Skip to content

Use Cases

This guide focuses on the most common everyday workflows in CraneSched and walks through each scenario from job submission to result retrieval. For specialized or large-scale workloads, follow the advanced guides linked in the Next Steps section.

Use Case 1: Basic Batch Job

Scenario: Run a simple computational task on a single node.

Task Description

You have a script that processes data and outputs results to a file. This is the most common use case for batch processing.

Step 1: Prepare Your Script

Create a file named simple_job.sh:

#!/bin/bash
#CBATCH --job-name=simple_test
#CBATCH --partition=CPU
#CBATCH --nodes=1
#CBATCH --ntasks-per-node=1
#CBATCH --cpus-per-task=1
#CBATCH --mem=100M
#CBATCH --time=0:05:00
#CBATCH --output=job_%j.out

# Print job information
echo "Job started at: $(date)"
echo "Running on node: $(hostname)"
echo "Job ID: $CRANE_JOB_ID"

# Simulate some computational work
echo "Processing data..."
sleep 5
echo "Data processing complete!"

# Print job completion
echo "Job finished at: $(date)"

Step 2: Submit the Job

cbatch simple_job.sh

Expected output: 

Job <12345> is submitted to partition <CPU>

Step 3: Check Job Status

cqueue -j 12345

or check all your jobs:

cqueue --self

Step 4: View Results

After the job completes, check the output file:

cat job_12345.out

Use Case 2: Interactive Session

Scenario: You need to interactively test code or debug an application on a compute node.

Task Description

Instead of submitting a batch job, you want direct command-line access to a compute node with specific resources.

Step 1: Request an Interactive Session

crun -p CPU -N 1 -c 2 --mem 500M -t 1:00:00 /bin/bash

This allocates: - 1 node from the CPU partition - 2 CPU cores - 500 MB memory - 1 hour time limit

Step 2: Work Interactively

Once connected, you'll see a shell prompt on the compute node:

# Check where you are
hostname

# Check allocated resources
echo $CRANE_JOB_NODELIST

# Run your commands
python test_script.py
./my_program --input data.txt

# When done, exit
exit

Resources are released automatically when you exit.

Common Patterns

Check Job History

After jobs complete, use cacct to view job statistics:

# View your recent jobs
cacct --user $USER --starttime $(date -d "7 days ago" +%Y-%m-%d)

# View specific job details
cacct --job <job_id>

Cancel Jobs

# Cancel a specific job
ccancel <job_id>

# Cancel all your pending jobs
ccancel --state=PENDING --user=$USER

Resource Efficiency

Check how efficiently your job used allocated resources:

ceff <job_id>

Tips and Best Practices

  1. Start Small: Test your job with minimal resources first, then scale up
  2. Set Accurate Time Limits: Overestimating wastes resources, underestimating kills jobs
  3. Use Job Names: Makes it easier to identify jobs in the queue
  4. Monitor Resource Usage: Use ceff after job completion to optimize future submissions
  5. Error Handling: Always specify separate --error file to catch stderr messages
  6. Checkpointing: For long jobs, implement checkpointing to resume from failures

Next Steps