I’m trying to build a simple sequential caching pipeline in Houdini TOPs using ROP Fetch nodes, but I clearly don’t understand the intended workflow.

My goal is straightforward:

I have about 10 - 15 different cache nodes
Most but not all are unrelated to each other
I want them to run one after another
Each stage should fully finish before the next begins

Example:

Cache A (240 Frames) → Cache B (only one "Frame" needed, not time dependant) → Cache C (a Sim with 240 Frames)

Some caches are frame ranges/sims, others are single-shot/static exports.

The problems I keep running into:

Chaining ROP Fetch nodes directly seems to multiply work items/frames unexpectedly
Wait for All doesn’t seem to behave the way I expect
Downstream nodes sometimes start cooking even when upstream isn’t finished
I don’t understand when to use:
All Upstream Items are Generated
All Upstream Items are Cooked
batching
Single Frame vs Frame Range
ROP Node Configuration

I feel like I’m missing the core mental model of how TOPs wants dependencies and work item generation to work.

What is the standard/simple production workflow for sequential cache stages in TOPs? Especially when mixing:

sims
frame-based caches
static exports

I’m not trying to build a giant distributed farm setup. I mostly just want a reliable “do this, then this, then this” cache pipeline without unexpected parallelism or frame explosions.

From your description, you do need to use Wait for All.

ROP Fetch -> Wait For All -> ROP Fetch -> Wait For All -> ROP Fetch.

All the other settings you mentioned are irrelevant for your goal. Unless, alternatively, you set all your ROP Fetches to cook all frames in a single batch, it would achieve the same effect since there would now be a queue of single work items I suppose.

Previous node's work items are like a multiplier for subsequent node's work items. If you use Wait For All you bring work items back to "1".