First test jobs

This page contains two batch job examples, that can be used to get started with batch processing:

First Slurm job job which could be run on any SLURM installation.
First Hummel-2 jobs which has the properties of all batch jobs to be run on Hummel-2.

First Slurm job

This job is generic, i.e. it contains nothing that is specific to Hummel-2. It contains resource specifications (--ntasks and --time) and one command (echo).

line no.	/sw/batch/examples/first-test-jobs/first-slurm-job.sh
`1` `2` `3` `4` `5` `6` `7` `8` `9`	`#!/bin/bash` `#SBATCH --ntasks=1` `#SBATCH --time=00:01:00` `# This batch job should run on every Slurm batch system.` `echo "Hello, world!"` `exit`

First Hummel-2 job

This job has all characteristic properties of a batch job to be run on Hummel-2:

First of all it is a parallel job. It has the smallest possible job size which is 8 CPU cores.
The recommended submit option –export=NONE is set.
The recommended initialization source /sw/batch/init.sh is included.

The job uses the srun command for demonstration purpose. (OpenMP programs do not need srun. MPI programs shall be started with mpirun.)

line
no. /sw/batch/examples/first-test-jobs/first-hummel2-job.sh

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48 #!/bin/bash
#
# Do not forget to select a proper partition and account if the
# default one is no fit for the job! You can do that with environment
# variables SBATCH_PARTITION and SBATCH_ACCOUNT, in the sbatch command
# line or here with the other #SBATCH settings.
#
# Never forget "--export=NONE" and "source /sw/batch/init.sh" !
# Strange happenings ensue otherwise.
#
#SBATCH --job-name=hello
#SBATCH --ntasks=8
#SBATCH --time=00:10:00
#SBATCH --export=NONE

source /sw/batch/init.sh

# Load environment modules for your application here.

# Useful settings:
set -e # Stop operation on first error.
set -u # Treat undefined variables as an error.
set -x # Print command lines as they are executed.

# Actual work starting here. You might need to call mpirun for MPI
# programs or srun depending on your type of application for proper
# parallel work. Examples of simple commands (that might themselves
# handle parallelisation):
echo "Hello World! I am node $(hostname -s) greeting you!"
echo "Also, my current TMPDIR is $TMPDIR"

# Let's pretend our started processes are working on a predetermined
# parameter set, and are finding their specific parameters using the
# set number and the process number inside the batch job:
export PARAMETER_SET=42

# Simplest way to run an identical command on all allocated
# cores. Environment variables can be used to tell apart the
# instances:
srun bash -c 'echo "process $SLURM_PROCID \
(out of $SLURM_NPROCS total) on $(hostname -s), \
parameter set: $PARAMETER_SET"'

# Demonstration of process binding: each "grep" uses
# only the CPU core printed:
srun --cpu-bind=cores grep Cpus_allowed_list: /proc/self/status

exit

line no.	/sw/batch/examples/first-test-jobs/first-hummel2-job.sh
`1` `2` `3` `4` `5` `6` `7` `8` `9` `10` `11` `12` `13` `14` `15` `16` `17` `18` `19` `20` `21` `22` `23` `24` `25` `26` `27` `28` `29` `30` `31` `32` `33` `34` `35` `36` `37` `38` `39` `40` `41` `42` `43` `44` `45` `46` `47` `48`	`#!/bin/bash` `#` `# Do not forget to select a proper partition and account if the` `# default one is no fit for the job! You can do that with environment` `# variables SBATCH_PARTITION and SBATCH_ACCOUNT, in the sbatch command` `# line or here with the other #SBATCH settings.` `#` `# Never forget "--export=NONE" and "source /sw/batch/init.sh" !` `# Strange happenings ensue otherwise.` `#` `#SBATCH --job-name=hello` `#SBATCH --ntasks=8` `#SBATCH --time=00:10:00` `#SBATCH --export=NONE` `source /sw/batch/init.sh` `# Load environment modules for your application here.` `# Useful settings:` `set -e # Stop operation on first error.` `set -u # Treat undefined variables as an error.` `set -x # Print command lines as they are executed.` `# Actual work starting here. You might need to call mpirun for MPI` `# programs or srun depending on your type of application for proper` `# parallel work. Examples of simple commands (that might themselves` `# handle parallelisation):` `echo "Hello World! I am node $(hostname -s) greeting you!"` `echo "Also, my current TMPDIR is $TMPDIR"` `# Let's pretend our started processes are working on a predetermined` `# parameter set, and are finding their specific parameters using the` `# set number and the process number inside the batch job:` `export PARAMETER_SET=42` `# Simplest way to run an identical command on all allocated` `# cores. Environment variables can be used to tell apart the` `# instances:` `srun bash -c 'echo "process $SLURM_PROCID \` `(out of $SLURM_NPROCS total) on $(hostname -s), \` `parameter set: $PARAMETER_SET"'` `# Demonstration of process binding: each "grep" uses` `# only the CPU core printed:` `srun --cpu-bind=cores grep Cpus_allowed_list: /proc/self/status` `exit`

The commands for trying this batch job are:

shell$ cd $BEEGFS
shell$ sbatch /sw/batch/examples/first-test-jobs/first-hummel2-job.sh 
Submitted batch job 12345
shell$ cat slurm-12345.out 
module: loaded site/slurm
module: loaded site/tmpdir
module: loaded site/hummel2
module: loaded env/system-gcc
++ hostname -s
+ echo 'Hello World! I am node n123 greeting you!'
Hello World! I am node n123 greeting you!
+ echo 'Also, my current TMPDIR is /tmp'
Also, my current TMPDIR is /tmp
+ export PARAMETER_SET=42
+ PARAMETER_SET=42
+ srun bash -c 'echo "process $SLURM_PROCID \
   (out of $SLURM_NPROCS total) on $(hostname -s), \
   parameter set: $PARAMETER_SET"'
process 5    (out of 8 total) on n123,    parameter set: 42
process 0    (out of 8 total) on n123,    parameter set: 42
process 4    (out of 8 total) on n123,    parameter set: 42
process 1    (out of 8 total) on n123,    parameter set: 42
process 2    (out of 8 total) on n123,    parameter set: 42
process 6    (out of 8 total) on n123,    parameter set: 42
process 3    (out of 8 total) on n123,    parameter set: 42
process 7    (out of 8 total) on n123,    parameter set: 42
+ srun --cpu-bind=cores grep Cpus_allowed_list: /proc/self/status
Cpus_allowed_list:      160
Cpus_allowed_list:      161
Cpus_allowed_list:      162
Cpus_allowed_list:      163
Cpus_allowed_list:      164
Cpus_allowed_list:      165
Cpus_allowed_list:      166
Cpus_allowed_list:      167
+ exit