4 ms·

A corollary to this might be "a new part on the shelf is not truly 'better' than a 90% worn out one still in service." Having ready spares is great but I'm sur

A corollary to this might be "a new part on the shelf is not truly 'better' than a 90% worn out one still in service." Having ready spares is great but I'm sure the HN crowd knows how much crib-death there is on new, replacement bits of all kinds. Until it's actually been run-in for a while how do you really know it's going to work when you unbox it to replace a truly dead piece? The upshot of this is' RAID-6' type designs are the most reliable in the real world since when one fails at least you are leaning on other parts that have been run-in and are past the leading edge of the bathtub curve.

dylan604 5y ago
>Having ready spares is great but I'm sure the HN crowd knows how much crib-death there is on new, replacement bits of all kinds. >Until it's actually been run-in for a while how do you really know it's going to work when you unbox it to replace a truly dead piece? We had the unfortunate experience of installing 4 new 16 bay chasis with brand new drives (10+ years ago now). We designated one of the 16s as hot spare for each chasis, plus had 2 cold spares for each as well. 72 brand new drives in total. All from the same batch of drives from the manufacture. Set them all up on a Friday and configured all for RAID5 (pre-RAID6 availability). Plan was to let them build and have some burn-in time over the weekend for possible Monday availability. Monday provided us with multiple drive failures in each chasis. Drive manufacturer confirmed a batch batch from whichever plant, replaced all and delivered larger sizes for replacements. Luckily, they failed during burn-in rather than 1 week after deploying.