9 min read

Learning Objectives

By the end of this module, you will be able to:

  • Choose the Block Storage class, size and zone for boot and data volumes, knowing which attributes are fixed at provisioning
  • Select an image source and plan around its location boundary
  • Design first-boot configuration with cloud-init, within its one-time and Linux-only scope
  • Attach and configure FinCorp's boot and data volumes, with credentials and cloud-init user data, in the Data Center Designer (DCD)

Unit 4.2: Images, Disks, and Cloud-Init

Introduction

A server shell from Unit 4.1 has cores, RAM and a CPU family, but nothing to boot. This unit adds the volumes, and with them a set of decisions that are hard to undo: a volume's storage type and Availability Zone are fixed when it is provisioned, an image is usable only in its location, and cloud-init runs only on the first provisioning of a volume.

Each of these is a design input. Get them right on the first build, and the scaling and migration units later in the course start from clean volumes. The unit ends by attaching FinCorp's boot and data volumes in the DCD.

1. Disks: Class, Size and the Fixed Attributes

1.1 How Block Storage Behaves

A Block Storage volume is equivalent to an Internet Small Computer System Interface (iSCSI) block device and behaves like direct-attached storage. It is managed independently of servers, and each volume is replicated four times and stored on distributed physical devices within the selected data center location. Replication stays inside the location; IONOS CLOUD does not create cross-region redundancy for you.

Unit 2.4 compared the three classes, Essential, Balanced and Performance, and their prices. The class rule for this build is the performance floor: Balanced and Performance scale with volume size, IONOS CLOUD recommends at least 100 GB to get the full benefit of the SSDs, and volumes above 600 GB reach the maximum expected performance. A second condition is easy to miss. The maximum full-random rates per volume are reached only with a minimum server size: 2 cores and 2 GB RAM for Balanced, 4 cores and 4 GB RAM for Performance. A Performance volume on a one-core server does not deliver its ceiling.

Every class takes volumes from 1 GB to 4 TB; larger volumes are available on request.

1.2 Boot and Data Volumes

A server boots from a storage volume, a CD-ROM or a NIC. You mark one volume as the boot device and choose its bus type: VirtIO, the recommended type, or IDE for troubleshooting when the operating system has no VirtIO drivers installed. Windows guests need the VirtIO drivers.

Keep the boot volume small and replaceable, and put state on separate data volumes. The reason is lifecycle. Boot volumes are rebuilt from images; data volumes are moved, resized and snapshotted on their own schedule. A separate data volume is also what the storage re-attachment path in Unit 4.1 moves between instances.

1.3 What You Cannot Change Later

The following table lists the volume attributes and when they can change.

Attribute After provisioning Design consequence
Storage type (class) Cannot be changed Choose the class per volume at creation; a class change means a new volume
Availability Zone Cannot be changed A new zone requires a snapshot and a new volume from it
Size Can be increased, even while the server runs, if the operating system supports it; cannot be reduced Start at the class floor, not at the eventual size
Cloud-init user data Cannot be set or changed Decide first-boot configuration before the first provisioning

As shown above, only size moves, and only upward. Block Storage offers Zones 1, 2 and 3 and Auto, while servers offer Zones 1 and 2 and Auto. Pin the volumes of a redundant pair to the same zone as their server, so that each half of the pair sits on separate physical resources, and avoid Auto for volumes whose placement you must document.

2. Images and the Location Boundary

Three image sources exist, and each is bound to a location.

Public images. IONOS CLOUD provides operating system images, including AlmaLinux, Debian, Rocky Linux, Ubuntu, Red Hat Enterprise Linux and Windows Server 2016, 2019, 2022 and 2025. All images are frequently updated. When a new patch is provided, a new image is built and the previous version is removed from the catalog, without a separate notice; the version number is in the image name. A build that names "the Ubuntu image" today gets a different image next month. For reproducible tiers, record the image name per build, or capture a known-good state as your own image or snapshot.

Private images. You upload your own images over FTP to the server of each data center location (Unit 1.3 lists the formats). Your own images are only available in the region where you uploaded them, so an image uploaded to Frankfurt can only be used in a VDC in Frankfurt. After upload, set the image metadata: OS type, Cloud Init Support and UEFI Support. Images for Intel Sierra Forest and AMD EPYC Turin servers must be UEFI-compatible, because those CPUs require UEFI for booting.

Snapshots. A snapshot captures a provisioned Block Storage volume. Snapshots are limited to the data center location where they were created, cover the entire capacity of the volume including unwritten space, are not incremental, and are not deleted automatically. A new volume from a snapshot must be at least the snapshot's size. Unit 5.7 places snapshots in the data-protection plan.

