Revisiting "too small to fail"

Posted Jun 2, 2017 18:11 UTC (Fri) by Wol (subscriber, #4433)
In reply to: Revisiting "too small to fail" by nix
Parent article: Revisiting "too small to fail"

I think a lot of people run raid-1 split over machines.

My moan at the moment is people who think that just because raid CAN detect integrity errors, then it SHOULDN'T. Never mind. I ought to use it as an exercise to learn kernel programming.

But it does appear that a lot of what the kernel does is still stuck in the POSIX mindset. It would be nice if people could sit down and say "POSIX is so last century, what should linux do today?". I think the problem is, though, as Linus said, it's like herding cats ...

Cheers,
Wol

Revisiting "too small to fail"

Posted Jun 3, 2017 21:55 UTC (Sat) by nix (subscriber, #2304) [Link]

That's not as much of a problem as I thought it was. raid6check does what's necessary (it's just, uh, not installed by default). Frankly, I'm happy to wait for the immensely rare occasion when mismatch_cnt rises above 0 on a RAID-6 array and then run raid6check on it: complete automation of events as rare as that isn't terribly important to me. (But maybe it is for others...)