search

ubccr-slurm-simulator / slurm_sim_tools

tools for slurm simulator
18 stars 14 forks source link

Is Preemption Avaliable? #4

Open jsearcy1 opened 5 years ago

jsearcy1 commented 5 years ago

Hi All, This is a great project. I'm trying to simulate our cluster here, and mostly everything is working well. However, I'm getting a number of errors whenever trying to use

PreemptType=preempt/partition_prio

in slurmctld.out I see Error connecting slurm stream socket at 127.0.0.1:6817: Connection refused

followed by all the jobs sitting in the queue with reason (FrontEndDown)

Could I be doing something wrong or is the simulator not ready for preemption yet?

Thanks, Jake

nsimakov commented 5 years ago

Hi Jake,

I didn't test preemption but it is very interesting feature.

With slurm simulator it is not obvious which feature would work right away and which one would need some attention. In this particular case, because there is no real slurmd and preemption require killing the job on compute node so there is a communication between slurm controller and slurm daemons, which needed to be faken for simulation.

I will take a look into what.

Thanks, Nikolay