Re: YA RAIDframe thought

[Greg Oster <oster%cs.usask.ca@localhost>]

der Mouse writes:
> I've been dealing with more RAIDframe stuff, and another thing occurs
> to me: after a member has failed and been reconstructed onto a spare,
> it would be nice - especially for autoconfigured RAIDs - if there were
> some way to update the live RAID to elimiate the spared member,
> converting the used_spare into a real replacement for it.  Otherwise,
> after a few failures with no intermediate reboots you can easily wind
> up with a RAID set half of whose members are used spares - and have
> trouble because there's a disk that was long ago spared out that you
> want to now add as a spare, but it's already a member....
> 
> Is there any such thing?  I haven't been able to find it, and was
> tempted to implement it even if just for my own use, but thought I'd
> (a) see if it's there and I just missed it and (b) ask if anyone else
> might be interested.

There isn't such a thing yet.... It's on my TODO list of items that 
need to be fixed someday... The one difficulty is that to fix this 
properly requires a major restructuring of how disks and spare disks 
are handled in the kernel.  I've not made time to analyse all the 
corner cases that will result from making such a radical change... 
(But yes, it would be good to do...)

> I'm thinking something like "raidctl swapspare /dev/sd0e raid2" (which
> would either eliminate spared sd0e and replace it with its spare, or
> move the used_spare sd0e into place instead of of the disk it's spared
> out, depending on whether sd0e is spared or used_spare).

My thought was to just have the 'used spare' disk show up as a 
full-fledged member of the array... The 'spared' disk could either 
show up in a 'spared' section, or just simply be turfed from the 
array.. 

Later...

Greg Oster