Debugging Open OnDemand

When you run into issues with jobs in Open OnDemand, review the output files from your session for important clues on what went wrong.

Review Session Output Files

  1. Go to My Interactive Sessions.

    My interactive sessions button in the upper OOD menu.

  2. Find the interactive session of interest from the list and click the Session ID link.

    Example interactive session with the Session ID field highlighted.

  3. Browse the list of output files for clues about what went wrong. The user_defined_context.json and output.log files are highlighted below.

user_defined_context.json

The user_defined_context.json file shows the options you input to launch the session.

Did you request the correct amount of RAM in the correct format? Use Slurm format (for example, 4096M or 10G). If this field is left blank, 1000 MB will be allocated per CPU core requested.

The following user_defined_context.json example is for a 30 minute, 1 CPU, JupyterLab job request with default memory.

bc_account   "bbka-delta-cpu"
bc_partition "cpu-interactive"
bc_duration  "00-00:30:00"
bc_reservation       ""
bc_num_slots "1"
bc_num_memory        ""
bc_num_gpus  "0"
bc_email_on_started  "0"
working_dir  ""

output.log

The output.log file shows what happened when the script.sh file was run. The last line of output.log should tell you why the job ended.

The following example is the last line of the output.log file of a job that ended because it reached its requested duration.

slurmstepd: error: *** JOB 4214572 ON cn001 CANCELLED AT 2024-07-24T14:06:53 DUE TO TIME LIMIT ***