Fusion FAQ

• 1 Fusion FAQ
  • 1.1 Which verions of Fusion are supported by Deadline?
  • 1.2 Can I use both workstation and render node licenses to render jobs in Deadline?
  • 1.3 Why is it not possible to have to 2 instances of Fusion running?
  • 1.4 Why do I get lots of Fusion Render Node icons showing up which go away when I move the mouse over them?
  • 1.5 Fusion alone renders fine, but with Deadline, the slaves are failing on the last frame.
  • 1.6 Is there a way to increase Deadline's efficiency when rendering Fusion frames that only take a few seconds to complete?
  • 1.7 Does Fusion cache data between frames on the network, in the same way it does when rendering sequences locally?
  • 1.8 Fusion seems to be taking a long time to start up when rendering. What can I do to fix this?

Which verions of Fusion are supported by Deadline?

Digital Fusion 4.x and Fusion 5.x are supported. For more information go to the Fusion Network Rendering Guide.

Can I use both workstation and render node licenses to render jobs in Deadline?

You can use workstation licenses to render, you just need to do a little tweaking to get this to work nicely. In the Plugin Configuration settings, you need to specify two paths for the RenderExecutable option, separated by a semicolon. The first path will be the render node path, and the second will be the actual Fusion executable path. You then have to make sure that the render node is not installed on your workstations. Because you have specified two paths, Deadline will only use the second path if the first one doesn't exist, which is why the render nodes can't be installed on your workstations.

Why is it not possible to have to 2 instances of Fusion running?

With Fusion there is only one tcp/ip port to which dfscript/eyeonscript (the scripting language used to run Fusion renders on a slave computer) can connect. If Fusion is open on a slave computer then the port will be in use and the Fusion Render Node will have to wait for the port to become available before rendering of Fusion jobs on that slave can begin.

Why do I get lots of Fusion Render Node icons showing up which go away when I move the mouse over them?

These are ghosts of the Fusion Render Node processes which Deadline was not able to shut down properly. There are two ways that Deadline can use to shut down the Fusion Render Node, either by running a DFScript which tells the Fusion Render Node to exit, or by sending a Windows message to the application telling it to close. Deadline attempts to do the second option, and generally succeeds at it, but occasionally there are some kinds of flows/comps which tend to prevent the Fusion Render Node from shutting down properly. In these cases, Deadline proceeds to terminate the Fusion Render Node process, and because of that the icon doesn't get cleaned up properly.

Fusion alone renders fine, but with Deadline, the slaves are failing on the last frame.

This is usally accompanied by this error message:

   * INFO: Checking file \\path\to\filename####.ext
   * INFO: Saver "SaverName" did not produce output file.
   * INFO: Expected file "\\path\to\filename####.ext" to exist.

The issue likely has to do with the processing of fields as opposed to full frames. When processing your output as fields, the frames are rendered in two “halves” (for example, frame 1 would be rendered as 1.0 and 1.5). This error often occurs when the Global Timeline is not set to include the second half of the final frame. Simply adding a “.5” to the Global End Time should resolve this issue.

For example, let us assume that you are processing fields and your output range is 0 - 100. If the Global Timeline is set to be 0.0 - 100.0, Fusion will render everything, but Deadline will fail on the last frame. If the Global Timeline is set to be 0.0 - 100.5, Deadline will render everything just fine.

Is there a way to increase Deadline's efficiency when rendering Fusion frames that only take a few seconds to complete?

Rendering these frames in groups (groups of 5 for example) tends to reduce the job's overall rendering time. The group size can be set in the Fusion submission dialog using the Task Group Size option.

Does Fusion cache data between frames on the network, in the same way it does when rendering sequences locally?

Deadline renders each block of frames using the dfscript flow.render function. The Fusion Render Node is kept running between each block rendered, so when Fusion caches static results, it can be used by the next block of frames to be rendered on the same machine.

Fusion seems to be taking a long time to start up when rendering. What can I do to fix this?

If you are running Fusion off a remote share, this can occur when there is a large number of files in the Autosave folder. If this is the case, deleting the files in the Autosave folder should fix the problem.