SDI Notes

Appearance

19 - Volumes (Hetzner Auto-Format) ​

Working Code: terraform/exercise-19-volume-manual/

The Problem: Server local disks are ephemeral. Delete the server, lose the data.

The Solution: Use external Volume (block storage) that survives server deletion.

Objective ​

Attach a 10GB volume using Hetzner's auto-format and auto-mount features.

How-to ​

1. Create & Attach Volume ​

Hetzner can format and mount the volume automatically:

hcl
resource "hcloud_volume" "data_volume" {
  name     = "data-volume"
  size     = 10
  location = "hel1"
  format   = "ext4"  # Hetzner formats it for you
}

resource "hcloud_volume_attachment" "main_attachment" {
  volume_id = hcloud_volume.data_volume.id
  server_id = hcloud_server.web.id
  automount = true  # Hetzner mounts it automatically
}

2. Verify ​

After terraform apply, SSH into the server:

bash
lsblk  # See the volume (usually /dev/sdb)
df -h  # Volume is already mounted (typically at /mnt/HC_Volume_*)

Hetzner automatically formats and mounts the volume, saving you manual steps.

3. Test Persistence ​

bash
echo "Important Data" > /mnt/HC_Volume_*/file.txt
# Destroy server but keep volume → data survives!

4. Identify the Volume ​

bash
lsblk

The Hetzner volume is the 10G disk — the device name (sda or sdb) can change between reboots and even from what Terraform's output shows. Always identify it by size via lsblk, not by assuming a fixed name.

5. Wipe and Partition ​

bash
sudo wipefs -a /dev/<device>

sudo fdisk /dev/<device>

Inside fdisk:

g        ← new GPT partition table
n        ← new partition
1        ← partition number
Enter    ← accept default first sector
+5G      ← last sector
n        ← new partition
2        ← partition number
Enter    ← accept default first sector
Enter    ← accept default last sector (remaining space)
w        ← write and exit

6. Format Partitions ​

bash
sudo mkfs -t ext4 /dev/<device>1
sudo mkfs -t xfs  /dev/<device>2

7. Mount Partitions ​

bash
sudo mkdir /disk1 /disk2
sudo mount /dev/<device>1 /disk1
sudo mount /dev/<device>2 /disk2
df -h | grep disk

8. Make Persistent with fstab ​

Always use UUIDs — device names change across reboots, UUIDs don't.

bash
sudo blkid /dev/<device>1 /dev/<device>2

echo "UUID=<uuid1> /disk1 ext4 defaults 0 0" | sudo tee -a /etc/fstab
echo "UUID=<uuid2> /disk2 xfs  defaults 0 0" | sudo tee -a /etc/fstab

sudo mount -a    # test — should produce no errors

9. Reboot and Verify ​

bash
sudo reboot
# reconnect with ./bin/ssh
df -h | grep disk    # both /disk1 and /disk2 should still be mounted

Problems & Learnings ​

Common Issues

  • Device name changes between reboots — the volume may be sda or sdb depending on boot order. Always use lsblk to identify it by size, and use UUIDs in /etc/fstab (never /dev/sdX)
  • automount = true pre-creates a partition — if the Terraform config uses automount = true, Hetzner will format the disk as a single partition before you log in. Use sudo wipefs -a /dev/<device> to clear it before running fdisk
  • fdisk +5G goes on the Last sector prompt, not First sector — press Enter to accept the default first sector, then type +5G

Key Takeaways

  • Use UUIDs in /etc/fstab for reliable persistent mounts regardless of device naming
  • lsblk is the right tool to identify disks; blkid gives you the UUIDs
  • sudo mount -a tests your fstab before rebooting — a typo here can make the server unbootable