According to the MTBF ratings neither should have failed, but both did.
Whilst not having any argument with deciding to put slightly more expensive drives in your array, especially with only single drive redundancy, you don't seem to understand what a MTBF rating means.
It is a
*mean* time between failures. It cannot tell you ANYTHING AT ALL about when any individual drive "should" fail! All it tells you is a statistical measure, given a LARGE number of drives, which will fail at all sorts of different times, what is the mean of those failure times.
A drive with a higher MTBF rating may be
less likely to fail per unit time than a drive with a lower rating, but either drive can absolutely fail at any time, and because the MTBF doesn't tell you anything about the distribution of failures with time, you can't conclude anything at all about when one can expect a drive failure.
A 5 year MTBF failure rate absolutely does NOT mean that you expect the drive to go fine for five years. It means that if you run hundreds and hundreds of drives to failure, the mean time at which they failed was five years. But the first drive might have packed up inside five minutes... and it tells you very little about the pattern of failure, either. For example it is sometimes claimed that drives tend to fail right at the start, or run for years... this could still give you a five year MTBF if half the drives fail in the first few weeks, but the rest all run for 10-15 years before failing. And one wonders how a manufacturer bringing out a new drive model can possibly have accurate ratings even for that, since by definition none of their drives will have been run for 15 years.
It may well make sense to buy drives with a higher MTBF rating to put in your arrays, but one should absolutely plan that a drive could fail at any time. Two drives are significantly less likely to fail at the same time if drive failure is stochastic, although if the failure is trigger by some correlated cause like a power spike on startup or dropping the enclosure, that can make multiple simultaneous failures a lot more likely.
The only robust solution is proper, full, regular backups, and offline backups of the backups offsite as well.
Cheers, Hywel.
