[BioBrew Users] Unable to get mpiblast running

[Bastian Friedrich]

Hi Glen,

thank you for your quick response.

On Thursday 06 April 2006 15:53, Glen Otero wrote:
>
> I usually don't see these types of errors. Here are a few questions:
>
> How did you format the database for mpiblast?

/usr/local/bin/mpiformatdb --nfrags=28 -i Hs.seq.uniq
was the latest call, but I had used
mpiformatdb -N 28 -i Hs.seq.uniq
earlier.

> Is the mpiblast database on a shared filesystem, like NFS (I don't
> think symlinks will work)?

Currently, I have created a /export/data/blastdb/ on the frontend; this 
was rsynced to /state/partition1/blastdb on the compute nodes; on the 
frontend, I had a directory /state/partition1 (on the root 
partition...) containing a symlink to /export/data/blastdb.

I have just used a bind mount on the frontend (no more symlinking), but 
this was not successful, either.

The first tests were done via NFS, which did not work either.

> How did you launch the job, SGE?

In the future, we surely want to use mpiblast in an SGE environment; 
currently, it was started from the command line.

> Can you try a smaller job using just the 6 compute nodes (and
> formatting the db into 6 pieces)?

Wow, I get a new one now:
===============================
bastian at frontend:/state/partition1/blastdb> mpiformatdb -N 6 -i 
Hs.seq.uniq
[...]
[... semi-manual distributing data to /state/partiton1/blastdb of all 
nodes ...]
bastian at frontend:/state/partition1/blastdb> cd ~/tmp03
bastian at frontend:~/tmp03> /opt/mpich/gnu/sbin/cleanipcs
bastian at frontend:~/tmp03> cluster-fork /opt/mpich/gnu/sbin/cleanipcs
[...]
bastian at frontend:~/tmp03> mpirun -np 6 /usr/local/bin/mpiblast -p blastn 
-d Hs.seq.uniq -i IL2RA -o blast_results
54p3_2934:  p4_error: : 0
3       0.078125        Bailing out with signal 11
[3] MPI Abort by user Aborting program !
[3] Aborting program!
2p1_28697:  p4_error: interrupt SIGx: 13
 p5_17962:  p4_error: : 0
        0.0742188       Bailing out with signal 11
[5] MPI Abort by user Aborting program !
[5] Aborting program!
p4_21219:  p4_error: : 0
rm_l_4_21279: (0.367188) net_send: could not write to fd=5, errno = 104
        0.078125        Bailing out with signal 11
[4] MPI Abort by user Aborting program !
[4] Aborting program!
p2_13443:  p4_error: : 0
        0.078125        Bailing out with signal 11
[2] MPI Abort by user Aborting program !
[2] Aborting program!
rm_l_3_2994: (0.644531) net_send: could not write to fd=5, errno = 104

 p1_28697: (7.242188) net_send: could not write to fd=5, errno = 32
rm_l_2_13503: (6.929688) net_send: could not write to fd=5, errno = 104
p2_13443: (6.929688) net_send: could not write to fd=5, errno = 32
p5_17962: (6.093750) net_send: could not write to fd=5, errno = 32
===============================

Signal 11 seems to be a segfault? Something's going awfully wrong 
here...

> Can you try a smaller blast job using p53, p53db from ftp://
> ftp.bioinformatics.org/pub/biobrew/ and blastp?

This works! :)) The first time I see mpiblast actually working :)

Unfortunately, we are looking forward to blasting against the 17 GB 
genebank... Any more ideas?

Thx again,
   Bastian

-- 
 Bastian Friedrich                  bastian at bastian-friedrich.de
 Adress & Fon available on my HP   http://www.bastian-friedrich.de/
\~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\
\ Computers make very fast, very accurate mistakes.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://bioinformatics.org/pipermail/biobrew-users/attachments/20060406/6f4156f9/attachment.bin