Im having an issue when sending out a set of renders via Unreal Engine plugin where I can only have one render machine picking up a project at one time.
When the second or third machine tries to load the project Unreal opens but then crashes… If I stop the one machine that is rendering, another instantly picks up and begins rendering.
Why might this be? Here is the error log output.
2023-11-07 17:46:13: 0: STDOUT: PermissionError: [Errno 13] Permission denied: ‘X:/Danny/Test_Deadline 5.3/Saved/MovieRenderPipeline[\QueueManifest.utxt’](file://queuemanifest.utxt’/)
All in latest versions, 5.3.2 and latest windows, latest Deadline etccc
Looks like an issue with the permission on the render node, but also the path has doesn’t translate to the actual path.
Are you using any relative path in the scene which needs to be mapped to an absolute path? If not, then the path in the error seems to be wrong and causing issue. I would like to isolate the submission from Deadline and try running the “Full Command” from the log file on the render node which failed the job.
Here is a documentation on isolating the submission: Troubleshooting — Deadline 10.3.0.15 documentation
Thanks for your message, I’m looking into the documentation you sent over but also wanted send you the full error report to see if that helps at all.
I can get all the machines to work if I submit a job to each one and no others, but if I send two jobs from the same project then I have this issue… one starts the job and one just keeps opening unreal and crashing until the first finishes. Once finished the second completes the job…
Please see paste bin for the error output i get…
And this is the second error report I have… Look at the bottom it talks about permission denied for the utxt file.
One of the folks on our side hit this as well and figured out a solution for this error - could you take a crack at his changes described here - UE5 Deadline Integration - #2 by Wookiecookie_27 - Rendering - Epic Developer Community Forums
and let us know how it goes? It’s pretty bleeding edge at the moment, but it tested well on his setup.
Interestingly this has made an improvements where I now have 4 of the 16 machines set up and all load a test scene together.
However, when I send a bigger scene I only have two of those 4 machines able to pick the job up and start rendering. Yet all four are able to manually open it?
So defiantly a better day but still have over 10 machines which just are unable to pick up the jobs sent… weird!
Do you think its a standard practice to install the latest version of Python on all the machines to make sure that is up to date?