Hello community! Does anyone have any experience with using Harvester with local HDDs configured as JBOD (using the RAID controller in HBA mode)?

with longhorn on top? why not just put them in longhorn directly?

Well longhorn is deployed with harvester. Otherwise, perhaps I don't understand your suggestion properly.

Yes, with Longhorn managing your redundancy you can use the onboard RAID controllers in HBA mode to achieve the maximum amount of storage for the device. We do the same thing in our server configurations.

oic

The osdisk is ssd, I plan to use the other disks as defaultdisk for the workloads

Only using SSD/NVMe disks.

Btw @strong-shoe-72392 ever noticed data loss due to not using RAID, or does Longhorn provide the safety net for it? I understand that using both RAID (e.g. RAID10) and Longhorn provides a tiered approach to data loss (at the node level and cluster level), however my question is do you feel safe enough for production only with Longhorn handling replication?

Yes - we use Longhorn replication for non-RAIDed setups on both enterprise server hardware and on edge device clusters. Longhorn handles the replication just fine. It is chatty and requires decent network bandwidth, but it has worked for us.

Thanks for the feedback @strong-shoe-72392

How did you configure your disks in LH? Through the Harvester (or embedded Longhorn) web interface, or through K8s manifests?

@abundant-camera-87627 I created the RAID10 array using the controller and then configured the disk during the Harvester setup.

Yeah, so only one disk to select. I am more interested in David's approach and how to install it fully unattended

I think it's just this?

k get lhn -A -o json | jq '.items[] | .spec.disks'

I have only manually added the additional drives through Harvester UI post-install/config. But, I'd be very interested in automating it!

There is this option in the Harvester settings:

auto-disk-provision-paths