AWS Thinkbox Discussion Forums

Resuming tasks vs frames after idle detection

My question is: is there a way to get jobs to resume ‘frames’ vs resuming ‘tasks’ after idle detection?

The scenario: We have 8K renders that take ~3 hours per ‘task’ (5 frames). so a worker chews on that for a couple hours and gets 3/5 frames done. Then it gets ‘awoken’ by idle detection etc. So that whole task gets queued again, and even though 3/5 frames are rendered, the next time any worker starts that ‘task’ again, it starts the whole thing over… re-rendering frames 1,2, 3… then it stops again etc… you can see the issue.

Aside from changing the frames per task on this job is there some other setting that will either check the drive for existing output or some how resume the task on the next unrendered frame?

thanks

Hello

I could not quiet get the Worker gets awoken by idle detection? Like the Worker shuts down if the machine is idle?

Please share a screenshot of the Idle Detection settings and more details of what is going on.

There is an option in the Worker scheduling> Miscellaneous Options:

You can check that off and that should do the trick.

is not an issue of idle detection persay, its just that when ‘idle detection’ kicks in, it stops the worker/task (as intended). But when that task is resumed, it starts at the beginning of the task, even if frames were already rendered.

I guess I’m looking for a more granular way of resuming tasks or some way to detect already rendered frames, before starting the task from the beginning.

If the application with which you are rendering support that then it might be possible. Some applications like Houdini writes a cache while rendering (bgeo cache) on the network. If the Worker is gone but the cache is still there, it could start from where it left off. Deadline does not have a functionality to resume from the completed frame.

Privacy | Site terms | Cookie preferences