From: Mike T. <mh...@us...> - 2005-07-18 19:39:05
|
On Fri, 2005-07-15 at 18:18, Michel Bouissou wrote: > Le Samedi 16 Juillet 2005 01:05, Michel Bouissou a =E9crit : > > > > > EVMS developers discussed this issue before... Why does the > > > administrator want to purposely degrade an raid1/raid5 array? It's > > > possible to run into I/O errors on another member while the array i= s > > > running in degrade mode, so we avoid this situation. If you want t= o > > > free up a member of raid5 array, you should replace it with a spare > > > (i.e. Mark Faulty). It's more safe this way, don't you think? > > > > And how if you want to replace a disk with another one, but cannot ad= d the > > other one FIRST as a spare, because you don't have room for another d= isk in > > your machine, or no more available controller, etc ? > > > > Then, you _need_ to first kick a drive out, then replace it. You can'= t do > > it the other way... >=20 > Maybe my example wasn't that good, because of course, you could simply = replace=20 > the disk you want to replace and let RAID discover that a disk has=20 > disappeared... >=20 > But in my precise case I wanted to replace the disk with _itself_ after= having=20 > repartitioned it. So I had to kick it out because I didn't need or want= to=20 > physically remove it. >=20 > You way also want to kick a disk out if you are in doubt and want to ru= n=20 > extended tests on it while offline, etc. And you may not have any avail= able=20 > spare -- but rely on existing good backups in case of a problem. >=20 > I like security features, but only as long as I can decide to override = them=20 > and stay master on board ;-) so I prefer a big ugly warning "are you re= ally=20 > sure you want to do this ?" rather than not to have the option at all... Your suggestion sounds reasonable. For evmsgui and evmsn, I can put up a big warning message. For evms, the default action is "Yes, go ahead" :) Look for this change in next release of EVMS. -- Regards, Mike T. |