lsf-slurm-wrappers

These are a set of wrapper scripts to common Slurm commands that execute LSF commands in the background. The scripts are intended as a migration aid for customers migrating from Slurm to LSF and not as a replacement for the LSF commands.

Abstract

Slurm Command Wrapper
Product/Component Release: 10.1
First Public Release: 01 April 2020

Abbreviations

Slurm: Simple Linux Utility for Resource Management

About Slurm command wrapper

The Slurm Command Wrapper provides a way for Slurm users, to submit and manage jobs in LSF cluster environment, with Slurm command syntax and options.

Commands included:

srun
sbatch
squeue
scontrol
sinfo
scancel

Not all Slurm command options are supported. If you don't find an option that you like, see the second Contributing Changes below.

Supported operating systems

RHEL 7 x86_64 and later
Generally any operating system supporting LSF and Python 2.7+

Contributing Changes

Changes from the broader Linux community are welcomed. If you would like to make a contribution to these scripts, simply open a pull request and in the body of the request, follow the DCO process documented in this and other IBMSpectrumComputing repositories. You essentially must "SignOff" using your real email address.

Wrapper command usage notes

srun - Run a parallel job.

Supported options:

Usage: srun [-N nnodes] [-n ntasks] [-i in] [-o out] [-e err] [-p partition]
            [--hold] [-t minutes] [-J jobname] [--open-mode=a|A|t|T] [-m dist]
            [--cpu_bind=...] [--mem_bind=...] [--ntasks-per-socket=n]
            [--ntasks-per-core=n] [--sockets-per-node=S] [--cores-per-socket=C]
            [--threads-per-core=T] [--ntasks-per-node=n] [--mail-user=user]
            executable [args...]

sbatch - Submit a batch script.

Supported options:

Usage: sbatch [-N nnodes] [-n ntasks] [-p partition] [--hold]
              [-t minutes] [-D path] [--workdir=directory]
              [-J jobname] [--mail-user=user] [--ntasks-per-node=n]
              [--export[=names]] executable [args...]

squeue - Submit a batch script.

Supported options:

Usage: srun [-N nnodes] [-n ntasks] [-i in] [-o out] [-e err] [-p partition]
            [--hold] [-t minutes] [-J jobname] [--open-mode=a|A|t|T] [-m dist]
            [--cpu_bind=...] [--mem_bind=...] [--ntasks-per-socket=n]
            [--ntasks-per-core=n] [--sockets-per-node=S] [--cores-per-socket=C]
            [--threads-per-core=T] [--ntasks-per-node=n] [--mail-user=user]
            executable [args...]

scontrol - Run a parallel job.

Supported options:

scontrol  [<OPTION>] [<COMMAND&>]
    Valid <OPTION> values are:
     -h or --help: equivalent to "help" command

    Valid <COMMAND> values are:
     resume <jobid_list>        resume previously suspended job (see suspend)
     suspend <jobid_list>       susend specified job (see resume)
     requeue <jobid_list>       re-queue a batch job
     release <jobid_list>       permit specified job to start (see hold)
     hold <jobid_list>          prevent specified job from starting.
                                <jobid_list> argument is
                                a comma separated list of job IDs.

sinfo - Report resource state.

Supported options:

Usage: sinfo [-alN] [-p partition]

scancel - Cancel the pending or running job.

Supported options:

Usage: scancel [-n job_name] [-p partition] [--usage]
               [-u user_name] [-w hosts...] [job_id]

Examples of command wrapper on LSF cluster

Job submission: sbatch, srun

sbatch example:

[hosta] # sbatch -n4 --ntasks-per-node=2 ./job.sh
Submitted batch job 695
[hosta] #

The job 695 requires 4 tasks in total and 2 tasks per node, to LSF cluster.

srun example:
[hosta] # srun -n4 --ntasks-per-node=2 ./job.sh
Thu Sep 1 01:12:51 EDT 2016: 0
Thu Sep 1 01:12:51 EDT 2016: 0
hosta
hosta
Thu Sep 1 01:12:53 EDT 2016: 0
Thu Sep 1 01:12:53 EDT 2016: 0
hostb
hostb
Thu Sep 1 01:12:53 EDT 2016: 1
Thu Sep 1 01:12:53 EDT 2016: 1
hosta
hosta
Thu Sep 1 01:12:55 EDT 2016: 1
Thu Sep 1 01:12:55 EDT 2016: 1
hostb
hostb
...

job.sh:
#!/bin/bash

((i=0))
while ((i<10));
do
        sleep 2
        echo `date`: $i
        hostname
        ((i=i+1))
done

Job management: squeue, scancel, scontrol

squeue example:

[hosta] # squeue
             JOBID PARTITION     NAME     USER ST       TIME  NODES NODELIST(REASON)
               696    normal ./job.sh  user1  R       5(s)      2 hosta,hostb 
[hosta] #

Compare this output to the LSF bjobs command output:

[hosta] # bjobs
JOBID   USER    STAT  QUEUE      FROM_HOST   EXEC_HOST   JOB_NAME   SUBMIT_TIME
696     user1 RUN   normal     hosta         hosta       ./job.sh   Sep  1 01:03
                                             hosta
                                             hostb
                                             hostb
[hosta] #

