Geordie Moffatt

GeordieM

About Me

EXPERTISE
CG Supervisor

Connect

LOCATION
United States
WEBSITE

Houdini Skills

Availability

Not Specified

Recent Forum Posts

Relationship between PDG Scheduler python and HDA July 15, 2024, 5:16 p.m.

Random notes:
  • Found this https://sourcegraph.com/github.com/CGRU/cgru/-/blob/plugins/houdini/pdg/types/afanasyscheduler.py [sourcegraph.com]

    Where they don't even use templateBody
  • Found that when you convert a Scheduler defined by a python file to a HDA there's two schedulers listed, until you restart Houdini
  • If you define a parm in templateBody after you've "converted" the script to HDA, it doesn't automatically show up in the UI, but it exists somehow in the context of the TOP node, since if you try and create a parm in the HDA properties with the same name it adds the 2 after it. So not sure how you're supposed to add new parms to templateBody after you've converted to HDA.

Relationship between PDG Scheduler python and HDA July 12, 2024, 8:21 p.m.

Hi all, I'm creating a new python PDG scheduler and trying to work out the exact relationship between it's python definition and it's TOP node HDA and UI in general.

If I register a custom python scheduler and want to edit the UI seems like I need to create a HDA from it. But once I've done that what's the relationship between the python script and the HDA and the "templateBody"?

I'm also hard pressed on finding any info about how templates work as well.

Looking at the tractor and deadline python code isn't much help, everything seems to have a mish-mash of some UI params defined in templateBody, some in the HDA. Then some params are prefixed with I PDG_ and some with the parmprefix defined in the scheduler python class.

Can someone help fit this all together? Thanks!!

Python Script TOP node will only cook on a single HQ client March 4, 2021, 4:51 p.m.

Hi all, I have a HQueue farm with 15 clients, server can detect them all and running render ROPs though PDG will use all the clients. But when I try and run a Python Script node in "out of process" mode it only uses a single client. I have the HQueue scheduler set to only use a single CPU, so I'd expect every CPU on every client to be given a workItem. When I try to process 1000+ Python workItems it seems to send them in batches of 30 and only to a single client. We think the batch of 30 thing is only happening when running the job in the TOPs UI, not when submitting the job from the HQueue server.

Anyone know about this behaviour?

Thanks!