/openmpi.gcc/bin/mpirun -np 3 -machinefile ./machines
       /home/local/bin/mpiblast --config-file=/home/zhu/.mpiblast.conf
       -p blastp -i ./bait.fasta -d ecoli.aa

and got following errors:

[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: P 
is not a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
BLOSUM80 is a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
BLOSUM62 is a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
BLOSUM50 is a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
BLOSUM45 is a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
PAM250 is a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
BLOSUM62_20 is a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
BLOSUM90 is a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
PAM30 is a supported matrix
[blastall] ERROR: gi|1786184|gb|AAC73114.1|: BlastKarlinBlkGappedCalc: 
PAM70 is a supported matrix
2       0.149567        Bailing out with signal 11
[node002:24414] MPI_ABORT invoked on rank 2 in communicator MPI_COMM_WORLD 
with errorcode 0

--- It hangs here for hours although I had only a few query sequences 
against a very small database (4000 sequences).  I had to ctr-C or kill 
the job.

To be sure, I have also put in my .bashrc following lines:

   export BLASTMAT="/home/src/ncbi/data"

and did "source ~/.bashrc", then re-run the mpiblast, and got the same 
error.

I have also googled for "[blastall] ERROR: BlastKarlinBlkGappedCalc: 
BLOSUM62 is a supported matrix", didn't get much useful information on my 
case.

Also, very strangely, when I tried to use the mvapich (v.0.9.9) to run:

bash> /opt/openmpi.gcc/bin/mpirun -np 5 -machinefile ./machines
       /home/local/bin/mpiblast --debug --config-file=/home/.mpiblast.conf
       -p blastp -i ./bait.fasta -d ecoli.aa

I got simply one line error:

  "End of file reached on hostfile at 4 of 9 hostnames"

Any hint?

Thanks in advance!

Zhiliang