{
  "slug": "fail-partial-failure-model",
  "term": "Fail-partial failure model",
  "aliases": [
    "Fail-partial model"
  ],
  "definition": "A fail-partial failure model includes inaccessible or silently corrupted blocks as well as complete device failure.",
  "note": "IRON distinguishes whole-device failure, inaccessible blocks and corrupted block contents; failures can be transient or persistent and may have spatial locality. A drive that still answers requests is not necessarily returning correct data. Detection and recovery are separate policies. [iron05]\n\nThis is a fault model, not an estimated failure rate or a claim that all faults are independent. Intra-disk redundancy cannot protect against losing the whole drive. [iron05]",
  "status": "research",
  "advantages": [],
  "limitations": [],
  "related": [
    "fault-model",
    "fault-injection",
    "latent-sector-error",
    "silent-data-corruption",
    "lost-write",
    "misdirected-write",
    "intra-disk-redundancy"
  ],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "citations": [
    {
      "id": "iron05",
      "title": "Prabhakaran et al. IRON File Systems. SOSP 2005, sections 2.3 and 3.",
      "url": "https://research.cs.wisc.edu/adsl/Publications/iron-sosp05.pdf",
      "accessed": "2026-11-11"
    },
    {
      "id": "hesela-integrity",
      "title": "Hesela: Filesystem integrity, detection and recovery boundaries",
      "url": "https://hesela.com/analyses/filesystem-checksum-error-detection/#fail-partial-evidence",
      "accessed": "2026-11-11"
    }
  ],
  "updated": "2026-11-11"
}