Users report intermittent I/O errors when writing to /dev/sdb on a database server. The filesystem occasionally remounts read-only, and journalctl shows repeated 'medium error' and 'unrecovered read error' messages referencing sdb. You suspect the physical drive is failing. Which command gives you the most direct insight into the drive's internal health and reallocated sector counts to confirm hardware failure?