[BioBrew Users] Trouble running mpiBlast

Glen Otero gotero at linuxprophet.com
Thu Mar 11 01:50:46 EST 2004


Hi Soumyadeep-

What application were you running with pvm, HMMER? If it was HMMER, 
which version?

And I think the problem with mpiBLAST (must be run on more than two 
nodes error) is that mpiBLAST is built with lam-mpi and hasn't been 
tested with vmi-mpich. Can you try using lam-mpi with mpiBLAST and let 
me know how it goes?

Glen


On Mar 10, 2004, at 3:26 AM, Soumyadeep nandi wrote:

> Hi Glen,
>
> We made no changes in the local .bashrc or profile
> files. The /etc/profile.d directory had two sets of
> pvm paths - one pvm_profile.csh and .sh and the other
> pvm.csh and .sh. The pvm_profile.sh had the following
> contents
>
> export PVM_ROOT=/usr/share/pvm3
> export PVM_RSH=/usr/bin/ssh
>
> while the pvm.sh had
>
> PVM_RSH=/usr/bin/rsh
> PVM_ROOT=/usr/share/pvm3
> export PVM_RSH PVM_ROOT
>
> as we require PVM_RSH to be ssh (rsh is not yet setup
> on the cluster) we removed the pvm.sh.
>
> We did have problems extracting the PVM environmental
> variables when setting up pvm on a cluster of
> workstations and got around it by sending these
> directives as part of the hostfile used to start pvm.
> This is our hostfile:
>
> * dx=/usr/share/pvm3/lib/pvmd
> ep=/usr/share/pvm3/bin/LINUX sp=2000
> frontend-0
> compute-0-0
> compute-0-1
>
>
>> Where did you install PVM on the compute nodes?
> We installed
> /home/install/7.3/en/os/i386/RedHat/RPMS/pvm-3.4.4-2.i386.rpm
> as "rpm -ivh pvm-3.4.4-2.i386.rpm"
> and it was installed by default in /usr/share/pvm3
>
> Regards,
> Soumyadeep
>
> --- Glen Otero <gotero at linuxprophet.com> wrote:
>> I'll try to duplicate your mpiBLAST scenario
>> tomorrow. In the meantime,
>> can you send me an example of a user's .bashrc and
>> .bash_profile from
>> your cluster? I'd like to see how you set the PVM
>> environment variables
>> for your users. Or did you put them in /etc/profile?
>> Where did you
>> install PVM on the compute nodes?
>>
>> Glen
>>
>> On Mar 9, 2004, at 11:16 PM, Soumyadeep nandi wrote:
>>
>>>
>>> Thanks Glen and Kip for your instructions.
>>>
>>> We could troubleshoot the earlier troubles.While
>>> working with pvm we found contirb-pvm is installed
>> in
>>> the frontend and no pvm was installed in the slave
>>> nodes. So we installed the normal pvm present in
>> the
>>> distro and it is working fine.
>>> Now we are trying to run blast on the cluster. We
>>> could ran the normal blast without any hiccups.
>> But
>>> the problem started with mpiblast.
>>> We have downloaded some of the example scripts
>> from
>>>
>>
> http://www.llnl.gov/computing/tutorials/mpi/samples/C/.
>>> Compiled the programs with "/usr/bin/mpicc",
>> created
>>> the machines file (with only two nodes,
>> compute-0-0,
>>> compute-0-1), started up ssh-agent and ran the
>> sample
>>> program. That ran fine in all the compute nodes
>>> without any error.
>>> As we try to run the mpiBlast with the following
>>> command:
>>> [andrew at frontend-0 mpi_test_blast]$
>>> /usr/local/mpich-vmi2/bin/mpirun -np 2
>> -machinefile
>>> ../machines /opt/BioBrew/NCBI/6.1.0/bin/mpiblast
>>>
>>
> --config-file=/opt/BioBrew/NCBI/6.1.0/etc/mpiblast.conf
>>> -p blastp -d ../db/nr -i test.seq
>>> We get the error as bellow:
>>> Sorry, mpiBLAST must be run on 2 or more nodes.
>>>
>>> With Regards,
>>> Soumyadeep
>>>
>>> __________________________________
>>> Do you Yahoo!?
>>> Yahoo! Search - Find what you’re looking for
>> faster
>>> http://search.yahoo.com
>>> _______________________________________________
>>> BioBrew-Users mailing list
>>> BioBrew-Users at bioinformatics.org
>>>
>>
> https://bioinformatics.org/mailman/listinfo/BioBrew-Users
>>>
>>>
>> Glen Otero, Ph.D.
>> Linux Prophet
>> 619.917.1772
>>
>
>
> __________________________________
> Do you Yahoo!?
> Yahoo! Search - Find what you’re looking for faster
> http://search.yahoo.com
>
>
Glen Otero, Ph.D.
Linux Prophet
619.917.1772




More information about the BioBrew-Users mailing list