Hi,

QUESTION 1:
———–

I think I'm having a misconception about how a PDG graph is supposed to run on HQueue. I have this scene where I need to send a single ROP Fetch node to the farm (rendering multiple frames). For now I have small testing environment with 2 render nodes only, and want to split renders between these machines. I also want to use the “Submit Graph as Job” section, so I can close and disconnect the current Houdini session from pdg.

The HQueue job hierarchy is created like this:

(parent) Top cook
(child) render frame 1
(child) render frame 2
(child) render frame 3
…

First client(A) gets assigned to the parent job, the machine is now locked to that job, waiting for all child jobs to terminate. All child jobs get assigned to second client(B). This is obviously wasting all power resources from (A).

I can make both clients work together by changing the number of CPUs per job on the “Job Parms” tab, but then I can't properly set a reasonable number of CPUs without having a portion of that locked to that parent job (which is, in theory, doing nothing but waiting on child jobs to finish).

So my question is: do that parent job really needs to lock resources while waiting for the child items to complete ? Any way to circumvent that ?


QUESTION 2:
———–

Are there any ways to gain control over job names created from built-in TOP nodes. For example: sending a ROP fetch node to farm would create a series of jobs called “render_1” or “render_1_2” or names like these on HQueue. Can I somehow (python or whatever) overwrite these workitem names before sending them to the farm ?



thank you!

any thoughts on this? I'm mainly concerned about the parent job locking resources on the farm.
thanks.