This message was deleted.

I do realize that longhorn is a storage solution. I am migrating from a baremetal nfs server and I can't move to a CSi model of pv/pvcs

The ideal deployment for Harvester would have many nodes to install Harvester on as a cluster. As you mentioned, Harvester uses Longhorn under the hood for VM disks, and in a cluster scenario, would replicate these disks across the Harvester nodes to prevent single-point-of-failure for storage.

the storage for the vms is very small, just enough for os. the applications are statefull sets that access a stand alone nfs server for their "home" dir. This has always been a stand alone box running ubuntu and nfs4. The baremetal machine has 8T of nvme storage across two disks. So the one disk would be used for harvester os and vm storage. I would like to use the other 4t disk for the nfs services. The question is, can I create a ubuntu vm and access this second disk effciently and reliablity.

You could use PCI passthrough to the Ubuntu VM that encapsulates the disk as an NFS server. For my homelab I pass through a SATA expansion card to my TrueNAS VM which is where all my large storage devices are plugged into.

This was pretty much the same with the other hyper visors. Is this hard to setup in Harvester

Not in my experience. Here's some additional info on it: https://docs.harvesterhci.io/v1.2/advanced/addons/pcidevices

Thank-you for the info. I am really hoping this will be easy to stand up a Harvester/rancher cluster. I am hoping that because they are siblings they should play nice together.

There's an experimental feature: https://docs.harvesterhci.io/v1.2/rancher/index#harvester-baremetal-container-workload-support-experimental But if you're using Rancher with Harvester anyways, Rancher is able to build out a k8s cluster for you using VMs as the underlying nodes.