Mount Point Naming

Technical Background

By default, when using Hetzner Cloud volumes with automount = true, Cloud-init generates a random mount point like /mnt/HC_Volume_<id>.
In many production systems, predictable mount points (e.g., /volume01) are preferred for simplicity and automation.

To achieve this, volume and server creation must be decoupled: - The volume is created independently from the server. - The server attaches the volume but does not automount it. - Cloud-init handles directory creation, /etc/fstab entry, and mount execution.

This avoids a cyclic dependency (server -> volume -> server) and ensures Cloud-init receives all required parameters for correct mounting.

Solution

Prerequisits

Create a main.tf, outputs.tf, /tpl/* variables.tf, network.tf,providers.tf and secrets.auto.tfvars like in 15 Partitions And Mounting.

Edit Volume

1. Disable automount for the volume resource inside volumes.tf:

resource "hcloud_volume" "volume01" {
    name      = "volume1"
    size      = 10
    location = "hel1"
    server_id = hcloud_server.web.id
    format    = "xfs"
    automount = false
}

1. Add volume attachment to volumes.tf:

resource "hcloud_volume_attachment" "volume01_attachment" {
    server_id = hcloud_server.web.id
    volume_id = hcloud_volume.volume01.id
}

Passing Parameters to Cloud-init

Edit your user data local_file ressource inside your main.tf to contain:

resource "local_file" "user_data" {
  content = templatefile("tpl/userData.yml", {
    host_ed25519_private = indent(4, tls_private_key.host.private_key_openssh)
    host_ed25519_public  = tls_private_key.host.public_key_openssh
    devopsSSHPublicKey   = hcloud_ssh_key.loginUser.public_key
    volume_id            = hcloud_volume.volume01.id
  })
  filename = "gen/userData.yml"
}

Updating Cloud-init Template

Modify the tpl/userData.yml by adding these:

runcmd:
  # Wait until the volume appears (max 20 seconds)
  - |
    for i in $(seq 1 20); do
      if [ -e /dev/disk/by-id/scsi-0HC_Volume_${volume_id} ]; then
        echo "Volume is available."
        break
      fi
      echo "Waiting for volume to appear..."
      sleep 1
    done

  # Create mount point
  - mkdir -p /${volume_name}

  # Trigger udev so volume is discoverable
  - udevadm trigger

  # Format the entire volume with the XFS file system
  - mkfs.xfs -f /dev/disk/by-id/scsi-0HC_Volume_${volume_id}

  # Write mount line into /etc/fstab
  - echo "/dev/disk/by-id/scsi-0HC_Volume_${volume_id} /${volume_name} xfs discard,nofail,defaults 0 0" >> /etc/fstab

  # Reload systemd and mount volumes
  - systemctl daemon-reexec
  - mount -a

  # Create Partitions
  - fdisk /dev/sdb
  - mkfs.ext4 /dev/sdb1
  - mkfs.xfs /dev/sdb2
  - mkdir /disk1 /disk2
  - mount /dev/sdb1 /disk1
  - mount /dev/sdb2 /disk2

  # Fail2ban setup
  - systemctl enable fail2ban
  - systemctl start fail2ban
  - >
    echo "[sshd]
    enabled = true
    port = ssh
    logpath = %(sshd_log)s
    maxretry = 3
    bantime = 3600" > /etc/fail2ban/jail.d/ssh.conf
  - systemctl restart fail2ban

  # Nginx setup
  - systemctl enable nginx
  - rm /var/www/html/*
  - >
    echo "I'm Nginx @ $(dig -4 TXT +short o-o.myaddr.l.google.com @ns1.google.com) 
    created $(date -u)" >> /var/www/html/index.html

  # Plocate indexing
  - systemctl enable plocate-updatedb.timer
  - systemctl start plocate-updatedb.timer
  - updatedb

Display Volume Infos

Add this output to your outputs.tf to expose the volume device path and id:

output "volume_device_name" {
  value       = hcloud_volume.volume01.linux_device
  description = "The volumes device name"
}

output "volume_device_id" {
  value       = hcloud_volume.volume01.id
  description = "The volumes id"
}

Apply and Verify Changes

1. Apply Terraform configurations:

terraform init
terraform apply

1. Reboot the server

sudo reboot

1. Check mounts:

df -h | grep volume01

Related Links

Hcloud Volumes

Terraform Hetzner Docs

Hcloud Volume Attachment