Backup software asks for a weird amount of trust. You set everything up while your files are perfectly fine, then months later you’re supposed to believe the whole system will still make sense when something actually breaks. I’ve tried enough backup apps that I don’t put much stock in one until I’ve restored files from it myself.

Some people are perfectly happy building their entire backup strategy around a command-line tool like rsync, but Duplicati won me over by giving me control without turning every backup into a Herculean task. I used it to encrypt and back up my Obsidian vault, then created a separate test specifically to see how well its versioning and restore process held up. That second part was more important to me than watching another progress bar reach 100%.

Setting up a backup is pleasantly methodical

Five screens and very little guesswork

I installed Duplicati on Windows, so I grabbed the x64 MSI and ran through the installer. Duplicati runs locally, but its interface opens in your browser, which can make it look like a cloud service at first glance. The GUI package actually includes a local web server and just uses the browser as its front end. Once you sign in, the sidebar keeps the important stuff close by with Home, Add backup, Restore, Destinations, and Settings.

I started with Add backup, chose New backup, and worked through the five setup stages covering General, Destination, Source Data, Schedule, and Options. For my first proper job, I named it Obsidian Vault and added a description so I wouldn’t have to remember what the backup was for months down the line. The General page also lets you choose between Backup and Sync modes, so I left it on Backup and enabled the built-in AES-256 encryption.

You create the encryption password right there, and Duplicati puts a warning underneath that deserves your attention. If you lose that password, you lose access to the encrypted backup with it, so keep it somewhere that won’t vanish along with the PC you’re trying to protect.

Destination comes next, and Duplicati immediately offered File system, Duplicati Storage, S3 Compatible, and SFTP, with more options tucked under Show all. I chose File system and pointed the job at a dedicated local folder for testing. There’s also a Test now button before you move on, and I’d absolutely use it. Discovering that a destination path or connection is wrong after you’ve configured everything else is just wasted effort.

My test destination was on the same physical drive as the source, which worked perfectly well for testing the workflow but would do nothing for me if that drive failed. For anything I actually cared about, I’d use another physical drive, a NAS, an off-site destination, or a mix of those depending on the backup. If you only need something simple on your home network, even a USB drive connected to your router can become basic network storage.

Duplicati lets you decide how much history you want

The useful controls are where you’d expect them

The Source Data screen uses a normal folder tree, so I selected my entire Obsidian Vault. That included the Markdown notes themselves along with the hidden configuration data that keeps the vault working the way I’ve set it up.

You can also add direct paths, local disks, and remote paths, while filters and exclusions let you skip anything you don’t want backed up. If you were using Duplicati for your broader Windows profile, this is where I’d be selective. Documents, photos, project folders, and anything you couldn’t easily replace belong in the backup. Temporary downloads, caches, and files you can recreate probably don’t.

Scheduling gets its own screen. The presets on my install included daily, weekdays, weekly, custom, and Don’t run automatically. I chose the manual option for testing because I wanted to control each run myself. For a long-term backup job, though, I’d schedule it. A good backup routine combines automatic backups with multiple copies on different storage, and Duplicati takes care of the automatic part once you give it a schedule. If the whole setup depends on you remembering to click Start, there’s a good chance you eventually won’t.

The final Options page handles retention. Mine defaulted to Keep all backups, which I left enabled during testing because I wanted every version available. Duplicati also supports Smart backup retention, which keeps more versions from the recent past and gradually thins them out as they get older.

That becomes important once a backup has been running for months. Duplicati uses incremental backups, compression, and block-level deduplication, so it avoids storing identical data repeatedly where it can. Storage use still grows as your files change, though, and keeping every historical version forever will eventually add up.

Once I submitted the job, Obsidian Vault appeared on the Home screen with its own Start button. While it ran, the card showed that Duplicati was processing files. A few seconds later, the dashboard reported one successful version. The dashboard also keeps the last successful backup time, duration, source size, destination size, and version count in one place. Those are the numbers I’d check first to make sure a backup job actually ran the way I expected.

The restore process gets the final vote

I wanted an old file back, not another green tick

For recovery testing, I created a separate job called Article restore test. I backed up a small document, edited it, then ran the backup again. After that, the Home screen showed two versions.

From Restore, Duplicati listed both of my configured jobs and also gave me Direct restore from backup files and Restore from configuration. Direct restore is especially useful if the original machine is gone because Duplicati can work from the backup files themselves instead of depending on the original local database.

I selected the test job and chose the earlier version from the timestamp menu. The next screen showed the backup as a normal folder tree, so I could recover a single file instead of restoring everything.

The final step let me choose between Original location and Pick location. I chose a separate folder, which is how I’d recommend testing restores. You can inspect the recovered copy first without touching the live file. Duplicati also asked what to do if a file with the same name already existed. I could overwrite it or save the restored copy with a timestamp added to the filename. I picked the timestamp option and started the restore.

A few seconds later, Duplicati reported Restore completed. That’s the part I care about most. Encryption options, storage providers, and retention settings are useful, but backup software only earns my trust once I can actually recover the file I thought I’d lost.

Duplicati 2.4 also includes disk and partition backup support, and the Source Data page exposes an Add local disk option. I tried it, but no selectable disk appeared on my machine, so I couldn’t test full-disk imaging. I wouldn’t rely on that part of Duplicati until I’d rehearsed the full recovery process on spare hardware anyway.

I’d trust it because I know how to get out

Most of the software on my PC is replaceable. Years of notes, photos, unfinished drafts, and project files are another story, and those are what I’d actually hate to lose.

Duplicati gives me enough control over how those files are protected without making the restore process hard to follow. More importantly, I’ve already tested the part I’d actually need on a bad day: choosing an older version, restoring it somewhere safe, and opening the recovered file again.

Before trusting it with everything, I’d run a larger recovery test and keep both the encryption password and configuration somewhere separate from the machine I’m backing up. And your second backup copy doesn’t have to live on another external drive; cloud storage, a NAS, or even another computer can give that copy somewhere else to live. Once that’s done, I’d be comfortable letting the backups run on their own.