It's bad practice primarily because RAID5 can only sustain one disk failure. The problem with this is that when replacing the disk, you're putting a huge amount of strain on the others - meaning that the chances of another failure (And therefore losing the whole array) during the rebuild is surprisingly high.
This is compounded even more by the fact that people tend to buy all their disks from the same supplier at the same time - meaning they all come from the same batch so will probably fail at around the same time anyway.
Depending on use, RAID1, RAID6 or RAID10 are considered good day to day setups now.
This is compounded even more by the fact that people tend to buy all their disks from the same supplier at the same time - meaning they all come from the same batch so will probably fail at around the same time anyway.
What do you recommend to prevent that?
Simply buying the same size drives from different manufacturers? Or the same drives from different resellers to try and avoid the same batch?
Either one of those is the theory, but in truth, few people do. It's even harder when you consider that, really, to get the best support and reliability you should be buying disks from your server / SAN manufacturer. So, for example, we put in a HP SAN last week with about 70 disks - we pretty much have to buy them from HP, so you're kind of stuck there.
It's just one of those things to be aware of, and a good reason to avoid RAID5 at nearly all costs.
Ahh, yes, I seem to recall that that's where I was heading...except that no one who pontificates over these things on t'webz actually makes a hard-and-fast recommendation that those of us with goldfish-like attention spans can follow.