By default, job output is named slurm-696.out, and saved in the current directory.

scancel example:

[hosta] # squeue
             JOBID PARTITION     NAME     USER ST       TIME  NODES NODELIST(REASON)
               697 interactive    sleep  user1  R      11(s)      1 hostb
[hosta] # scancel 697
[hosta] # squeue
             JOBID PARTITION     NAME     USER ST       TIME  NODES NODELIST(REASON)
[hosta] #

scontrol example:

Suspend a running job, and then resume it:

[hosta] # bjobs
JOBID   USER    STAT  QUEUE      FROM_HOST   EXEC_HOST   JOB_NAME   SUBMIT_TIME
1421    llbld   RUN   normal     hosta       hosta       sleep 999  Oct 19 22:20
[hosta] # scontrol suspend 1421
[hosta] # bjobs
JOBID   USER    STAT  QUEUE      FROM_HOST   EXEC_HOST   JOB_NAME   SUBMIT_TIME
1421    llbld   USUSP normal     hosta       hosta       sleep 999  Oct 19 22:20
[hosta] # scontrol resume 1421
[hosta] # bjobs
JOBID   USER    STAT  QUEUE      FROM_HOST   EXEC_HOST   JOB_NAME   SUBMIT_TIME
1421    llbld   RUN   normal     hosta       hosta       sleep 999  Oct 19 22:20

Resource status: sinfo

Example:

[hosta] # sinfo
PARTITION    AVAIL  TIMELIMIT    NODES      STATE NODELIST
priority        up   infinite        2       idle all
admin           up   infinite        2       idle all
interactive*    up   infinite        2       idle all
normal*         up   infinite        2       idle all
night           up        5.0        2       idle all
test1           up       20.0        2       idle hosta,hostb
idle            up   infinite        2       idle all
test2           up       25.0        1       idle hosta
short           up   infinite        2       idle all
owners          up   infinite        2       idle all
test3           up       30.0        2       idle hosta,hostb
[hosta] #

Compare this output to the LSF bqueues command output:

[hosta] # bqueues
QUEUE_NAME      PRIO STATUS          MAX JL/U JL/P JL/H NJOBS  PEND   RUN  SUSP
admin            50  Open:Active       -    -    -    -     0     0     0     0
owners           43  Open:Active       -    -    -    -     0     0     0     0
priority         43  Open:Active       -    -    -    -     0     0     0     0
night            40  Open:Active       -    -    -    -     0     0     0     0
short            35  Open:Active       -    -    -    -     0     0     0     0
test1            35  Open:Active       -    -    -    -     0     0     0     0
test2            35  Open:Active       -    -    -    -     0     0     0     0
test3            35  Open:Active       -    -    -    -     0     0     0     0
normal           30  Open:Active       -    -    -    -     0     0     0     0
interactive      30  Open:Active       -    -    -    -     0     0     0     0
idle             20  Open:Active       -    -    -    -     0     0     0     0
[hosta] #

Installation and Configuration

Before installation (LSF_TOP=Full path to the top-level installation directory of LSF.)
1. Log on to the LSF master host as root
2. Set your environment:
- For csh or tcsh: % source LSF_TOP/conf/cshrc.lsf
- For sh, ksh, or bash: $ . LSF_TOP/conf/profile.lsf
Installation steps
1. Download the command wrappers and make them executable, for example:
```
chmod 755 srun
```
1. Copy the command wrappers to the binary directory $LSF_BINDIR/
Ensure LSF is installed and working properly.
You are now able to use the srun, sbatch, squeue, scontrol, sinfo, and scancel commands.

Copyright

License

Apache License

Version 2.0, January 2004

http://www.apache.org/licenses/

TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION. For details check the LICENSE file in the root of this repository.

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
cmd		cmd
IBMDCO.md		IBMDCO.md
LICENSE		LICENSE
README.md		README.md

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

lsf-slurm-wrappers

Abstract

Contents

Abbreviations

About Slurm command wrapper

Supported operating systems

Contributing Changes

Wrapper command usage notes

srun - Run a parallel job.

sbatch - Submit a batch script.

squeue - Submit a batch script.

scontrol - Run a parallel job.

sinfo - Report resource state.

scancel - Cancel the pending or running job.

Examples of command wrapper on LSF cluster

Installation and Configuration

Copyright

License

About

Releases

Packages

Languages

License

IBMSpectrumComputing/lsf-slurm-wrappers

Folders and files

Latest commit

History

Repository files navigation

lsf-slurm-wrappers

Abstract

Contents

Abbreviations

About Slurm command wrapper

Supported operating systems

Contributing Changes

Wrapper command usage notes

srun - Run a parallel job.

sbatch - Submit a batch script.

squeue - Submit a batch script.

scontrol - Run a parallel job.

sinfo - Report resource state.

scancel - Cancel the pending or running job.

Examples of command wrapper on LSF cluster

Installation and Configuration

Copyright

License

About

Resources

License

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages