AWS Thinkbox Discussion Forums

Texture conversion triggers every time when rendering with AWS

I’m trying to reduce render time for my Redshift Standalone jobs when using AWS Portal. Right now when I render a new job I get messages like this:

STDOUT: Texture file '/mnt/Data/LProjects31a1defa2b7eeac7db882ca3532023d4/ABC.png' conversion triggered, because: it was not found!

And spend 12 minutes processing textures, which is fine. But then I resubmit the same job, on the same instance, without modifying any textures and get this:

STDOUT: Texture file '/mnt/Data/LProjects31a1defa2b7eeac7db882ca3532023d4/ABC.png' conversion triggered, because: date stamp did not match source!

And still spend 4 minutes on texture conversion. I believe it’s because Asset Server uploads textures every time job is started, without actually checking if they were modified. Is there a way to prevent this so textures don’t get converted every time?

Hello,

the texture conversion, is on the redshift side, not the deadline side.
texture file are re converted if the original file changes, ie if the creation date of the png, is newer than the rstexbin it converts.
you could avoid this conversion time, if you pre convert the rstexbin files and have then in the same folder as the png, just make sure they get uploaded to the farm/ instance and redshift, will use the rstexbin, instead of the png file, therefore no conversion needed.

The thing that AWS Portal may make more difficult here is that we avoid copying back the TX files as they can be fairly large, and each of the Linux machines is an island as far as assets are concerned (they checkout their own copy from S3). Converting the textures up front may cause the assets to take longer to make it to the render machines but would save the conversion time.

It’s something to experiment with.

Privacy | Site terms | Cookie preferences