nate <debian-user@aphroland.org> [2002-08-15 13:21:21 -0700]: > Chris Jantzen said: > > > In RAID you can sustain a single drive failure and continue functioning. > > A master/slave drive relationship guarantees a two drive failure. So if > > you're in it for the redundancy at all, setting drives up on > > master/slave with IDE is a recipe for disaster. Period. > > I've had a drive fail in a master/slave relationship and it did > not affect the other drive in any way The type of failure is important. If the *media* fails then what you say will certainly be true. If the *controller* fails most of the time what you say will be true. But if the *controller* fails in such a way as to short out the bus then that will no longer be true. (I have had that happen once to me.) Redundancy there will protect you from that failure mode. You just need to decide how much redundancy you need to achieve your reliability goals. Certainly bad controller failures which take down the bus are more rare than other general failures which don't. So you might not need that level of redundancy. Life is a tradeoff and this is just another one of those judgement calls. I mean in the extreme case you would need off-site fail over replacement systems in the case that fire or flood takes out your main site but most people don't need that either. Normal drive failures are the typical case and worthwhile to protect against in isolation of other problems. Bob
Attachment:
pgppFqiADHrX9.pgp
Description: PGP signature