Problem with slurm backend when calling scancel

HenrikBengtsson / future.batchtools

:rocket: R package future.batchtools: A Future API for Parallel and Distributed Processing using batchtools

https://future.batchtools.futureverse.org

84 stars 9 forks source link

Problem with slurm backend when calling scancel #45

Open statquant opened 5 years ago

statquant commented 5 years ago

Hello, When I call scancel to cancel all tasks submitted to a slurm cluster by future.batchtools then hit CTRL-C to get the terminal back, R displays an error per task and is painfully slow to come back.

HenrikBengtsson commented 5 years ago

How many jobs are we talking about? Basically, is it "painfully slow to come back" for a single job or for 100's of jobs?

What's probably going on is that futures are being garbage collected by R and we cleaned up they reach out to Slurm to query about each job's status. Doing lots of these queries might take some time. At least that's my guess based on the limited info I have at hand.

statquant commented 5 years ago

Hello, I was referring to 1000 jobs. Interrupt with Ctrl-C becomes useless I would say I get an error message per second. I have little information as I am forced to kill -9 the master R process. I can probably replicate this this easily, is there any information you'd like me to report?