For a multi-region design, the boundary is the point. FinCorp's recovery VDC fc-prod-txl-dr in Berlin (Unit 3.6) cannot use an image uploaded to Frankfurt or a snapshot taken there. Upload golden images to each location the design uses, and keep the build pipeline, not one location's image store, as the source of truth.

3. Cloud-Init for First-Boot Configuration

Cloud-init automates the initialization of a server during boot. On IONOS CLOUD, it is available on all public Linux images; a custom Linux image must support it, and Windows images and Linux LVM are out of scope. User data is written as a shell script or as cloud-config directives in YAML syntax, and existing cloud-init configurations from other providers are compatible.

The scope rule shapes the design: cloud-init can only be set at initial provisioning. It is a property of the volume, set when the volume is created, and it cannot be applied to instances that have already been provisioned or changed afterwards. Cloud-init is therefore a bootstrap, not a configuration manager. Use it to bring a server to the point where your configuration tooling takes over: users, keys, an agent, a repository, a first package set. Keep anything that changes over the server's life in that tooling.

A short cloud-config document shows the shape:

#cloud-config
package_update: true
packages:
  - nginx
runcmd:
  - systemctl enable --now nginx

Two operational details matter for the build. The DCD reports provisioning as complete before bootstrapping finishes, so allow extra time before testing. And the results are in /var/log/cloud-init-output.log, with a more detailed record in /var/log/cloud-init.log.

SSH key injection follows the same boundary: it is available for public IONOS CLOUD Linux images. For a private image, build the access configuration into the image itself.

DCD Implementation Walkthrough

You will extend the servers sized in Unit 4.1: a Balanced boot volume with a public Linux image, credentials and cloud-init on each portal server, and a boot volume plus a separate Performance data volume on each data-tier server. This realises the boot-versus-data split and the class floor from Section 1.

Prerequisites: the sized servers in fc-prod-fra-core from Unit 4.1, the SSH key in the SSH Key Manager, and the cloud-config document reviewed by the platform team.

Build goal: Attach and configure storage; supply cloud-init user data.

Steps (in the Data Center Designer):

  1. Open fc-prod-fra-core. Drag a Storage element from the Palette onto fc-app-01. The server expands with a storage section; select it to open its options in the Inspector pane.
  2. Enter a Name unique within the VDC, such as fc-app-01-boot. Set Availability Zone to Zone 1, the server's zone. This cannot be changed later.
  3. Set Size in GB, and in Performance select Balanced. The class cannot be changed after provisioning.
  4. In Image, select IONOS CLOUD Images and choose the Linux image. Record its full name, including the version.
  5. Set the root password, and select the SSH key stored in the SSH Key Manager.
  6. Select No configuration next to Cloud-Init user data. In the Cloud-Init User Data window, paste the cloud-config document.
  7. Select Boot from Device and choose VirtIO as the Bus Type.
  8. Repeat steps 1 to 7 for fc-app-02 in Zone 2, and for the boot volumes of fc-data-01 and fc-data-02.
  9. Drag a second Storage element onto fc-data-01. Name it fc-data-01-db, Zone 1, class Performance, at least 100 GB, and select no image, which creates an empty volume. Do not mark it as the boot device. Repeat for fc-data-02 in Zone 2.
  10. Select PROVISION CHANGES, confirm with your password and provision. Wait for bootstrapping, then check /var/log/cloud-init-output.log on one server before you register it in the ALB target group (Unit 3.3).

Common mistakes:

  • Planning to add cloud-init to a volume after provisioning. It is set only at creation; a change means a new volume.
  • Choosing Auto as the zone for volumes of a redundant pair, then being unable to prove their placement.
  • Sizing a database volume at its current data size on Balanced or Performance. Below 100 GB, you pay for the class without its performance.
  • Expecting a Frankfurt image or snapshot to appear in the Berlin VDC. Upload or recreate it there.
  • Testing a server the moment the DCD reports success, before cloud-init has finished.

Summary

Volumes carry the decisions that are hardest to undo. The storage type and the zone are fixed at provisioning, size only grows, and cloud-init runs once, on a new volume. Images and snapshots are bound to their location, and public images are replaced as they are patched. Design the boot volume as rebuildable and the data volume as the thing you protect, and make the build pipeline the source of truth for images in every location you use.

Key Points:

  • A volume's storage type and Availability Zone cannot change after provisioning; its size can grow but never shrink.
  • Balanced and Performance volumes need at least 100 GB, and a minimum server size, to deliver their performance.
  • Private images are usable only in the region where they were uploaded; snapshots only in the location where they were created.
  • Public images are rebuilt with each patch, and the previous version leaves the catalog.
  • Cloud-init is Linux-only and can only be set at the initial provisioning of a volume.

Important Terminology:

  • Boot device: the one volume, CD-ROM or NIC a server starts from, with its bus type (VirtIO or IDE).
  • User data: the shell script or cloud-config document that cloud-init applies on first boot.