Currently, OSP Director assumes a statically configured, uniform hardware profile for all storage nodes. Rather than require this profile be defined upfront in the Puppet/Heat templates, this feature is provide a facility to allow OSP Director to dynamically discover the hardware profile of a storage node. The user could then select which drives perform which function: OS, journal or OSD and the relations between them. This would also allow the customer to have more than one type of hardware profile for their storage node, recognizing that some customers use varying server models over time.
From OSPD8 we have a general mechanism to distribute node-specific hieradata regardless of the resource group a node belongs to [1] and it allows the deployer to push, for example, a specific disk configuration on a node-by-node basis. The node-specific hieradata is provisioned based on the machine unique UUID, which has to be gathered from the Ironic introspection data, i'm trying to push some documentation for this in the upstream docs [2]. We *could* add a Mistral workflow to remove some manual steps and drive it from the UI. 1. https://bugzilla.redhat.com/show_bug.cgi?id=1238807#c14 2. https://review.openstack.org/#/c/332142/