Re: BBLISA: data recovery services for oracle datafiles

[Rick Perron]

Just for the record, in case anyone is following this thread to pick up some ideas about Oracle.  I just want to be clear about terminology here. Do not shut down just the listeners for backups.  If you really can make the database unavailable, shut the entire database down and take a cold backup.  Otherwise, if you cannot make the database unavailable, and need to be able to restore the database run in ARCHIVELOG mode and take Hot Backups. Also, depending upon how quickly you need to get your database back up and running, exports may work for some situations.  Exports can be used in either ARCHIVELOG mode or NOARCHIVELOG mode.  It can just take a hell of a lot longer to import them later on. We purposefully and knowingly did not take proper backups of this particular database.  We got very lucky that we were able to restore it.  It was only because our user activity had finished before the backups kicked off , and the backups finished before the nightly scripts kicked off.  And even then, it was still luck. Daniel Hagerty wrote: Lisa Koch <[EMAIL PROTECTED]> writes: I was once Unix admin for a company that did Oracle front-end development. We ran Oracle in NOARCHIVELOG mode too. But we shut it down regularly for cold backups. If you stop the Oracle "listeners" before your backup and start them again afterward, you can get perfectly useable backups. I have recovered from such backups many times without problems (we were always When you write threaded code but don't worry about locking between the threads, the code will work *almost* every time. You're doing the same thing with your backups. It might work almost every time. In my mind, that's not a good property for disaster recovery. Even if you shut down the listener, e.g. oracle *might* decide to schedule a checkpoint in the middle of your backup. At which point you lose; if you are moving bits before and after they will reflect different SCNs. I'd imagine there are even situations where oracle can't detect a problem; you just get corrupted data at logical layer. Oops. Oracle doesn't even like to make promises about databases backed up during "alter database suspend"; at least in that case your gambling with the oracle's IO layer being quiescent. Just because something works doesn't mean it will work when you actually need it. As someone else said, it's clear that you could actually do the right thing here. Why gamble against stacked odds?