MPI problem

Top Page > Browsing

MPI problem

Date: 2025/01/17 19:46
Name: Md Aktar Hossain <mdhossainbhu@gmail.com>: Dear OpenMX Developers,

I recently started using OpenMX for my calculations. I am working on heterostructures, which require large computational resources. To address this, I have been submitting jobs to our institute's supercomputer. However, I am encountering issues whenever I request more than 40 CPUs. For jobs with fewer than 40 CPUs, everything works fine.

Below is the submission script I am using for 80 CPUs (2 nodes × 40 CPUs per node):

#!/bin/bash
#SBATCH -J mpi_80_m # Name of the job
#SBATCH -o test.log # Standard output log
#SBATCH -e test_err.log # Error log
#SBATCH -q iitkgp_freeq # Queue name
#SBATCH -p medium # Partition name (options: standard, standard-low, gpu, gpu-low, hm)
#SBATCH -N 2 # Number of nodes
#SBATCH --ntasks-per-node=40 # Number of tasks per node
#SBATCH -t 72:00:00 # Walltime (Max: 72 hours)
#SBATCH --mem=192000 # Max memory in MB

source /home/apps/spack/share/spack/setup-env.sh
spack load openmx@3.9 /2dvn4px

exe=openmx
mpirun slurm -n $SLURM_NTASKS $exe cu_111_22l_3x3_soc.dat > cu_111_22l_3x3_soc.std

When I attempt to submit this job, I encounter the following error:

There are not enough slots available in the system to satisfy the 80
slots that were requested by the application:

This issue prevents me from utilizing more than 40 CPUs. I would greatly appreciate it if someone could suggest a solution to overcome this limitation or help me identify what I might be doing wrong.

Thank you in advance for your help!

Best regards,
Aktar

Page: [1]

Re: MPI problem ( No.1 )

Date: 2025/01/18 17:12
Name: Naoya Yamaguchi

Hi,

This problem is caused by the fact that your job submission is not appropriate.
Such an error will occur if you submit a job with a number of MPI processes that is greater than the number of physical cores in the CPU.

Workarounds can be found by searching for that error message, for example:
https://stackoverflow.com/questions/35704637/mpirun-not-enough-slots-available

I recommend that you contact the supercomputer administrator to learn how to properly submit jobs.

Regards,
Naoya Yamaguchi

Re: MPI problem ( No.2 )

Date: 2025/01/18 18:08
Name: Md Aktar Hossain <mdhossainbhu@gmail.com>

Hi,

Thank you for your prompt response.

I would like to clarify that the same script works properly for other DFT packages like Quantum Espresso, where 80 processors are allocated (40 per node, as specified). However, the issue arises specifically with OpenMX.

I have also verified the specifications of our institute's supercomputer, which confirm 40 physical processors per node. This matches the configuration used in the script.

Below are the details of the supercomputer specifications for reference
"CPU only Compute Nodes : 384

2* Intel Xeon SKL G-6148

Cores = 40, 2.4GHz

Total Cores = 15,360 cores

Memory= 192 GB, DDR4 2666 MHz

Total Memory=73,728 GB

SSD = 480 GB (local scratch) per node "

Additionally, I have already contacted the supercomputer administrator to help resolve this issue, as OpenMX was installed by them. Unfortunately, I have not yet received a clear resolution.

I would greatly appreciate any further suggestions or guidance to resolve this issue.

Best regards,
Aktar

Re: MPI problem ( No.3 )

Date: 2025/01/19 00:19
Name: Naoya Yamaguchi

Dear Aktar,

>I would like to clarify that the same script works properly for other DFT packages like Quantum Espresso, where 80 processors are allocated (40 per node, as specified). However, the issue arises specifically with OpenMX.

I suggest you install the OpenMX properly from the source; you seem to be using the one from Spack, but the fact that the MPI calculation does not work means that it is not compatible with your computing environment.

Regards,
Naoya Yamaguchi

Re: MPI problem ( No.4 )

Date: 2025/01/19 15:48
Name: Md Aktar Hossain <mdhossainbhu@gmail.com>

Dear Yamaguchi,

Thank you for your suggestion to install OpenMX from source. I will contact the administrator of our institute's supercomputer to proceed with this installation. If the problem persists, I will reach out to you again for further assistance.

Best regards,
Aktar

Re: MPI problem ( No.5 )

Date: 2025/01/21 23:04
Name: Md Aktar Hossain <mdhossainbhu@gmail.com>

Dear Yamaguchi,
I am reaching out as I have not yet received a response from the administrator of our institute's supercomputer. I have gathered some details about our system, and I would like to seek your guidance on installing OpenMX.

Here are the system details:

Operating System:

NAME: CentOS Linux
VERSION: 7 (Core)
ID: centos
ID_LIKE: rhel fedora
VERSION_ID: 7
PRETTY_NAME: CentOS Linux 7 (Core)
CPE_NAME: cpe:/o:centos:centos:7
HOME_URL: https://www.centos.org/
BUG_REPORT_URL: https://bugs.centos.org/
CENTOS_MANTISBT_PROJECT: CentOS-7
CENTOS_MANTISBT_PROJECT_VERSION: 7
REDHAT_SUPPORT_PRODUCT: centos
REDHAT_SUPPORT_PRODUCT_VERSION: 7

Hardware:
Processor: 2x Intel Xeon SKL G-6148 cluster platform

I would greatly appreciate any instructions or steps you could provide to help me with the installation of OpenMX on this system.

Thank you for your time and assistance. I look forward to your reply.

Best regards,
Aktar

Page: [1]