When submitting jobs via submitit with timeout_min=0 to a Slurm partition that has a timelimit, the jobs indefinitely remain in a pending state. However, if timeout_min is not explicitly set, the jobs execute as expected based on the partition's timelimit.
Steps to Reproduce
Create a Slurm partition with a timelimit (e.g., 1 hour).
Set timeout_min=0 to executor
import submitit
executor = submitit.AutoExecutor(folder="submitit_jobs")
executor.update_parameters(
slurm_partition="your_partition", # Set timelimit to 1 hour
timeout_min=0 # Set to 0
)
def my_function():
print("Hello, world!")
job = executor.submit(my_function)
Expected Behavior
- The job should execute based on the Slurm partition's timelimit.
Actual Behavior
- The job remains in a pending state indefinitely.
---
When I don't set timeout_min, it works as it should. So, is this a bug, or is it supposed to work this way?
Summary
When submitting jobs via submitit with timeout_min=0 to a Slurm partition that has a timelimit, the jobs indefinitely remain in a pending state. However, if timeout_min is not explicitly set, the jobs execute as expected based on the partition's timelimit.
Steps to Reproduce
executor = submitit.AutoExecutor(folder="submitit_jobs") executor.update_parameters( slurm_partition="your_partition", # Set timelimit to 1 hour timeout_min=0 # Set to 0 )
def my_function(): print("Hello, world!")
job = executor.submit(my_function)