Longhorn volume size

[transmogrifix]

Hi,

I have the following configuration in my kube.tf file:

agent_nodepools = [
    { # Node for TimescaleDB with 500GB volume.
      name        = "agent-cx41-l500",
      server_type = "cx41",
      location    = "nbg1",
      labels      = [
        "node.kubernetes.io/server-usage=tsdb",
        "node.kubernetes.io/storage-type=longhorn",
        "node.kubernetes.io/longhorn_volume_size=500"
      ],
      taints      = [],
      count       = 1,
      longhorn_volume_size = 500
    },
    { # A general node for application pods and PostgresSQL with local NVME storage.
      name        = "agent-cx51-l0",
      server_type = "cx51",
      location    = "nbg1",
      labels      = [
        "node.kubernetes.io/server-usage=general",
        "node.kubernetes.io/storage-type=longhorn",
        "node.kubernetes.io/longhorn_volume_size=0",
        "node.longhorn.io/create-default-disk=config"        
      ],
      taints      = [],
      count       = 1,
      longhorn_volume_size = 0
    }
  ]
  enable_longhorn = true

I was under the impression that when I deploy, node from the first pool will get a 500GB Hetzner volume created and attached. What actually happens is longhorn uses local node disks and no Hetzner volume gets created.

I have tried with versions 1.10.9 (some others from 1.9.x range as well but not sure which ones exactly) and 2.0.0.

I have searched through the issues and discussions on this project and tried to follow the excellent best practices gist https://gist.github.com/ifeulner/d311b2868f6c00e649f33a72166c2e5b by ifeulner but I can't seem to get a volume created and used automatically.

I have worked around the problem by creating a volume manually, attached it to the node and added it manually via the Longhorn UI. It works fine this way but I would really like to avoid this.

Any ideas or suggestions?

Thank you!

[mysticaltech]

@transmogrifix This is the code taking care of this:

terraform-hcloud-kube-hetzner/agents.tf

Lines 99 to 142 in 192866e

	resource "hcloud_volume" "longhorn_volume" {
	for_each = { for k, v in local.agent_nodes : k => v if((lookup(v, "longhorn_volume_size", 0) >= 10) && (lookup(v, "longhorn_volume_size", 0) <= 10000) && var.enable_longhorn) }
	labels = {
	provisioner = "terraform"
	cluster = var.cluster_name
	scope = "longhorn"
	}
	name = "${var.cluster_name}-longhorn-${module.agents[each.key].name}"
	size = lookup(local.agent_nodes[each.key], "longhorn_volume_size", 0)
	server_id = module.agents[each.key].id
	automount = true
	format = var.longhorn_fstype
	}
	resource "null_resource" "configure_longhorn_volume" {
	for_each = { for k, v in local.agent_nodes : k => v if((lookup(v, "longhorn_volume_size", 0) >= 10) && (lookup(v, "longhorn_volume_size", 0) <= 10000) && var.enable_longhorn) }
	triggers = {
	agent_id = module.agents[each.key].id
	}
	# Start the k3s agent and wait for it to have started
	provisioner "remote-exec" {
	inline = [
	"mkdir /var/longhorn >/dev/null 2>&1",
	"mount -o discard,defaults ${hcloud_volume.longhorn_volume[each.key].linux_device} /var/longhorn",
	"${var.longhorn_fstype == "ext4" ? "resize2fs" : "xfs_growfs"} ${hcloud_volume.longhorn_volume[each.key].linux_device}",
	"echo '${hcloud_volume.longhorn_volume[each.key].linux_device} /var/longhorn ${var.longhorn_fstype} discard,nofail,defaults 0 0' >> /etc/fstab"
	]
	}
	connection {
	user = "root"
	private_key = var.ssh_private_key
	agent_identity = local.ssh_agent_identity
	host = module.agents[each.key].ipv4_address
	port = var.ssh_port
	}
	depends_on = [
	hcloud_volume.longhorn_volume
	]
	}

It should work and was working in the past. I will test myself and let you know.

[mysticaltech]

@transmogrifix Thanks for finding this bug! It's been fixed in v2.0.1. You should be able to just terraform init -upgrade to install the missing volumes and config.

[transmogrifix]

Thank you for fixing it so fast!

Also a big thank you to everyone who contributed to this great project!

[valkenburg-prevue-ch]

Oh wow, and here I was wondering why ssd storage seemed just as fast as node storage (and why longhorn storage size did not reflect the ssd size)....!
I guess this was the fix? 47dc93c

[mysticaltech]

@valkenburg-prevue-ch Exactly!