Serial Jobs¶

Video

Watch this in our courses: 2022 February, 2021 January

Abstract

Batch scripts let you run work non-interactively, which is important for scaling. You create a batch script, which runs in the background. You come back later and see the results.

Example batch script, submit with sbatch the_script.sh:

#!/bin/bash -l
#SBATCH --time=01:00:00
#SBATCH --cpus-per-task=2
#SBATCH --mem=4G

module load anaconda
python my_script.py

See the quick reference for complete list of options.

Prerequisites¶

Why batch scripts?¶

You learned, in Slurm: the queuing system, how all Triton users must do their computation by submitting jobs to the Slurm batch system to ensure efficient resource sharing. This lets you run many things at once without having to watch each one separately - the true power of the cluster.

A batch script is simply a shell script (remember Using the cluster from a shell?), where you put your resource requests and job steps.

Your first job script¶

A job script is simply a shell script (Bash). And so the first line in the script should be the shebang directive (#!) followed by the full path to the executable binary of the shell’s interpreter, which is Bash in our case. What then follow are the resource requests, and then the job steps.

Let’s take a look at the following script

#!/bin/bash
#SBATCH --time=00:05:00
#SBATCH --mem-per-cpu=100M
#SBATCH --output=pi.out

echo "Hello $USER! You are on node $HOSTNAME.  The time is $(date)."

# For the next line to work, you need to be in the
# hpc-examples directory.
srun python slurm/pi.py 10000

Let’s name it run-pi.sh (create a file using your editor of choice, e.g. nano; write the script above and save it)

The symbol # is a comment in the bash script, and Slurm understands #SBATCH as parameters, determining the resource requests. Here, we have requested a time limit of 5 minutes, along with 100 MB of RAM per CPU.

Resource requests are followed by job steps, which are the actual tasks to be done. Each srun within the a slurm script is a job step, and appears as a separate row in your history - which is useful for monitoring.

Having written the script, you need to submit the job to Slum through the sbatch command. Since the command is python slurm/pi.py, you need to be in the hpc-examples directory from our sample project:

$ cd hpc-examples       # wherever you have hpc-examples
$ sbatch run-pi.sh
Submitted batch job 52428672

Warning

You must use sbatch, not bash to submit the job since it is Slurm that understands the SBATCH directives, not Bash.

When the job enters the queue successfully, the response that the job has been submitted is printed in your terminal, along with the jobid assigned to the job.

You can check the status of you jobs using slurm q/slurm queue (or squeue -u $USER):

$ slurm q
JOBID              PARTITION NAME                  TIME       START_TIME    STATE NODELIST(REASON)
52428672           debug     run-pi.sh             0:00              N/A  PENDING (None)

Once the job is completed successfully, the state changes to COMPLETED and the output is then saved to pi.out in the current directory. You can also wildcards like %u for your username and %j for the jobid in the output file name. See the documentation of sbatch for a full list of available wildcards.

Setting resource parameters¶

The resources were discussed in Slurm: the queuing system, and barely need to be mentioned again here - the point is they are the same. For example, you might use --mem=5G or --time=5:00:00. Always keep the reference page close for looking these up.

Checking your jobs¶

Once you submit your jobs, it goes into a queue. The two most useful commands to see the status of your jobs with are slurm q/slurm queue and slurm h/slurm history (or squeue -u $USER and sacct -u $USER).

More information is in the monitoring tutorial.

Cancelling a job¶

You can cancel jobs with scancel JOBID. To obtain job id, use the monitoring commands.

Full reference¶

The reference page contains it all, or expand it below.

Slurm quick ref

Command	Description
`sbatch`	submit a job to queue (see standard options below)
`srun`	Within a running job script/environment: Run code using the allocated resources (see options below)
`srun`	On frontend: submit to queue, wait until done, show output. (see options below)
`sinteractive`	Submit job, wait, provide shell on node for interactive playing (X forwarding works, default partition interactive). Exit shell when done. (see options below)
`srun --pty bash`	(advanced) Another way to run interactive jobs, no X forwarding but simpler. Exit shell when done.
`scancel` JOBID	Cancel a job in queue
`salloc`	(advanced) Allocate resources from frontend node. Use `srun` to run using those resources, `exit` to close shell when done (see options below)
`scontrol`	View/modify job and slurm configuration

