Saturday 14th July 2018: 11.42pm. Link shared: https://www.backblaze.com/blog/hard-drive-stats-for-q1-2018/
Decided to fire the starting pistol on the next round of hard drive upgrades for my ZFS array, I've got 800Gb still free which won't be filled for over a year yet, but it takes me at least three months to buy drives as I need to stagger their purchase over a long time to ensure I never buy two drives from the same production batch, plus there is a non-trivial soak testing period after arrival to ensure the postman didn't drop them too hard. Plus I can see that the 8Gb Seagate ST8000DM004, which is very reliable according to https://www.backblaze.com/blog/hard-drive-stats-for-q1-2018/, has taken a recent price tumble under 200 euro which suggests it is finally time to buy.
As an update to 1520138851.html, my current ZFS array looks as follows:
1. Mirrored 3Tb WD30EFRX 38k hours + 28k hours
2. Mirrored 3Tb WD30EFRX 28k hours + 32k hours
3. Mirrored 4Tb ST4000DM000 3k hours + 25k hours
As much as those 3Tb WD Red's failed in droves before, their warranty issued replacements appear to be much better. 38k hours is 4.3 years of continuous operation. Those WD Red's failed before at around 17k hours, so we're heading past double that now. And I still have a spare WD Red with 40k hours on the clock, but had shown no signs of old age when I replaced it.
Now I appreciate a mere 10Tb array could fit onto a single mirrored pair of 12Tb drives with room to spare, but that would be well over a grand to buy, and as much as it would save me 20w of electricity or so which is €45/year in electricity, the math says that I'm better replacing one of the 3Tb pairs with 8Tb pairs instead, thus adding 5Tb. It'll take eight years (yes, eight!) to fill 5Tb at our current rate, during which I'd expect those 3Tb drives to fail, in which case we'll have plenty of ready to go spares. And if one of the 4Tb pairs go, I can always fake a 6Tb drive spare by joining two 3Tb spares.
So, in short, I don't expect to have to think about that ZFS array again after this upgrade for many years to come. If anything, in fact, I really ought to start considering mirror triples, as they vastly reduce the risk of drive failure ever being a problem, plus they eliminate the need for keeping spares at the cost of extra electricity. It's a tough call though, no drive failures since 2015 makes triple mirrors seem a bit overkill for my needs. I guess, though, everybody thinks that until they experience a dual mirror pair failure, and then they're hosed.