MonitorRunningJobs

Once you've launched a batch job on the high performance computer, you will want to know if it has started, if it is still running, and what happened once it ran. 

To see if a job has started or is still running:

1. Open a terminal on red.uits.iu.edu

2. Check if you have any jobs in the slurm queue

   1. squeue -u $USER
      
      

You may see one of a few things. In all cases the first row you see will look like this

If that is all you see, your job is gone. See the next section about determining the fate of a job for next steps.

If there are lines below this, there is one line per job you have active. For example, you might see this:

This shows that I have 6 jobs running and tells me about each one. The '3474427_5' is the job number, this is helpful for checking in on your jobs while they run. The 'gpu-debug' is type of compute node it is on, yours should always be gpu. Let me know if they aren't. The 'lblVids' is the name of your job. The 'ehnewman' is you :-) . The 'PD' or 'R' indicates the job status. 'PD' means it hasn't started yet whereas 'R' means it is running. The numbers '0:00' or '0:01' here show how long your job has been running. The last two columns tell you about the what compute nodes are working on your job.

Determining the fate of a job

Once your jobs are out of the queue, you'll want to know what they accomplished or if they ended in error. To do this, you'll look the *.out files the jobs left behind:

1. Open a terminal on red.uits.iu.edu

2. Navigate to your slurm job directory

   1. cd ~/slurm_DLC
      
      

3. Check what *.out files are there

   1. ls -lt *.out
      
      

4. Scroll to the top and look at which files were edited most recently

5. Look at the last few lines of the most recent file (For example, if this was a file called 'slurm-3474427_6.out')

   1. tail slurm-3474427_6.out
      
      

6. At this point you could see several possible things. Here are a few possibilities based on a key word you might see among the text:

   1. 'COMPLETED!' If you were labeling videos, it means it finished labeling all videos! Hurrah!

   2. 'MEMORY' This shows up if your job ran out of memory. We can fix this by asking for more memory in the *.sh script.

   3. 'OUT OF TIME' This means your job didn't finish before the clock time ran out. If you were training a network, this is expected.

   4. random text pointing to random lines of code of random functions - This means there is a problem in the python script. Tell me about it!