1. Qube
2. Nuke Job Submission

Nuke Job Submission

What's the difference?

loadOnce will open a copy (instance) of Nuke on the Worker that picks up the job, and that instance will render frames as assigned by the Supervisor. Since the scene file is only being opened one time, this can be faster to generate the full sequence.

cmdline will use the command line to open a new instance of the Nuke file for each frame that it renders. Although this can be slower for large scenes, it can also be more flexible in large, complex environments.

While not strictly required, the following settings are useful for getting a better result. You will need to turn on "Expert Mode" (check box at the bottom of the submission UI) in order to get access to them.

• Make sure that the Nuke version and the OS are set correctly for the Worker(s) that will execute the job.
• Optimize the use of cores. Set the "Slots = Threads" checkbox and then set the "Specific Thread Count" to a value like 8 (assuming you have 8 cores). (If you have Designer licenses, only set the thread count. Do not check "Slots = Threads")
• By default, Qube! will not retry failed frames. So set the "Retry frame/instance" value to 3 or 5, to get 3 or 5 retries before it gives up. Related to that, the default wait between retries is zero seconds, which is typically not useful. Set the "Retry Work Delay" to a value like 5 or 10 to allow machines time to recover from temporary problems such as network errors.
• Because the path to the Nuke executable is part of the submission parameters, it is not possible to mix OSs with Nuke jobs - that is, although you can submit from, say, Windows to OS X, you cannot submit to both Windows and OS X at the same time.