wblock, one of my past jobs was qualifying hard drives for a vendor of NAS systems. Every single component had to pass my qualification tests before it went into our NAS systems. I believe we were the first vendor to use "enterprise grade" SATA drives in a NAS system. One of the things that my initial qualification test for hard drives did was read every single sector of the drive. What I found was that approximately 40% of the hard drives that came into our shop from both of the major vendors we worked with -- Seagate and Western Digital -- had at least one bad sector on the initial read right out of the box from the vendor and failed the initial qualification test that I devised. Once we ran a disk scrub on the drives (i.e., wrote zeros to every sector) the sectors were replaced from the relocation list and all was well, they passed further tests without any issues. So the qualification test was modified to do the final read test after the scrub and checked the SMART relocation list counts and rejected only those drives which had a number of relocations which exceeded our strict standards, rather than requiring a zero error count fresh from the manufacturer. (Note that this process occurred for roughly three days straight -- yes, three days straight of reading and writing patterns to the hard drives before they were allowed to be shipped).
In case you're wondering, we had 1 (one) system arrive DOA at a customer site during the entire time I was in charge of manufacturing quality at that manufacturer, and we don't know what happened out there at the customer site because when it arrived back, it booted up immediately when we plugged it in. That's how persnickity I am about quality -- out of the first two hundred systems we shipped, not a single "real" DOA.
In any event, SMART implementations vary wildly between hard drive vendors, so it's always iffy reading too much into what they report unless you have extensive experience with that specific vendor (as in, you've qualified hundreds of drives of that type in the past). If read errors happen *WHEN READING DATA THAT HAS ALREADY BEEN WRITTEN* then I view the hard drive as pretty much toast, if it's in service in a storage array the RAID attempts to re-build that stripe in hope that the bad block will get relocated, logs it, and if a certain number of these errors happen we eventually fail the drive. But if read errors happen reading sectors that have never been written, my experience is that this is completely and totally normal for SATA hard drives, "enterprise" or not, and that the correct solution to avoid this happening if it is an issue for a specific application is to scrub the drives prior to putting them into service.
