Basically, I had this drive in my windows build and it had ~ 7 pending sectors, I got my unraid server set up and I migrated it over to my unraid server (After I copied all the data off it). I half completed a preclear but I had to restart my server (So I lost all logs of that preclear) and then I started another one.

 

Anyway, I checked the smart report (Preclear isn't done yet, on it's final read) and the reallocated sectors is now 0 and so is the current pending sectors. Can I ask what happened in this case?

The controller tests the sectors again before reallocating them. In this test they passed and it marked them as good.

The controller tests the sectors again before reallocating them. In this test they passed and it marked them as good.

 

Fair enough. Can I ask in what situations a sector would first be invalid then later be valid?

Could be a marginal sector or could have even been when reading that sector there was a bump or vibration that caused it to misread and mark it for reallocation.

 

It happened to me twice on another disk and was bugging me so I researched it quite a bit and from what I found there is really not definite way to tell what caused it.

Another more common cause for this is when a sector is corrupted by a power spike, such as a sudden power outage during a write (sectors under the writing head may be corrupted), or just a huge electrical spike through the head.  Sectors carry a pretty good ECC section (Error Correction data), so that they can fix a certain percentage of corruption, but if there is too much, the UNC flag (UNCorrectable) is raised and the sector becomes a Current_Pending_Sector, one that needs additional testing.  Once tested, it quickly determines that the magnetic media underneath the sector is fine, so then writes good data (with updated ECC info) to the sector.  Therefore no need to remap the sector, and it's no longer considered bad, as a 'Current Pending sector'.  I like to call these soft errors, as opposed to hard errors where the magnetic media has become too weak to save data accurately and 'readably'.  Hard errors have to be remapped to a spare good sector, with good media underneath it.