I suppose everybody is entitled to their own opinion. Mine is that a drive dropping out of an array every once in awhile for no good reason is not OK. I think we all realize that during a rebuild you are at a greater risk to lose the whole array if there is an error (or another drive dropout) during the process.
Unnecessary rebuilds don't seem like a good thing. Don't you wonder why this happens?
Maybe because manufacturers are *deliberately* programming RAID-related "bugs" (maybe masked under TLER-like names...) into the non-enterprise class drives, just to earn some more money? Most of the people would never buy enterprise drives just for the sake of it, if RAID functions would function correctly in mainstream consumer drives, for third the price... Enterprise drives seem not to be more error-free than standard drives, all they have is extra warranty. Anyway... 5 year WD Black warranty, for example, would surely be enough for anyone.
So the issue here is: money, as usual... By relabeling a drive, writing a good firmware on it (i.e. without RAID errors included) and extending its warranty + triple the price... they call these "Enterprise" drives...
Just look at some forums on the net of people flashing RE4 firmware to Green drives (on WD20EADS more specifically) and having no problems whatsoever in RAID configurations.
Then go ahead and check the Seagate 1.5 TB (7200.11) drives fiasco, for which they have changed so many times, so many firmwares, some of them supporting RAID, others not; some of them having performance issues (CC1G) under RAID, some not (CC1H and SD1A). Of course, for the CC1H firmware... it depends... some of them work perfectly, others not at all. And there it is a newer CC1J firmware on some drives, but Seagate support assures people that it is similar to CC1H... I'd prefer to believe this is not exactly accurate. Why make the same firmware under 2 different names ?
For example (somewhat off-topic), I had a Seagate with CC1G F/W. I could set SATA 1.5 Gbps with the jumper, as specified on its label. In RAID, I've seen only dropouts, totally worthless. Then, under an update to CC1H, the SATA 1.5 Gbps jumper did not work anymore, but the drive became usable in RAID without issues. On the other hand, a factory-released CC1H drive drops out of RAID constantly.
The original CC1H drive in an external Vantec Nexstar 3 eSATA enclosure works perfectly on SATA 3.0 Gbps. On the other hand, the CC1H-updated CC1G drive gives me CRC error rates as soon as I try to write to the drive (and SATA 1.5 Gbps can't even be activated anymore...) I've even read somewhere that it "might" be an issue with the Spread Spectrum (Disabled/Enabled)... which is clearly connected to the dropouts (see WD Green/TLER and Spread Spectrum connections on some forums).
I have never seen this specified anywhere before, but maybe Enabling/Disabling (don't know which one) Spread Spectrum would also solve the dropout issue on Hitachi drives ?? Anyone tried it ? It can be done via Hitachi FTool. Interestingly, my drives are neither Enabled, nor Disabled in FTool, which seems quite strange to me (Ftool can't re-read the last value set...?)
It may seem that tiny hardware changes are constantly made in the same drive model, by most manufacturers; meanwhile firmware updates are sometimes solving the problems only partially... while (re)creating others.
WD20EADS TLER-enabled drives, for example, are performing a so-called "deep sector scan", which is quite annoying (constant seek noise produced by these drives, they wear down quickly... interestingly: no head parking); on WD20EARS, on the other hand, one can't even enable TLER. It may seem that these deep sector scans are enabled in order to prevent head parking on WD20EADS... which is a rather crude solution of a shoddy firmware... as someone already pointed out here.
Just dropped in these lines to scratch some wounds