CSE Roadmap

Common tier · about 2 hours · 5 parts

Part 4 of 5 · 2 min read

Disk failure

The storage media itself is damaged, so even the data that was supposedly durable is now lost or corrupted.

Disk failure

Think of an actual fire destroying the store's entire filing cabinet, not just interrupting today's business, meaning every previously filed receipt is now gone too, no matter how carefully it had been filed.

In plain words: a disk failure happens when the storage media itself is damaged or destroyed, meaning the data that was supposedly durable, written to disk and therefore assumed safe under the durability guarantee covered in Level 7, is now lost or corrupted anyway. This is fundamentally more severe than a system crash, since the write-ahead log that a crash recovery relies on may itself be unreadable or gone; recovering from a disk failure instead requires restoring from a separate backup, covered later in this level, stored on different, undamaged media entirely.

A worked example. A failed disk drive can make both the data files and the write-ahead log stored on it permanently unreadable, meaning the only path to recovery is restoring from a backup kept on a physically separate storage device.

The key lines. No new SQL; disk failure recovery depends entirely on backups and, where available, a remote standby copy, both covered later in this level.

Disk failure

A disk failure is damage to the storage media itself, making data that was supposedly durable actually unrecoverable from that media, requiring restoration from a backup on separate, undamaged storage instead.

Example: a fire destroying an entire filing cabinet, losing every previously filed receipt regardless of how carefully it was filed, mirrors a disk failure.

Remember this

A transaction failure is contained to one transaction; a system crash loses memory but keeps disk intact, recoverable from the log alone; a disk failure damages the disk itself, requiring a separate backup. Matching the recovery mechanism to the right failure type is exactly what the rest of this level builds out.

Quick notes

  • A transaction failure affects only one transaction; the rest of the system keeps running normally.
  • A system crash loses everything in memory, but disk-resident data, including the log, survives intact.
  • A disk failure damages the storage media itself, making even previously durable data unreadable.
  • Crash recovery relies on the durable write-ahead log; disk-failure recovery relies on a separate backup.
  • Each failure type demands a fundamentally different recovery mechanism, covered across the rest of this level.