24-hour RPO generates 365 annual snapshots requiring explicit rotation to bound storage growth

The Filipovski post recounts the incident and derives three rules: maintain a copy elsewhere, avoid live mirrors such as RAID 1, and retain time-stamped snapshots. It defines RPO as the maximum acceptable data loss window, citing <30 seconds for financial systems versus 24 hours for small enterprises. Without pruning, 365 daily snapshots accumulate, each consuming incremental storage until rotation or deletion intervenes.

Bit-rot studies document annual uncorrectable error rates of 0.5–2 % on consumer HDDs and NAND charge leakage on SSDs after 12–18 months of unpowered storage. The 3-2-1 rule, codified in NIST SP 800-53 and implemented in BorgBackup and restic, mandates three copies on two media types with one offsite, directly addressing both the single-point failure and the ransomware vector the post identifies.

Operationally this means scheduled snapshot jobs must embed retention policies that delete objects older than the chosen window while verifying checksums on each cycle. Failure to automate verification converts the backup set into an undetected archive of corrupted data, negating the RPO guarantee.

Next steps include adoption of tools that combine deduplication with automatic integrity checks, reducing the manual oversight burden the post flags as unsustainable for non-technical users.