Quote Originally Posted by rysiekmus
Hi ,

the main reason is that if you get a redolog corrupt, the mirrored redolog will also be corrupt. If you set up a mirrored "by Oracle" (another phisical drive) set of redologs and one file in a set becomes corrupt, Oracle will not crash but will continue to work with the other set of redo logs until you fix the problem. Otherwise - when one redolog is corrupt, the mirrored by the operating system is corrupt as well and this could lead you to having the database crashing.

The problem these days is that - like in my work I get a Windows server with all drives initially mirrored (typically only C: and D: ) and it's hard to follow Oracle recommendations.

O hope this helps,

Regards,
Richard.
I would have to disagree with you Richard, and say that the main reason for not using RAID5 with redo logs is for the write overhead as previously stated.

The fact that if you "get a redolog corrupt", you'll lose them all doesn't really compute with me. What type of corruption are you talking about? Corruption from data? If you are talking about hardware corruption then i'm not sure you have properly considered the benefits of RAID in that statement. If a disk fails then you hot swap it out (no data loss). If a RAID controller dies, the backup controller should kick in (no data loss). Please expand on your statements so that I fully understand.