When SMART Lies: A Firsthand Look at Unrecoverable Drive Failures
Late last night, ordinary laptop operations started failing in odd ways: logins stalled, system calls dragged, and the network grew unreliable. By morning, the signs pointed to the hard drive—and despite SMART reporting the disk as healthy, read tests failed almost immediately. The drive had died.
The good news: most of the home directory was backed up to a work machine before the drive gave out (a process that surfaced NTFS's restrictions on common filenames along the way). After swapping in a fresh drive and installing a clean Ubuntu image, most files were restored using a customized tar variant that tolerated the corruption in the backup archive. Still, a meaningful chunk of data did not survive.
A Practical Gap in Archive Recovery
That loss raises a straightforward engineering question: why doesn't tar support an option to skip past damaged regions of a partially readable archive? A handful of bad sectors near the start of an archive does not necessarily condemn the remaining gigabytes of data—yet standard tooling offers no way to salvage them.
Lessons from a Well-Timed Failure
After several busy hours, the system is back to normal: all critical work files are recovered, and the asset-tracking application is being polished ahead of a Friday deployment. If anything, the timing was fortunate—losing data in the final hours before a trip would have been far worse than losing a morning to it now.
