4.3 deprecates inventory items and suggests Modules

[leapswitch-repos]

https://netboxlabs.com/docs/netbox/models/dcim/inventoryitem/

Deprecation Warning
Beginning in NetBox v4.3, the use of inventory items has been deprecated. They are planned for removal in a future NetBox release. Users are strongly encouraged to begin using modules and module types in place of inventory items. Modules provide enhanced functionality and can be configured with user-defined attributes.

Will Netbox-sync evolve to use Netbox-Inventory, or use Modules ?

[bb-Ricardo]

probably not: #474

[marcinpsk]

I do have working modules in my fork, I'll try to get a PR at some point, @semx

[semx]

@marcinpsk a Modules-based PR would be very welcome; please target development. Since inventory items stay in use until NetBox removes them, keeping the current inventory-item path working alongside would make it straightforward to land.

[marcinpsk]

@semx sure thing, I've put it there. It needs a flag to enable, and Netbox 4.3, so it coexists with inventory path. Disabling flag flips it back to write inventory.

[semx]

@bb-Ricardo you answered "probably not" here in July 2025 when asked whether netbox-sync would move to modules, and I have been treating that as the standing decision. Three things have changed since, so it is worth asking again rather than leaving #560 open indefinitely.

NetBox deprecated inventory items in 4.3 and says they are planned for removal, so the current path has an end date that is not ours to control. @marcinpsk built the modules path and it is in #560. It is opt-in (model_components_as_modules, off) and additionally gated on NetBox 4.3, so the inventory-item path stays exactly as it is for everyone who does not ask for it.

I ran it against a real NetBox 4.7 rather than the test inventory and fixed the two things that were wrong: modules were not read back from NetBox, so every second run threw 230 errors trying to recreate them, and the module type catalog was being filled with per-instance names instead of hardware models. Both are fixed on the branch, second run is now 0 errors and 0 changes, and switching the option on tags the old inventory items orphaned so the migration cleans up after itself.

So the question is just the direction, not whether the code holds: do you want this in v1.9.0, kept on the branch until inventory items actually go away, or dropped? Happy with any of the three, I would just rather not leave it hanging.

[marcinpsk]

@semx completely missed that. Thanks for catching this.
One remark - I didn't include ignoring speed on Fans - and issue I had with it - it was getting live readings from some - so constant churn - but at the same time - it was in inventory as well.

[semx]

@marcinpsk good catch, and it was worse than just the modules: both paths took speed from the same item data, so the fan was rewritten on every run as an inventory item too. Reproduced it against a real NetBox by bumping the reading between two runs, one write per fan per run, all of it in the changelog. #568 drops the fan reading and is merged; the other components keep speed because theirs describes the part (DIMM clock, drive rpm rating, NIC link speed) rather than measuring it. The fan still syncs with name, health and context.

One I left alone: a CPU's current_speed could drift the same way on hardware that reports the live clock. If you have seen that in your fleet, say so and I'll treat it the same.

[bb-Ricardo]

Hi,

the "probably not" was due to the lack of time to implement all this. As you both seem to have done that, I don't see a reason not to include it.

The config switch is a good option to let the user decide when they want to move from inventory to modules. Once it is removed, we can drop the NetBox inventory part and remove the config option.

[semx]

Thanks @bb-Ricardo, appreciate you reconsidering.

Where it stands: #560 has the switch @marcinpsk built — needs NetBox 4.3, coexists with the inventory path, and flipping it off writes inventory items again. I ran both directions end-to-end against a live NetBox this week: with the option on the second pass is a no-op, and turning it off resolves cleanly and re-creates the inventory items. I also pushed one fix there — with the option off, existing module references on interfaces and power ports were not being read back, so the migration-off path logged resolve errors. That is sorted now.

So both the forward path (inventory -> modules) and the way back are covered. #560 should be ready for review whenever you have time.