For dependent jobs, it can sometimes take up to a minute for deadline to change the status from pending to active after dependencies have completed. Is there anything I can do to reduce this lag?
One of the reasons this is a problem is that if another job is submitted while the parent job is still running, if another job is submitted, it typically kicks off and may consume all the workers before the previous pending jobs are activated, leaving me in this blocking state:
- job2: active (running)
- job1.0: complete
- job1.1: pending
- job1.2: pending
after a minute:
- job2: active (running)
- job1.0: complete
- job1.1: active (queued)
- job1.2: active (queued)
Now job1.1 and job1.2 must wait for job2 to complete before they can begin.
You could reduce the ‘Pending Job Scan Interval’ in your repository config, if that is what you’re looking for?
1 Like
Thank you! This doesn’t solve my blocking problem, but I was able to reduce the lag significantly. I guess, the best plan is to dedicate a separate pool to these secondary jobs.
Yep, or maybe something could be done with weighting the jobs perhaps!
Ah, I didn’t know about weighting. Implementing FIFO would definitely do it, thanks!