Difference between revisions of "Annotated SLURM Script"

Revision as of 20:33, 7 November 2016

HiPerGator 2.0 documentation

This is a walk-through for a basic SLURM scheduler job script for a common case of a multi-threaded analysys. If the program you run is single-threaded (can use only one CPU core) then only use '--ntasks=1' line for the cpu request instead of all three listed lines. Annotations are marked with bullet points. You can click on the link below to download the raw job script file without the annotation. Values in brackets are placeholders. You need to replace them with your own values. E.g. Change '<job name>' to something like 'blast_proj22'. We will write additional documentation on more complex job layouts for MPI jobs and other situations when a simple number of processor cores is not sufficient.

Download raw source of the [{{#fileLink: run.sh}} run.sh] file.

Set the shell to use

#!/bin/bash

Common arguments

Name the job to make it easier to see in the job queue

#SBATCH --job-name=<JOBNAME>

Email: Your email address to use for all batch system communications

##SBATCH --mail-user=<EMAIL>

What emails to send: NONE - no emails; ALL - all emails; END,FAIL - only email if the job fails and email the summary at the end of the job

#SBATCH --mail-type=FAIL,END

Standard Output and Error log files

Use file patterns

%j - job id

%A-%a - Array job id (A) and task id (a)

You can also use --error for a separate stderr log

#SBATCH --output <my_job-%j.out>

Number of nodes to use. For all non-MPI jobs this number will be equal to '1'

#SBATCH --nodes=1

Number of tasks. For all non-MPI jobs this number will be equal to '1'

#SBATCH --ntasks=1

Number of CPU cores to use. This number must match the argument used for the program you run.

#SBATCH --cpus-per-task=4

Total memory limit for the job. Default is 2 gigabytes, but units can be specified with mb or gb for Megabytes or Gigabytes.

#SBATCH --mem=4gb

Job run time in [DAYS]: HOURS:MINUTES:SECONDS; [DAYS] are optional, use when it is convenient

#SBATCH --time=72:00:00

Optional: A group to use if you belong to multiple groups. Otherwise, do not use.

#SBATCH --account=<GROUP>

A job array, which will create many jobs (called array tasks) different only in the '$SLURM_ARRAY_TASK_ID' variable, similar to Torque_Job_Arrays on HiPerGator 1

#SBATCH --array=<BEGIN-END>

Example of five tasks: #SBATCH --array=1-5

END OF PBS SETTINGS

Recommended convenient shell code to put into your job script

Add host, time, and directory name for later troubleshooting

date;hostname;pwd

Below is the shell script part - the commands you will run to analyze your data. The following is an example.

Load the software you need

module load ncbi_blast

Run the program

blastn -db nt -query input.fa -outfmt 6 -out results.xml --num_threads 4

date

@@ Line 1: / Line 1: @@
 [[Category:SLURM]]
 {{HPG2}}
-This is a walk-through for a basic SLURM scheduler job script. Annotations are marked with bullet points. You can click on the link below to download the raw job script file without the annotation. Values in brackets are placeholders. You need to replace them with your own values. E.g. Change '<job name>' to something like 'blast_proj22'. We will write additional documentation on more complex job layouts for MPI jobs and other situations when a simple number of processor cores is not sufficient.
+This is a walk-through for a basic SLURM scheduler job script for a common case of a multi-threaded analysys. If the program you run is single-threaded (can use only one CPU core) then only use '--ntasks=1' line for the cpu request instead of all three listed lines. Annotations are marked with bullet points. You can click on the link below to download the raw job script file without the annotation. Values in brackets are placeholders. You need to replace them with your own values. E.g. Change '<job name>' to something like 'blast_proj22'. We will write additional documentation on more complex job layouts for MPI jobs and other situations when a simple number of processor cores is not sufficient.
 Download raw source of the [{{#fileLink: run.sh}} run.sh] file.
@@ Line 33: / Line 33: @@
 :: %j - job id
 :: %A-%a - Array job id (A) and task id (a)
+:: You can also use --error for a separate stderr log
 {{#fileAnchor: run.sh}}
 <source lang=make>
 #SBATCH --output <my_job-%j.out>
-#SBATCH --error <my_job-%j.err>
 </source>
-;Number of nodes to use
+;Number of nodes to use. For all non-MPI jobs this number will be equal to '1'
 {{#fileAnchor: run.sh}}
 <source lang=make>
 #SBATCH --nodes=1
 </source>
-;Number of tasks (usually translate to processor cores) to use
+;Number of tasks. For all non-MPI jobs this number will be equal to '1'
 {{#fileAnchor: run.sh}}
 <source lang=make>
 #SBATCH --ntasks=1
+</source>
+;Number of CPU cores to use. This number must match the argument used for the program  you run.
+{{#fileAnchor: run.sh}}
+<source lang=make>
+#SBATCH --cpus-per-task=4
 </source>
 ;Total memory limit for the job. Default is 2 gigabytes, but units can be specified with mb or gb for Megabytes or Gigabytes.
@@ Line 92: / Line 97: @@
 {{#fileAnchor: run.sh}}
 <source lang=make>
-blastn -db nt -query input.fa -outfmt 6 -out results.xml
+blastn -db nt -query input.fa -outfmt 6 -out results.xml --num_threads 4
 date
 </source>

Difference between revisions of "Annotated SLURM Script"

Revision as of 20:33, 7 November 2016

Navigation menu

Search