I have had four 500GB WD drives that have failed on me about the same time after a couple of years of operation. They click-o'-died on me by the end of 2009/beginning of 2010.
After that I built a raidz2 storage pool (OSOL/OI) consisting of 8 1.5TB Samsung EcoGreen (HD154UI) drives on an LSI1608 SAS HBA. It has been in operation since January 2010 and during that time
two Samsung drives have failed on me and one showed "odd" behavior during resilvering of the pool. This odd behavior means that some (correctable) errors occurred during the resilvering process but has showed no malfunction during normal operation. The first faulty Samsung drive was replaced with a 1.5TB WD Caviar Green (WD15EARS) drive and it also showed some odd behavior during the resilvering process after I replaced the second corrupted Samsung drive.
The system drive which is a 1TB Samsung Spinpoint F3 (HD103SJ) has corrupted sectors which has seriously affected one of my virtual machines on the system. It needs to be RMAed but I'm not willing to go through the hassle of reinstalling the entire system. The motherboard is really fishy too and probably need to be RMA'd as well....
So far so good, two drives RMA'd, one in need to be RMA'd and an additional two drives that are destined to fail eventually, that is my experience with my current setup. Otherwise the RAIDZ2 pool is still going strong since the past 6 months and there are no signs of instabilities in the hard drives. The best part of it all is that even though some drives have failed, I have suffered no data loss on the RAIDZ2 pool. The system drive is a different story though, but I don't have any vital data on it anyway...
After this experience I'm actually considering buying some Hitachi DeskStars for my next storage pool. I've heard a lot of good things about them and the large ones (2TB+) don't have that weird 4K sector setup that the WD drives have which causes some compatibility issues on Linux/Unix systems. They come with 3 years of warranty. If I have the money I might consider buying some UltraStars. It's not that they would be considerably more reliable than the DeskStars but they come with 5 years of warranty and maybe that might be worth the extra price premium...
If you consider building a RAID pool it is recommended to have
at least two drive redundancy simply because when a drive fails and you replace it with a new one, the pool needs to be rebuilt (or "resilvered" as it is called in the Solaris/ZFS world, ya know mirrors... smoke'n'mirrors
![;) ;)]()
). The rebuild process may take several days and put
severe strain on the other drives as data is copied onto the new drive. This strain may cause an
additional drive to fail during the rebuild process. If that happens and you don't have an additional redundant drive that can take care of the problem than you are literally smoked. This is an admin's nightmare and there is a whole hate website against RAID5 because of this (I think it is on baarf.org you can google it. It means "Battle Against Any Raid Five", and the website has research papers on why RAID5 is bad ...).