Is there a way to render only 1 job at a time like in Keyshot 6 NR?

Started by willchocks7, August 20, 2017, 10:02:24 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

willchocks7

Hi everyone,

Keyshot 7 Network Rendering appears to render multiple images at once. Is it possible to set it up so that all slaves work on the one image like in Keyshot 6? Individual renders seem to be slower than ever and I can't see how long each job is taking because the total 'elapsed time' is just a sum of the pending+processing time.

Thanks for reading

DriesV

Hi,

You can change the priority by right clicking on a job. If you set a job's priority to urgent, then as soon as Workers are free, they will process that job.
If there aren't any free Workers, then you can manually abort the tasks of Workers in other jobs. After aborting, the Workers will process the urgent job first.

I hope that helps.

Dries

Morten Kristensen

Hi Will,

What do you mean with this?
Quote from: willchocks7 on August 20, 2017, 10:02:24 PM
Individual renders seem to be slower than ever

The elapsed times are actually correct, if a bit confusing, because the first task of a job is synchronizing the data onto a node in the setup. This means that the job is started at this point but since the sync tasks always, initially, have critical priority, the render tasks will be done after them (if you add more jobs at the same time, for instance). We'll give this some thought for 7.1 if it could be improved how this information is shown.

Gordon

I am tempted to make all my users go back to KS v6 just so I can get NR v6 back :)   I was just told they are working on bringing back v6 functionality where free workers would compete on a job in a queue to see which would finish it first before moving on to the next job.  v7 just starts on the next job as soon as any worker is free.   In our case we have  two 48 core workers and  four 8 core workers in our pool so invariably all my jobs get stuck waiting for the 8 core machines to process a region, even when both of the 48 core machines are sitting idle because they finished all their regions.

DMerz III

I know what you mean, but I think there's a better solution than going back to NR6

I feel like this could be resolved if the tasks were broken up into smaller components. (ie, smaller region dicing) if there's a region which is problematic and a 'slower' machine is tasked with churning it out, the ability to have other machines chime in and help out vs moving on to the next task could be super beneficial. Smaller regions = more tasks = less 'problematic' regions assigned to just one machine.