search

bartongroup / slivka

http://bartongroup.github.io/slivka/
Apache License 2.0
7 stars 3 forks source link

Scheduler cannot recover running jobs #52

Closed warownia1 closed 5 years ago

warownia1 commented 5 years ago

Scheduler can't recover running jobs because run_configuration is None. run_configuration field of slivka.db.models.Request is not set when the job is created. When the requested job is picked-up for execution and the proper configuration is selected and the runner is created, the scheduler should save configuration name-id to the database.

Traceback (most recent call last):
...
  File "/cluster/rs_lab/mmwarowny/miniconda3/envs/slivka/lib/python3.5/site-packages/Slivka-0.3a4-py3.5.egg/slivka/command.py", line 167, in scheduler
    scheduler = Scheduler()
  File "/cluster/rs_lab/mmwarowny/miniconda3/envs/slivka/lib/python3.5/site-packages/Slivka-0.3a4-py3.5.egg/slivka/scheduler/scheduler.py", line 46, in __init__
    self._restore_jobs()
  File "/cluster/rs_lab/mmwarowny/miniconda3/envs/slivka/lib/python3.5/site-packages/Slivka-0.3a4-py3.5.egg/slivka/scheduler/scheduler.py", line 88, in _restore_jobs
    .get_runner_class(request.run_configuration)
  File "/cluster/rs_lab/mmwarowny/miniconda3/envs/slivka/lib/python3.5/site-packages/Slivka-0.3a4-py3.5.egg/slivka/scheduler/execution_manager.py", line 92, in get_runner_class
    return self.configurations[name].runner
KeyError: None
warownia1 commented 5 years ago

Inserting run configuration values to the database by hand allows restoring running jobs.