The Myth of the Perfect Snapshot: In Defense of the Partial Backup
We are taught, from our earliest days tinkering with systems, that a backup is a sacred, holistic thing. It is a perfect, frozen moment. A complete capture of state. The doctrine is clear: anything less than a full-system image is a gamble, a half-measure destined to fail you at the precise moment of maximum need. We chase this totality with religious fervor, investing in ever-larger storage arrays and more complex snapshotting systems, all in service of this uncompromising ideal.
I want to propose a heretical thought: the perfect, complete backup is often a myth, and the relentless pursuit of it can be a dangerous distraction. Worse, it can become a single point of failure for our own operational discipline. The pressure to maintain these monolithic archives can be so immense that it leads to paralysis—if we can’t back up the entire mammoth, we back up nothing at all. The perfect becomes the enemy of the good, the complete, and the actually useful.
Consider the humble, partial backup. Not a system image, but a targeted salvage operation. The curated collection of configuration files from `/etc`, the precious application state from a single database, the unique source code from a home directory. These are not second-class citizens in the archive. They are often the *only* things you truly need to recover from a catastrophe. The operating system itself is, frankly, disposable; it can be reinstalled. The data and the configuration that make it *your* system are irreplaceable.
By fetishizing the complete snapshot, we blind ourselves to the elegant simplicity of the partial. A script that diligently tarballs a dozen critical directories every night is more valuable than a grandiose, petabyte-spanning backup system that is too complex to test and too fragile to trust. The partial backup is comprehensible. Its scope is known. Its restoration can be practiced and perfected. It is a tool you understand intimately, rather than a black-box miracle you hope will work.
This isn’t an argument for negligence. It is an argument for pragmatic, focused resilience. It is the understanding that recovery is not a binary switch from ‘total loss’ to ‘perfect restoration,’ but a spectrum. The goal is to get back to a functioning state as quickly as possible. Often, a few critical files are all that’s required to cross that threshold. So, by all means, strive for comprehensive coverage if you can. But if you cannot, do not let the impossibility of perfection prevent you from achieving something profoundly good. Embrace the intentional, documented, and reliable partial backup. It might just be the most complete solution you ever create.
Notes & further reading
A few pages I came back to while writing this: