We have this error since many years, each time we had new computeurs to the remote render farm.
The error happen when we requeue a job. The worker will be stuck for day on the task, with this error printing in an infinite loop.
2023-04-05 20:29:54: 0: WARNING - Waiting for unresponsive Render thread "Waiting for cancel for 01:24:42.3278660". Current plugin is "MayaBatch" and job "642da7ee88982d7a948bc1cb". Cannot proceed until it exits.
2023-04-05 20:30:15: 0: SandboxedPlugin still waiting for SandboxThread to exit
We have found that re-installing deadline on the worker two or three times, usually fix the issue.
Once the issue is fixed, the worked never make this error again.
We don’t have steps to fix that issue other than a re-install, we’ve never managed to re-create it locally so we can pull the issue apart. For our records, which version of Deadline are you seeing this happen in?
Thanks for you reply, and sorry about the delay of mine !
It’s deadline 10.1.21.4 e (but happen with a previous version aswell).
I found that there is a mayabatch.exe still alive in the Task manager.
killint it stopped the Sandbox pluggin wait.
Reinstalling deadline worker is not a 100% success rate, so I’m using the Execute Command from deadline monitor to run a bat that kill mayabatch.exe to free up the worker when i saw a stuck one.
I wish there was a way to autokill mayabatch.exe if the error sandbox plugin waiting to exit start happening. Is there ? I’m open to hacky solution =)
speaking of hacky solution…
I’ve wrote a script lopping every 3min. It’s read the log of the worker and kill mayabatch.exe if a sandbox waiting to exit line is find. It looks like it’s working, let’s see if that will do the trick long term…
Faced the same problem. Deadline can’t unload mayabatch.exe after render.
I noticed that during the render, the mayabatch.exe is launched as a child process of deadlinsandbox.exe. However, after rendering, the deadlinesandbox.exe ends, and the mayabatch.exe falls out of the tree and remains hanging as a separate process.
The screenshots above are taken during a normal render with option “Reload Plugin Between Tasks” enabled, not a cancelation or a require.
The steps are:
I have a paused job in the deadline. I start it. The first frame starts rendering, successfully finishes rendering, the deadline switches to the second frame and hangs in the “Waiting to start” status (and never starts rendering the second frame).
During the frame change, I took the screenshots above, in which the “old” Maya remains hanging outside the deadline process tree.
(I tested it on the one single machine from the entire farm).
I tested rendering on Maya 2022.0 and maya 2022.4.
Test disabling ‘Reload plugin between tasks’ as that might not be triggering the sandbox close correctly, causing the orphaned mayabatch.exe.
If you uncheck ‘Use Batch Plugin’ when you’re submitting the job (which will create a MayaCmd job) does Maya behave the same way by dropping out of the process tree on cancellations?
I’m trying to see if it’s Maya that’s failing to be shut down correctly, or how the sandbox is being shut down.
I enabled this option (Reload Plugin…) on purpose since in my scenes unclosed maya gives strange results. So, Maya does not close without it (= no orphaned processes).
I test my scene with MayaCmd and it works correctly. All processes terminate as expected. Looks like it is a MayaBatch issue.
I am also having issues with this (still)! There’s a couple of threads all with this issue, no solutions outlined yet except for the brute-force workaround someone posted above with a script that checks the log output:
(Mine):
