Backing up virtual servers has one peculiar characteristic: until you need it, functional backups look exactly the same as non-functional ones. In both cases, the overview shows green lights and in both cases files are accumulating. The difference only becomes apparent when someone needs to restore something from them, and that is the worst possible moment to discover it cannot be done.

Three most common ways to break down your advance payments

The first is a backup stored where the original also runs. If the storage fails or ransomware infiltrates the environment, both disappear at once. A backup on the same drive as production disks protects only against a deleted file: not against the event for which backups are established.

The second is backing up a running machine without coordinating with what is running inside. For a database, this means capturing the state midway through an operation: the backup is created and appears fine, but the database cannot be restored from it. Therefore, before taking a snapshot, a command is sent to the machine to briefly complete any pending work; without this, it is merely a copy, not a backup.

The third is a silent end. The task has changed, the disk is full, the password has expired, and backups simply stop being created. Without a warning, you won't notice this, because non-functional backup makes no noise. That is precisely why it is also monitored that the backup was established, not just that the last known one exists.

The 3-2-1 Rule and Why It Still Holds True

Three copies of data on two different types of storage, one off-site. This rule is decades old and has survived every trend because it addresses scenarios rather than technology: human error, hardware failure, and events affecting the entire location, fire, water damage, theft, or ransomware attacks.

The third copy kept off-site is also the one most often missing. Not due to ignorance, but because it is the only part requiring regular attention or a monthly fee. Companies that lost their data permanently almost always had backups, just all in one place.

What Makes Proxmox Backup Server Different

The backup server for Proxmox works with data in blocks and stores each unique block only once. In practice, this means that dozens of similar servers: typically the same operating system but different applications, take up a fraction of the space full copies would require. Daily backups can therefore be retained for months without the storage growing to absurd proportions.

The second aspect is verification. The server can regularly scan stored data and check that checksums match, ensuring that what resides on the disk remains exactly what was recorded. Otherwise, silent data corruption in long-term backups is an issue one only discovers during restoration.

The third is synchronisation to a second server elsewhere. Backups are made locally where it is fast, and from there transferred off-site. This creates the third copy of the 3-2-1 rule without anyone carrying anything home on a disk.

Recovery test: the only part that confirms everything.

Backups can be scheduled for the afternoon. What makes it a backup is the regular test that data can be restored from it, ideally into an isolated environment where the restored machine cannot cause harm if it comes online with the same IP address as the original.

The test answers two questions that otherwise can only be estimated. How much data will be lost in the worst case, i.e., how old the last usable backups are. And how long it will take for the company to resume operations. The second figure is often a surprise: restoring the file itself takes minutes, but getting the entire environment running properly, including dependencies between servers, takes significantly longer than anyone thinks until they try it once.

Summary

A backup that no one has ever attempted to restore is merely a collection of files we hope will work. Keep three copies on two types of storage with one off-site, monitor that the backup actually occurred, and at least occasionally test a real restoration, including measuring how long it took. How we implement backups in virtual environments is described on the netlabs.cz website. virtualisation and Proxmox, with a connection to supervision then in the article about network monitoring.