CGRU Forum

Hi there!

I did already ask this a couple of years ago but cannot find anything in my mails anymore. My question concerns the correct setup to run an alembic ROP for multiple frames on the farm. I tried to set different values on the render range of the ROP and of the Afanasy node, but I always get unexpected results like

* I only get the last frame stored in the alembic
* Houdini runs all frames for each frame on the farm

I'm sorry if that is obvious, but I tried all kind of combinations and settings and don't get it to run correctly.

Any help is welcome.

Thanks,
Carlo

Hi!
Alembic is like a movie. All frames are paked in to a single file. And this movie should be encoded on a single machine by a single task.
So if you are going to render 1-100 frame range, frames per task should be 100. This way Afanasy ROP will produce 1 task job. You can also use Single Task checkbox for this.

Hi Timur!

Thanks for your feedback. Yes, indeed, that is exactly what I have done. But the problem is that the execution time gets longer with more frame which does not make any sense. I tried to run a single frame on the Afanasy node with frame range strict on the ROP and the number of tasks set to the number of frames. The result is a single task running the entire frame range, which is good. But when I launch n frames, the execution time per frame is n times the duration of a single frame. So, there is something fishy or I really do not understand how to set this up correctly.

But maybe there is something fishy in my network. When I try to do a simple setup, the result works as expected. I will need to dig into my stuff again to see if there is anything else that I missed. For some reason, my execution time increases exponentially as if for each frame Houdini executes all frames.

I attached a simple scene just to see if that looks correct to you.

Greets,
Carlo

And what I forgot is that I do not have the same effect when running the ROP locally. Every frame takes more or less the same time.

So, I continued my little research and I don't know if that is related to CGRU or Houdini, but when I run the abc writing job on the farm, the time per frame increases with the number of frames to render:
But when I run the same scene locally, it scales perfectly linearly.

Can you watch the output file?
May be Afanasy task rewrites it on each frame?

Hi!

I'm not really sure with what you mean by 'watching the file'. How do you want me to distinguish and update and a recreation of the file? Sorry, I'm not an IT specialist.

Greets,
Carlo

Try to watch file size.

Hi Timur!

Yes, it seems to rewrite the entire file on each frame. When I launch the job locally, my file size grows constantly but when running on the farm with CGRU, the file size starts from 0 for each frame. So, is the problem on the CGRU side or in my Houdini graph setup?

Greets,
Carlo

Hm. Indeed.
I have no deal with alebmics myself.
I see that other users wrote something alembic special.
Try create alemic export node in the ROP context and connect it with Afanasy ROP directly.