> Yes. And no. Yes and yes. This stuff is binary.
> Maybe I wasn't all too clear? My expectation is not (yet) the automatic > recovery of the respective half mirror! Sure not! I don't expect > miracles. What I do expect, though, is a consistent, defined and > predictable state. Your expectations are out of whack with reality. > Please, try to view it from a different perspective. Nobody would > voluntary pull out disk A, plug it back after 20 seconds, expecting it > to recover the mirror, pull out disk B after another 10 seconds, and > plug it back after 20 seconds, and still expect a full mirror! The plugging in of the disk is a non-event. The disk is dead to the OS and by extension to softraid. > But, and that's a big 'but' for me: some fault might do exactly that, a > flimsy controller, a faulty power supply. And then I don't want I/O > errors, and neither a panic at reboot. My expectations are much lower, > but based on consistency: Panics shouldn't happen and I'd love to see traces so that I can fix those. > 0. Running sane raid > 1. One drive goes offline > What I'd expect, personally, would basically be minimally: > A. Immediate info about a drive lost. That is there. And you haven't shown me any evidence it isn't. > B. 2 half mirrors remaining that I can plug into another box, at least > to access the data on either. No, 1 half mirror; the other one is basically lost. You got an IO error for some reason. There is no telling what didn't get written to it after the remaining chunk continued on its merry way. > C. No further attempt to use that drive that went offline any longer, at > least not until a reboot. Right, and softraid will detect that it went tits up prior and ignore it. > D. That means, I won't have I/O errors, but the system running happily > from the active drive, Right. > E. And it means that a reboot will go through smoothly. Right. > > I am aware that this implies, that when the second drive goes offline as > well, that NO more drive is available (even if either came back!). As I > mentioned, I request consistency of data, not necessarily uptime. I want > to be abe to retrieve the data from the drive that went offline first, > and I want to be able to retrieve data from the drive that went offline > later. Personally, to me RAID is not failover, or availability, but > access to the data up to and until that moment when a drive goes offline. > And I want a clean reboot, irrespective of all ups and downs of the drives. There is no way for softraid to know if you ripped out the last drive and then use it on another box or reboot it with it reinserted. Softraid can only keep state on its disks and if you rip them all out state can no longer be kept. If you need way more fancy raid then this you should use mfi or ami. Inherent to software RAID there isn't any hardware assist such as flash to keep state.
