How do I set up Nuke and Toolkit to run on a render farm?

I have troubles setting up Nuke to render correctly on the farm.

The WriteTank nodes cause two problems:

1. Sgtk is not initialized when using the command line nuke -x to launch a batch render. I get errors like the following:

Traceback (most recent call last): 
File "<string>", line 1, in <module>
AttributeError: 'module' object has no attribute '_tank_write_node_handler'
'module' object has no attribute '_tank_write_node_handler'

What is the proper setup for Nuke to be able to render Tank nodes in batch?

2. If I send temp files to the render nodes, then Sgtk can't be initialized
because the context can't be determined from the /tmp file. This causes
troubles as we don't want to submit the work file directly (imagine an user
sending a job to the farm, then modifying his work scene and saving the
changes. the render nodes should therefore receive a copy of this file)



  • 0
    Manne Öhrström

    Hello Benoit! Great questions! I'll do my best to answer! :)

    When you run nuke in batch render mode and your scene contains Shotgun Write Nodes, these write nodes are connected to code which handles refresh logic and other functionality. This basically means that there needs to be an initialised toolkit instance running while you are rendering the scene. This is not very complicated to set up and we have documented the process over in the app store documentation for the Write Node. Scroll down to the Render Farm Integration section and there are some code examples!

    For the second question - what happens when your nuke file is copied to a temp location for rendering:

    By having a special script that starts up the nuke engine on the farm (as described in the tank write node docs), you can get around the dependency on the current file path, so that you can start nuke even if you scene is in a temp folder. All the tank write nodes will be locked at that point - meaning that you can still render to them but if you try to do a publish the file when it resides in the temp location, toolkit will issue warnings about these locked nodes. The idea is that you should still be able to do the rendering from a temp folder, but publishing and sending to review happens from the work file's original location. Again, there is some documentation about this in the app store documentation for the Write Node. Do you think that workflow will work for you?

  • 0
    Benoit Leveau

    Looks like I missed an important point in the documentation! This workflow should definitely work for me, if it doesn't for some reason I'll ask more details.

    And thanks for explaining that part about the locked nodes.



Please sign in to leave a comment.