Re: Recover from cold backup + archive logs question

From: Mark D Powell <Mark.Powell_at_eds.com>
Date: Wed, 20 Feb 2008 07:01:23 -0800 (PST)
Message-ID: <70f45a65-e2ae-45f8-8270-8884d81bc1d1@v3g2000hsc.googlegroups.com>


On Feb 18, 3:33 pm, "mpokr..._at_gmail.com" <mpokr..._at_gmail.com> wrote:
> On Feb 18, 9:21 am, Mark D Powell <Mark.Pow..._at_eds.com> wrote:
>
>
>
>
>
> > On Feb 18, 7:41 am, GD <goran99_rem..._at_yahoo.com> wrote:
>
> > > mpokr..._at_gmail.com wrote:
> > > > I have a new 10.2 DB setup on Linux.  I had an issue on 2/16 that
> > > > requires a recovery.  RMAN is not set up and working properly (yet).
> > > > That is job one once I get this sorted out.
>
> > > > I have a cold backup from 2/9 + a weeks worth of archive log files.  I
> > > > have copied the data files and control files from the cold backup into
> > > > the proper location.  I have left the archive log files in place,
> > > > assuming that the DB will look for them in the correct location.
>
> > > > In sqlplus I do
>
> > > > startup mount
>
> > > > recover database
>
> > > > ORA-00283: recovery session canceled due to errors
> > > > ORA-00264: no recovery required
>
> > > > How do I get my cold backup to process all the archive logs?  They are
> > > > in the default location under the flash_recovery_area/
>
> > > The database (cold backup) is consistent and requires no recovery to be
> > > opened.
>
> > > To make it apply the archived logs, you should:
>
> > > recover database until cancel using backup controlfile;
>
> > > Regards- Hide quoted text -
>
> > > - Show quoted text -
>
> > GD, that is how I would have done the recovery also.  I do believe
> > that if the current control file had been available that Oracle on
> > startup would have recognized that recovery was necessary and
> > assumming all archiveded redo logs were available been able to perform
> > automatic recovery.
>
> > HTH -- Mark D Powell --
>
> You know, I did not try that approach (old backup datafiles + current
> control files).  Our restore process on 8i has been to take old
> control files + old datafiles and then apply the archive logs.  Each
> transaction will update the SCN in the control files.  My
> understanding is that Oracle will not open the DB if the SCN of the
> control files differ from that of the data files, but I have no ever
> researched the specifics.
>
> What I did discover as I applied the archive logs was that the final
> log was corrupt.  It is possible that, had I removed that, a more
> conventional restore process would have worked.
>
> Has anyone ever seen a situation where two archive logs, in sequence
> (60 and 61, for example) were created at exactly the same time?  In my
> case, 61 was the bad one and the last one.  So I applied up through 60.- Hide quoted text -
>
> - Show quoted text -

If the Control file SCN is higher than the datafile SCN Oracle will not open the datafiles because recovery is required, which is what you want in this case. Now if you have a bad log and know it then you would want to perform a point in time recovery. Doing so will allow Oracle to open the database upon successful completion of the recovery even though the control file had a higher SCN before recovery started. PITR requires resitting the online redo logs on open.

HTH -- Mark D Powell -- Received on Wed Feb 20 2008 - 09:01:23 CST

Original text of this message