Migrating from Longhorn longhorn_volume_size=N to longhorn_volume_size=0

[Taronyuu]

When setting up my cluster I didn't really realise the difference yet, but I am running some databases in my cluster too and the performance on Longhorn volumes is abysmal (understandably so). Other databases run very decent on separate Hetzner volumes, but a volume has an under limit of 10gb so every database needs a 10gb volume, along with volumes not having the best performance anyway I am looking for an alternative. For this I want to move to node storage because of performance reasons, and for this I will need to migrate from Longhorn using a Hetzner volume to node storage.

I've got 2 questions

1. Generally speaking, is it correct to assume that Longhorn node storage will perform better then non-Longhorn Hetzner volumes?
2. What would be the best way to approach this change?
   If I change the longhorn_volume_size to 0 it will delete all the volumes:

  # module.kube-hetzner.hcloud_volume.longhorn_volume["0-0-n1-fsn1"] will be destroyed
  # (because key ["0-0-n1-fsn1"] is not in for_each map)
  - resource "hcloud_volume" "longhorn_volume" {
      - automount         = true -> null
      - delete_protection = false -> null
      - format            = "ext4" -> null
      - id                = "100901689" -> null
      - labels            = {
          - "cluster"     = "k3s"
          - "provisioner" = "terraform"
          - "scope"       = "longhorn"
        } -> null
      - linux_device      = "/dev/disk/by-id/scsi-0HC_Volume_100901689" -> null
      - location          = "fsn1" -> null
      - name              = "k3s-longhorn-k3s-n1-fsn1-dqg" -> null
      - server_id         = 49200439 -> null
      - size              = 80 -> null
    }

This is probably not a good solution because I love my data 🥺

The only reasonable way that I see is the following steps for my 10 node cluster;

1. Cordon an drain node 10
2. Remove node 10 by lowering the count in terraform to 9.
3. Wait for Longhorn to restore degraded volumes.
4. Add a new node in a different pool having set longhorn_volume_size=0
5. Repeat step 1 to 4 another 9 times, every time make sure that every Longhorn volume is healthy before I remove a node.

Are there better solutions?