Command	Option	Description
`sbatch`/`srun`/etc	`-t`, `--time=`HH:MM:SS	time limit
	`-t, --time=`DD-HH	time limit, days-hours
	`-p, --partition=`PARTITION	job partition. Usually leave off and things are auto-detected.
	`--mem-per-cpu=`N	request n MB of memory per core
	`--mem=`N	request n MB memory per node
	`-c`, `--cpus-per-task=`N	*Allocate n* CPU’s for each task. For multithreaded jobs. (compare ``–ntasks``: ``-c`` means the number of cores for each process started.)**
	`-N`, `--nodes=`N-M	allocate minimum of n, maximum of m nodes.
	`-n`, `--ntasks=`N	allocate resources for and start n tasks (one task=one process started, it is up to you to make them communicate. However the main script runs only on first node, the sub-processes run with “srun” are run this many times.)
	`-J`, `--job-name=`NAME	short job name
	`-o` OUTPUTFILE	print output into file output
	`-e` ERRORFILE	print errors into file error
	`--exclusive`	allocate exclusive access to nodes. For large parallel jobs.
	`--constraint=`FEATURE	request feature (see `slurm features` for the current list of configured features, or Arch under the hardware list). Multiple with `--constraint="hsw\|skl"`.
	`--array=`0-5,7,10-15	Run job multiple times, use variable `$SLURM_ARRAY_TASK_ID` to adjust parameters.
	`--gres=gpu`	request a GPU, or `--gres=gpu:`n for multiple
	`--gres=spindle`	request nodes that have disks, `spindle:`n, for a certain number of RAID0 disks
	`--mail-type=`TYPE	notify of events: `BEGIN`, `END`, `FAIL`, `ALL`, `REQUEUE` (not on triton) or `ALL.` MUST BE used with `--mail-user=` only
	`--mail-user=`YOUR@EMAIL	whome to send the email
`srun`	`-N` N_NODES hostname	Print allocated nodes (from within script)

Exercises¶

The scripts you need for the following exercises can be found in this git repository: hpc-examples. You can clone the repository by running git clone https://github.com/AaltoSciComp/hpc-examples.git. This repository will be used for most of the tutorial exercises.

Serial-1: Basic batch job

Submit a batch job that just runs hostname and pi.py.

Set time to 1 hour and 15 minutes, memory to 500MB.
Change the job’s name and output file.
Check the output. Does the printed hostname match the one given by slurm history/sacct -u $USER?

Serial-2: Submitting and cancelling a job

Create a batch script which does nothing (or some pointless operation for a while), for example sleep 300 (this shell command does nothing for 300 seconds). Check the queue to see when it starts running. Then, cancel the job. What output is produced?

Serial-3: Checking output

Create a slurm script that runs the following program:

for i in $(seq 30); do
  date
  sleep 10
done

Submit the job to the queue.
Log out from Triton. Log back in and use slurm queue/squeue -u $USER to check the job status.
Use cat NAME_OF_OUTPUTFILE to check at the output periodically. You can use tail -f NAME_OF_OUTPUTFILE to view the progress in real time as new lines are added (Control-C to cancel)
Cancel the job once you’re finished.

Serial-4: Constrain to a certain CPU architecture

Modify the script from exercise #1 to run on only one type of CPU using the --constraint option. Hint: check Triton quick reference

Serial-5: Why you use sbatch, not bash.

(Advanced) What happens if you submit a batch script with bash instead of sbatch? Does it appear to run? Does it use all the Slurm options?

Solution

It looks like it runs, but actually is only running on the login node! If you used srun python3 slurm/pi.py 10000, then it would request a Slurm allocation, but not use any of the #SBATCH parameters, so might not request the resources you need.:

$ bash run-pi.sh
Calculating Pi via 10000 stochastic trials
{"successes": 7815, "pi_estimate": 3.126, "iterations": 10000}

(advanced) Serial-6: Interpreters other than bash

(Advanced) Create a batch script that runs in another language using a different #! line. Does it run? What are some of the advantages and problems here?

(advanced) Serial-7: Job environment variables.

Either make a sbatch script that runs the command env | sort, or use srun env | sort. The env utility prints all environment variables, and sort sorts it (and | connects the output of env to the input of sort.)

This will show all of the environment variables that are set in the job. Note the ones that start with SLURM_. Notice how they reflect the job parameters. You can use these in your jobs if needed (for example, a job that will adapt to the number of allocated CPUs).

What’s next?¶

There are various tools one can use to do job monitoring.