Unit 5.7: Data Protection and Lifecycle
Introduction
Units 5.1 to 5.6 each built a store with its own protection: volumes, shares, buckets, three database engines and a stream. A risk office does not audit stores; it audits one question: for each dataset, what is the newest copy that survives a given failure, where is it, and who can delete it? This design unit answers that question by assigning each dataset to the mechanism built for it.
Three mechanisms are easily confused. Backup Service protects machines through an agent. Snapshots copy a volume. The DBaaS engines back themselves up. Each covers what the others do not, and the plan only works when every dataset has an owner.
1. The Mechanisms and Their Boundaries
1.1 Backup Service: Agent-Based Protection of Machines
Backup Service is built with Acronis. It protects a server through the Acronis backup agent: installed automatically when you activate backup on a bootable volume of a new server (IONOS CLOUD public images only), or manually at any time, which is required for private images. The agent can also protect resources outside IONOS CLOUD, such as on-premises servers, workstations and virtual machines in other environments. The service also includes backup support for software you run yourself on a server, including Microsoft SQL, MariaDB, MySQL, Oracle DB and SAP HANA.
Backups land in a Backup Unit, created in the DCD's Backup Unit Manager. Several units separate backup sources and their access, and only contract owners, administrators and users with the Create Backup Units privilege can use the Backup Service Console. Backup data is stored in the default Backup Storage, or in Object Storage or a Network File Storage share defined as a storage target. The agent needs an internet connection to the backup system.
Three behaviours belong in the runbook. If you encrypt a backup plan with a password, the agent encrypts locally, and a lost password makes the backups permanently irretrievable. Deleting a server, its storage or the whole VDC does not delete its backups. Deleting a backup unit deletes them. Backup Service is also outside the scope of the IT-Grundschutz certificate BSI-IGZ-0730-2025 (Unit 1.4), so FinCorp's compliance map records it separately.
1.2 Snapshots: Volume Copies, Not Backups
A snapshot copies a provisioned Block Storage volume. It covers the entire capacity, including unwritten space, is not incremental, is limited to the data center location where it was created, and has no retention period, so it is never deleted automatically. Data still in a VM's RAM is included only if you flush it with sync or shut down the guest first.
Those properties define its role: a rollback point before a change, or a template for new volumes in the same location. They also define its limits. A snapshot of a running database server holds what was on disk at that moment, not a transaction-consistent database. And a snapshot that never expires keeps costing until someone deletes it. Snapshots stay in the location of the volume; for a copy elsewhere, use Backup Service or Object Storage.
1.3 Database Continuity: The Engines' Own Backups
The managed databases bring their own backups, so database continuity runs on the engine's backup and point-in-time recovery (PITR), plus dumps:
| Engine | Built-in backup | Recovery | Retention |
|---|---|---|---|
| DBaaS for PostgreSQL | Daily base backup plus WAL, uploaded every 5 minutes | PITR to a target time | 1 to 365 days |
| DBaaS for MariaDB | Periodic full backups plus binary logs | PITR to a target time | 1 to 365 days |
| MongoDB | Daily base and weekly full snapshots | Snapshot restore; PITR up to 24 hours on Enterprise | 7 days |
| DBaaS for In-Memory DB | Snapshots at chosen hours | Nearest preceding snapshot | 1 to 365 days |
As shown above, MongoDB is the outlier at seven days, and destroying a MongoDB, PostgreSQL or MariaDB cluster deletes its backups. Dumps (pg_dump, mariadb-dump, mongodump) close the gaps: retention beyond the engine's window, survival of a deleted cluster, and a copy outside the provider's backup chain.
1.4 Object Storage: The Archive and the Immutable Copy
The fourth layer is an Object Storage bucket with Object Lock in Compliance mode (Unit 5.2). Nobody can shorten its retention period, neither the bucket owner nor IONOS CLOUD. Where the requirement is that a copy cannot be deleted, even by a compromised administrator account, this is the layer that meets it: dumps, exported activity logs and archived events go here, and lifecycle rules remove them once retention ends.
2. The FinCorp Continuity Plan
FinCorp maps each dataset to its mechanism and its failure:
| Dataset | Primary protection | Second copy | Immutable copy |
|---|---|---|---|
Portal servers fc-app-* |
None: rebuilt from image and cloud-init (Unit 4.2) | Images uploaded per location | Not needed |
Payment switch servers fc-data-* |
Backup Service agent, own Backup Unit | Backup Storage | Nightly database export to the locked bucket |
fc-pg-portal (PostgreSQL) |
PITR, off-site backup location | New cluster from backup in recovery (Unit 5.3) | Nightly pg_dump to the locked bucket |
fc-mongo-cases (MongoDB) |
Snapshots and PITR | Daily mongodump |
Dump in the locked bucket |
fc-imdb-portal (In-Memory DB) |
AOF persistence, snapshots | Refilled from PostgreSQL for cache keys | Not needed |
| Kafka topics | Replication factor 3 | Archive objects in Object Storage (Unit 5.6) | Archive bucket under Object Lock |
NFS share portal-assets |
Source in the build pipeline | Object Storage copy for Berlin (Unit 5.1) | Not needed |
Two rules make the plan testable. Every row names a restore drill with a measured recovery time. And no row relies on a snapshot as its only copy.
Decision Summary
| Question | Use | Not |
|---|---|---|
| Protect a VM's files and applications, including self-managed databases | Backup Service with the agent | Snapshots as the only copy |
| Roll back a volume before a risky change, in the same location | Snapshot, deleted after the change | Long-lived snapshots as backups |
| Recover a managed database to a point in time | The engine's PITR, off-site backup location | Volume snapshots |
| Keep database copies beyond the engine's window or a deleted cluster | Scheduled dumps to Object Storage | Engine backups alone |
| Make a copy that no one can delete before its retention ends | Object Lock in Compliance mode | Access controls alone |
| Satisfy an IT-Grundschutz scope requirement | Services in the certificate's scope; record Backup Service separately | A provider-level claim |
Summary
Data protection on IONOS CLOUD is a composition. Backup Service protects machines through an agent, snapshots give location-bound rollback points, each DBaaS engine provides its own backups and PITR, dumps extend what the engines keep, and Object Lock in Compliance mode provides the copy no one can remove. Assign every dataset to these layers explicitly, drill every restore, and keep the compliance scope of each layer on record.
Key Points:
- Backup Service works through the Acronis agent; deleting a backup unit deletes its backups, deleting a server does not.
- Snapshots are full, not incremental, location-bound, never expire on their own, and are not transaction-consistent.
- Database continuity is the engine's PITR plus dumps; MongoDB keeps snapshots for seven days only.
- Object Lock in Compliance mode is the immutable layer of the plan.
- Backup Service is outside the IT-Grundschutz scope; record it in the compliance map.