The Unseen Anchor: A Critique of the 'Set and Forget' Backup
There is a quiet, pervasive belief in our circles, a piece of received wisdom so comforting we rarely question it. It’s the idea of the ‘set and forget’ backup. We configure our cron jobs, our rsync scripts, our cloud sync tools, and we walk away. A green checkmark in a dashboard, a log entry stating success—these become our anchors, the proof that should disaster strike, we are prepared. We trust the machine to remember what we have so conveniently forgotten. This trust, I’ve come to believe, is a dangerous form of self-deception.
The fallacy isn't in the automation itself; automation is the bedrock of reliability. The fallacy is in the ‘forget’ part. We treat the backup system not as a living, breathing component of our infrastructure, but as a black box of salvation. We assume that because data is being moved from point A to point B, it is therefore good, restorable, and complete. We conflate process with outcome. The system is running, therefore we are safe.
But a backup is only as good as its last successful restoration. The ‘set and forget’ mentality divorces us from this ultimate truth. It creates a comforting illusion of safety that can shatter at the worst possible moment. I’ve seen it happen: the backups ran without error for years, diligently writing data to a mounted volume. Only when the primary storage failed did anyone discover the mount had quietly failed six months prior. The logs were a chronicle of success, writing to a void.
The Ritual of Remembering
True reliability isn’t born from automation alone, but from the ritual of engagement with that automation. It is the deliberate, scheduled act of ‘remembering’ what we’ve ostensibly ‘forgotten’. This means periodically, and with a solemn sense of duty, attempting to restore a file—any file—from that backup. It means validating the checksum of a restored tarball against the original. It means firing up a minimal instance from a system image to ensure it actually boots.
This is not exciting work. It is the antithesis of the glamorous new feature launch. It is the meticulous, boring work of a archivist checking the humidity levels in a vault. There is no applause for a successful validation test, only the quiet satisfaction of a promise kept.
The ‘set and forget’ backup is a myth we tell ourselves to ease the cognitive load. But in our world, the things we forget are the things that eventually fail. Our systems require not our constant attention, but our periodic, thoughtful witness. The anchor isn’t the script or the cloud service; the anchor is us, choosing to look, to verify, to remember that data has no value until it can be brought back from the silence we’ve sent it to.
Notes & further reading
A few pages I came back to while writing this: