Rudder

No, we should definitely not delete reports!

I favour moving the node inventory into another branch, like "Removed inventories". This way, if the node ever comes back (was only removed temporarily, or user made a mistake deleting it), we'll be able to restore it easily.

What needs to be done to achieve this?

As discussed this morning, we can't address this in the current run, retargeting for alpha5.

I think the first step should be your points 1 & 2 above, to at least make sure we don't completely remove the data. Handling restoration and conflict resolution would be great, too, but could be done later, and benefit from the "saved" data on existing instances.

I'm starting the point 1 & 2, which should also magically solve the #2093 as a side effect (if I understand what I'm doing)

The following steps will be done :
- Create the branch "Removed inventories"
- update the Inventory api (or find where it is) to have for a given MachineId all the attached NodeId in a DIT
- change all call to the delete so that it will rather be move or copy or nothing (if I can refere to an inventory from the accepted branch)

I created a new branch in the DIT, Removed Inventories, and when a machine is deleted, now it is rather moved to the Removed Inventories branch

Technical review OK, really nice patch to read, thanks !