Re: Backup/Recovery Question

Rogers' first rule of backups is that you can *never* hot copy any Oracle file without the results being internally inconsistent.

The second rule is that Oracle provides mechanisms for getting Control Files consistent (alter database backup controlfile to 'blah') and for getting Data Files consistent (application of redo) from a known starting point (alter tablespace blah begin backup).

But online redo logs will be internally inconsistent (rule 1), and there is NO mechanism for getting them sorted afterwards. That means (unless you've missed out a detail or two) your controlfiles will be unuseable (because you aren't doing the 'alter database' command), your data files will be unuseable (because you're not doing the 'begin backup' stuff) and your online redo logs will be unuseable (because they just are).

So no, it's not a backup strategy at all, and it's worse than useless (as described). Export won't help you: import can only work when the database is in the fully open state, and with your present backup strategy, you haven't a hope in hell of ever getting it there. What you'd be forced to do is to create a brand new database, and import everything into that. And that's not exactly a 24x7 recovery strategy.

Frankly, it's a mess.

Regards
HJR

--

Oracle Resources : http://www.geocities.com/howardjr2000
========================================


"Buck Turgidson" <jc_va_at_hotmail.com> wrote in message
news:f98999c8.0111020523.15b23b19_at_posting.google.com...


> Our DBA's are using various backup/recovery strategies, but one is the

> following:

>

> A complete server (HP Unix) backup is done, including OS, and all

> relevant Oracle files.  Basically, it is a complete volume backup of

> the server.  On one of the volumes is a nightly export file of the

> database.

>

> The database itself is used 24/7, and is always in a state of flux.

> And it is not taken offline during the server backup.

>

> If we had to restore to another server in a disaster, it sounds to me

> like all the SCN's in the files, versus the control file would be out

> of sync, and there'd be problems.

>

> But would it be enough for them to get back and running if they used

> the Oracle OS files, and overlaid the Oracle export on top of them?  I

> would imagine they could get the database to a mount state, but

> probably not open it, prior to starting the import.

>

> Thanks for any replies.

>

> Buck