article --summary
The summary
This podcast conversation starts with a familiar complaint: backups get set up once and forgotten until a crisis, and almost nobody tests whether a restore actually works. Gilles Chehade, who has been doing backups for about 30 years, explains how the scale of data and the shift to virtualized infrastructure have made the old tools and habits fall short.
The discussion separates two reasons to restore, disaster recovery and undoing damage caused by people, and argues that the second is now more likely because of ransomware and AI agents that can delete data from inside a system. Gilles makes the case for treating backups as a reconciliation loop, with an inventory drawn from the same source that creates infrastructure, policies for frequency and retention, and automated checks that fail a build when reality drifts from expectations.
The last part covers Plakar, an open source tool with end-to-end encryption and deduplication, and a closed-source control plane aimed at larger organizations. Gilles's closing advice is to keep backups regardless of which tool you choose.
Editorial note: This is a summary of an article first published by Arrested DevOps. The full piece is on their site.