[mdadm]

Originally Posted by Rogean [You must be logged in to view images. Log in or Register.]

Well I just can't catch a fucking break...

[You must be logged in to view images. Log in or Register.].

Red's Array.

Get comfortable guys.

Were the red server drives purchased at the same time as the blue drives?

Originally Posted by mdadm

I've been watching this thread for some time but I've not seen anyone point this out, so I'm taking the opportunity. If this ain't news to you, then, excuse the noise.

Drives are usually rated with an MTTF (mean time to failure), which gives an estimate of how many hours the drives will last before they fail. Usually, the variance associated with an MTTF is not very large. So now consider if you buy 6 drives, from the exact same batch (where the real times to failure are much likely to be closer together than the variance the manufacturer might claim with their MTTF), and put them in the same server at the same time. Now these drives each have the same number of hours on them... so even though you have your nice RAID5 so you can lose a drive without losing data, suddenly, 3 years in, you get multiple drives failing within a week of each other.

I suspect that's what's happened here, but knowing very little about the actual setup, I can't realistically make assumptions. Anyway, this kind of situation actually happens quite commonly; I've heard numerous horror stories.

Hope this is helpful information; if not to Rogean, then to someone else.

Originally Posted by dhoushi

Sounds like to me greater stability is then formed by spreading the MTTF around. If that is the case, the way to go is to upgrade a drive in 6 months to a year on a continuing basis (a drive here, a drive there, obviously not the same one), which will progressively create better variance in the MTTF and reduce the strength of the failures when they occur.

The obvious downside to this method is the additional costs this method would require in the initial steps of the upgrade, (I might make an educated guess it might be possible to use the older replaced drives later in rotations, if the MTTF still has a good lifespan, to continue variance and save money) long term this would even out as failures become weakened in strength (replacing one drive instead of 2 or more).

Either that, or just, as soon as you see a single failure, go "OMFG" and haul ass to get that drive replaced and hope another one doesn't fail while you're rebuilding the array.