[maker-devel] Running MAKER with MPI support and SLURM scheduler

[Siller, Karsten H (khs3z)]

Ok, we’ll try it with mpiexec. Thanks for your help.

Karsten

________________________________
From: Carson Holt <carsonhh at gmail.com>
Sent: Tuesday, March 12, 2019 11:07 AM
To: Siller, Karsten H (khs3z)
Cc: maker-devel at yandell-lab.org
Subject: Re: [maker-devel] Running MAKER with MPI support and SLURM scheduler

Launch with OpenMPI’s mpiexec command. The slurm stun command you are using won’t launch it as expected.

—Carson

On Mar 12, 2019, at 6:39 AM, Siller, Karsten H (khs3z) <khs3z at virginia.edu<mailto:khs3z at virginia.edu>> wrote:

Hi,

We have configured and installed MAKER 2.31 with OpenMPI 2.1.5 on our HPC cluster.  During build/installation, the MPI libs and mpicc wrapper are found.  This is the output of ./Build status

==============================================================================
STATUS MAKER v2.31.10
==============================================================================
PERL Dependencies:VERIFIED
External Programs:VERIFIED
External C Libraries:VERIFIED
MPI SUPPORT:ENABLED
MWAS Web Interface:DISABLED
MAKER PACKAGE:CONFIGURATION OK

This is how we run MAKER (the number of nodes and tasks are set via #SBATCH directives):

srun maker > maker_small_PARALLEL

This launches the correct number of MAKER processes across nodes but we’re seeing messages in the output "WARNING: Multiple MAKER processes have been started in the same directory.”  The processes appear to run ok, but we want to make sure there’s nothing fundamentally wrong with how we run MAKER.

Browsing the mailing list, I came across similar posts and it was suggested that there’s an issue either with the MAKER MPI configuration or with the scheduler.  We’re positive that the scheduler is configured properly since many other MPI applications run without problems on the cluster with the same scheduler and MPI implementation.

Is it possible that launching MAKER processes via srun is not supported?  Or is the warning harmless and can be ignored?

Thanks,
Karsten

—

Karsten Siller
University of Virginia


_______________________________________________
maker-devel mailing list
maker-devel at box290.bluehost.com<mailto:maker-devel at box290.bluehost.com>
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20190313/cfe389de/attachment-0003.html>