Mirko Venturi

Hi there.
Strange issue with Pdg and Deadline workers on different machines seemingly randomly choosing to not start all the concurrent tasks, but just one.
I am using a deadline scheduler in a houdini topnet with concurrent tasks parameter set to 4, 8 or 12, but i get random effects:
a) It works fine, each worker starts the amount of tasks assigned.
b) One worker starts multiple tasks, the other starts only 1 task (there seems to be no rule for which worker starts multiple tasks or just one)
c) SOMETIMES it seems to work like a) after restarting Houdini.

I’ve had this problem intermittently throughout the years so I’d love to find out what it is that I’m doing wrong.

Hi. I have a strange issue with path mapping in a mixed Windows/Linux environment that works fine in a situation but does not in another.
I’ll try to explain.

I have successfully installed and configured Deadline and Houdini(for headless use) on a Linux Ubuntu machine as a test.
My main Houdini workstation where i submit the jobs from has Windows installed.
If i submit a file cache node from Windows using the manual "Render -> submit to deadline” submitter, everything works fine.
If i submit a PDG job using the “submit” button from the deadline scheduler in the TOP context, everything works fine.
If i have a PDG job where i want to start the task directly from (I.E.) a ropfetch node (right click - cook or shift-g) and have it use a deadline scheduler instead of a local one, the PDGDeadline.py in the Linux worker’s plugin folder still has Windows paths, and the task fails.
How can i solve this?
I suspect it has to do with the Paths parameters in the Deadline Scheduler node that I'm cooking the node with, but between that and Deadline's repository path mapping I'm losing my orientation a little.

Hi
is it possible to apply a per-particle divergence attribute and have them react according to that value?
I.e. every particle has a random value, so particles with 0.1 will stick closer compared to particles with 0.5 divergence.