Thufir:
I'm getting SMART errors, but I'm not sure how much credence to give them. It seems to be the same two e-mails over and over.
The following warning/error was logged by the smartd daemon:
Device: /dev/hdb, 2 Currently unreadable (pending) sectors
Bruno Wolff III:
It might be that the drive firmware has a bug where as the pending sector count doesn't always get cleared when the sectors are reallocated. I have a Maxtor drive that has that problem.
It might also be that you have never written of the bad sectors so the drive can't reallocate them. If they really can't be read a long scan should be showing up a bad sector that you can then find the file it is contained in (so you know what you are losing) and then rewrite that sector (actually you want to rewrite the whole 8 sector block to keep the OS from trying to read the surrounding sectors). If the sector is permanently bad the drive should reallocate it. Sometimes just an isolated write was bad and the sector doesn't need to be remapped.
I had a system using LVM fail to boot, and when assessed using another drive (that would boot), got error reports like the above. Wherever the errors were, it was some place that LVM really did not like. I used dd to overwrite the entire drive, to try and force a write to wherever it was, and force the drive to try and fix what it could, and the errors got cleared up.
Nothing else I had tried cleared up the errors. The system got reinstalled, without LVM, just to see if the drive would keep on working, and it has. It's many months later, and there's no error reports, including while making deliberate checks.