Distributed bucked - slave ignores GI pass?

[vlado]

Ok then. I've modified the code for the render slaves so that if they fail to load an irradiance map file, or fail to receive it from the client machine, they will at least cancel the render so that they do not contribute wrong buckets.

Best regards,
Vlado

[KiboOst]

Yes nice, but could you find/fix why it doesn't receive entire file ? Not having wrong bucket is better indeed, but having good bucket is even better and faster, which is the purpose of DRs (faster)

[vlado]

Yes nice, but could you find/fix why it doesn't receive entire file?

No, I could not reproduce it unfortunately, although I tried quite hard (using 8 machines from our blade server with quite large irradiance maps), and I did not see an obvious problem in the code (of course, it's probably non-obvious ). I added a message to print out some error codes in case this happens, so it might help to track the issue if it happens again on your side.

Best regards,
Vlado

[KiboOst]

Well, it happens everytime we try to use DRs. We usually use one host and 4 DRs, on complexe interior scenes (I can send one if you need).
If you need any beta version or debug utility to be run here, let me know, I can deploy it, do some test to report errors to you, and deploy back actual version. We really need to find a way to use DRs (we are looking for alternative solution, but not convinced yet), so if I can help you to track the bug, I will do.

[vlado]

You can email me to vlado@chaosgroup.com and I'll get you access to the nightly builds; then we can work from there - I could add some debug code to track down the issue as needed.

Best regards,
Vlado