AWS Thinkbox Discussion Forums

New to Deadline, Questions

We’re rolling out deadline and I have a few new guy questions. If some of these have been answered before, can you point me to the thread and I’ll go through it?

We pretty much only render with Vray DR since we do still images. I’ve been noticing in our testing with deadline that it seems more stable if we submit the DBR jobs with “Use Local Host” unchecked. Is that correct? We’re using DBR with dynamic start.

We’re seeing a lot of our servers either stall or hang onto jobs far past the job completion. Is there anyway around this? Just as a note, our deadline Manager is not in our office. We have to connect to a machine in another office that serves as the manager for all of the satellite offices. I would have to guess this is causing an awful lot of slowdown and traffic as commands go some distance. Does anyone run Deadline like this?

For example, one job completes in 12 minutes. If we send the exact same job to start right after the first, that follow up job takes 17 minutes. Most of the DR slaves that need to pick up are not, or get stalled out. We also had a job “render” for 18 hours even though it completed and saved the final image out. One of the slave nodes just got stuck onto the job.

Thanks for all of your help!

Well, the way DR works it’s only slightly surprising that the job runs forever. Under the hood they are just opening the spawner and watching for some text mentioning that the render completed. If they don’t see a message they’ll run forever. In fact, you can configure just that in the V-Ray DR settings:

If you do have the “on job complete” set to “Do Nothing” then that’s pretty normal.

The variability is also not that unexpected either as V-Ray usually streams the required assets over the connection. If there is something else being read (via Deadline or anything else) over the network connection that would affect things.

Interesting thought though, assuming the assets are available on the other end you could try tile rendering in Max, Maya or Cinema4D which would split the image up into smaller pieces without DR.

We’re not streaming the assets, we leave transfer missing assets off if that is what you mean. The biggest issue is that when we did vray DR with crappy ole backburner, it worked flawlessly. If a render took 30 minutes, you can send it again to start right after and it will take the same 30 minutes. DR boxes would respond on demand. However, we were just leaving the DR spawners running at all times since that was the way to do it with Backburner.

Could it be related to our repository is not in our office? We connect to another office’s deadline manager, then render back to machines that are local in this office. Our IT is pushing all offices to use just one repository/manager, is that normal?

Well, if we focus on the performance here I don’t think Deadline should make much difference.

I think there’s a lot of possible ground to cover here (peformance, job logs, etc) and it may be faster to just have a call on this and I can debrief in this thread. Would you have some time next week? I’m in central Canada so usually that works out for a chunk of Europe and most of North America.

I’m usually free Wednesdays from 9 to 5 so you can just try giving that number in my signature a call and we can set up a Chime remote session.

Privacy | Site terms | Cookie preferences