How to debug

Dask jobqueue has been developed and tested by several contributors, each having a given HPC system setup to work on: a job scheduler in a given version running on a given OS. Thus, in some specific cases, it might not work out of the box on your system. This section provides some hints to help you determine what may be going wrong.

Checking job script

Dask-jobqueue submits “job scripts” to your queueing system (see How this works). Inspecting these scripts often reveals errors in the configuration of your Cluster object or maybe directives unexpected by your job scheduler, in particular the header containing #PBS, #SBATCH or equivalent lines. This can be done easily once you’ve created a cluster object:

print(cluster.job_script())

If everything in job script appears correct, the next step is to try to submit a test job using the script. You can simply copy and paste printed content to a real job script file, and submit it using qsub, sbatch, bsub or what is appropriate for you job queuing system.

To correct any problem detected at this point, you could try to use job_extra or env_extra kwargs when initializing your cluster object.

In particular, pay attention to the python executable used to launch the workers, which by default is the one used to launch the scheduler (this makes sense only if python is on a shared location accessible both to the Dask scheduler and the Dask workers). You can use the python argument in SLURMCluster to specify the python executable you want to use to launch your workers.

The typical error you might see is a ModuleNotFoundError, even if you loaded the right module just before:

Loading tensorflow-gpu/py3/2.1.0
  Loading requirement: cuda/10.1.2 cudnn/10.1-v7.5.1.10 nccl/2.5.6-2-cuda
    gcc/4.8.5 openmpi/4.0.2-cuda
distributed.nanny - INFO -         Start Nanny at: 'tcp://10.148.3.252:39243'
distributed.dashboard.proxy - INFO - To route to workers diagnostics web server please install jupyter-server-proxy: python -m pip install jupyter-server-proxy
distributed.worker - INFO -       Start worker at:   tcp://10.148.3.252:42618
distributed.worker - INFO -          Listening to:   tcp://10.148.3.252:42618
distributed.worker - INFO -          dashboard at:         10.148.3.252:36903
distributed.worker - INFO - Waiting to connect to:    tcp://10.148.0.20:35926
distributed.worker - INFO - -------------------------------------------------
distributed.worker - INFO -               Threads:                          1
distributed.worker - INFO -                Memory:                   10.00 GB
distributed.worker - INFO -       Local Directory: <local-dir>
distributed.worker - INFO - -------------------------------------------------
distributed.worker - INFO -         Registered to:    tcp://10.148.0.20:35926
distributed.worker - INFO - -------------------------------------------------
distributed.core - INFO - Starting established connection
distributed.worker - WARNING -  Compute Failed
Function:  train_dense_model
args:      (None, False, 64)
kwargs:    {}
Exception: ModuleNotFoundError("No module named 'tensorflow'")

slurmstepd: error: *** JOB 1368437 ON <node> CANCELLED AT 2020-04-10T17:14:30 ***
distributed.worker - INFO - Connection to scheduler broken.  Reconnecting...
distributed.worker - INFO - Stopping worker at tcp://10.148.3.252:42618
distributed.nanny - INFO - Worker closed

This happens when you created the cluster using a different python than the one you want to use for your workers (here module load python/3.7.5), giving the following job script (pay attention to the last line which will show which python is used):

#!/usr/bin/env bash

#SBATCH -J <job_name>
#SBATCH -n 1
#SBATCH --cpus-per-task=10
#SBATCH --mem=10G
#SBATCH -t 1:00:00
#SBATCH --gres=gpu:1
#SBATCH --qos=qos_gpu-dev
#SBATCH --distribution=block:block
#SBATCH --hint=nomultithread
#SBATCH --output=%x_%j.out
module purge
module load tensorflow-gpu/py3/2.1.0
/path/to/anaconda-py3/2019.10/bin/python -m distributed.cli.dask_worker tcp://10.148.0.20:44851 --nthreads 1 --memory-limit 10.00GB --name name --nanny --death-timeout 60 --interface ib0

Activate debug mode

Dask-jobqueue uses the Python logging module. To understand better what is happening under the hood, you may want to activate logging display. This can be done by running this line of python code in your script or notebook:

import logging
logging.basicConfig(format='%(levelname)s:%(message)s', level=logging.DEBUG)

Interact with your job queuing system

Every worker is launched inside a batch job, as explained above. It can be very helpful to query your job queuing system. Some things you might want to check:

  • are there running jobs related to dask-jobqueue?

  • are there finished jobs, error jobs?

  • what is the stdout or stderr of dask-jobqueue jobs?

Other things you might look at

From here it gets a little more complicated. A couple of other already seen problems are the following:

  • The submit command used in dask-jobqueue (qsub or equivalent) doesn’t correspond to the one that you use. Check in the given JobQueueCluster implementation that job submission command and arguments look familiar to you, eventually try them.

  • The submit command output is not the same as the one expected by dask-jobqueue. We use submit command stdout to parse the job_id corresponding to the launched group of worker. If the parsing fails, then dask-jobqueue won’t work as expected and may throw exceptions. You can have a look at the parsing function JobQueueCluster._job_id_from_submit_output.