On Mon, Dec 22, 2008 at 7:59 AM, Niall Litchfield < niall.litchfield@xxxxxxxxx> wrote: > I have however an SR about this with Oracle, since in 10g EM you cannot > apparently configure monitoring thesholds for the recovery area. In our > client case we have test databases that generate circa 30gb redo/day but > sporadically this jumbs to circa 0.5TB per day, not bad for a 100gb db with > 4 users!. They have 5 test databases and storage of 1.5Tb available to them. > Their complaint, quite correctly, is that we should spot when the flash > recovery area is filling up and do something about it. To quote from the SR > however we should just monitor disk space. Now it is true that one db > crashing is better than 5, but that isn't exactly a ringing endorsement > really. Oh and of course we'd get 5 notifications for the same low disk > space on archive dest issue from each of the 5 databases if I followed the > recommendation. > > I wonder how much of this is brought about by the 'one server, one database' philosophy? If that is what the Oracle architects have in mind, it certainly does not fit reality. These are good arguments for virtualization, though the only 'supported' model for that is based on Xen, aka Linux only. There are problem exceptions to that, such as LPARs on big iron, but most of us seem to be running on Windows of *nix. Jared Still Certifiable Oracle DBA and Part Time Perl Evangelist