Results 1 to 4 of 4
  1. #1
    Join Date
    Dec 2003
    Posts
    1,074

    Unanswered: Restoring Controlfile from backup

    If you have to resort to restoring your controlfile from backup, the documentation mentions that you must perform media recovery, then resetlogs.

    What I'm confused about is the state of the database afterwards. The datafiles are intact, but during media recovery, would RMAN still restore datafiles from backup and then apply redo logs, but only up to the SCN listed in the backup controlfile? Or are the intact datafiles kept in place, but if so, how would any redo be applied during the required media recovery since the SCN in the controlfile is likely earlier than the latest SCNs in the datablocks in the tablespaces?

    No catastrophe, just still trying to get a handle on what I'm reading in the manuals.

    -cf

  2. #2
    Join Date
    Jul 2003
    Posts
    2,296
    once you restore the control file(s) you will need to do media recovery.
    you can recover up to the most recent SCN or until-time based on your archive-redo logs

    There are differences if you have a recovery catalog or not. If you do not have a recovery catalog you will be pretty screwed since all the RMAN information was logged onto that controlfile that you no longer have.

    Code:
    http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96566/rcmrecov.htm#1009235
    - The_Duck
    you can lead someone to something but they will never learn anything ...

  3. #3
    Join Date
    Dec 2003
    Posts
    1,074
    As far as using a Recovery Catalog or not, we're trying to find the utility of the Catalog here. If you set the RMAN 'CONFIGURE CONTROLFILE AUTOBACKUP ON;' it sounds like will create a backup whenever there's a change to the database that a redo log cannot recover.

    And I found that '...if you are using a flash recovery area, RMAN implicitly crosschecks backups and image copies listed in the control file, and catalogs any files in the flash recovery area not recorded in the restored control file. This improves the usefulness of the restored control file in the restoration of the rest of your database.'

    So it sounds like the backup controlfile updates itself with any redo logs that it was previously missing.

    At this point, we're getting the impression that the recovery catalog in 10g is more of an additional piece of maintenance rather than a necessary component, especially when the backup controlfile has the properties mentioned above, in addition to the original controlfile residing on a RAID 10 array.

    -cf

  4. #4
    Join Date
    Sep 2003
    Location
    Virginia, USA
    Posts
    246
    This is a good question, but please allow this one reality check ... how do you anticipate losing every multiplexed copy of your control file simultaneously??? Let's say you have between 3 and 6 control files stored on multiple, different disks (not all on one RAID array), so even if you lose two disks simultaneously you still have not lost your control file. You only lost some expendable copies of it. After you replace the disks, simply copy the remaining control file to the new disks and edit your init.ora file to reference the new control file copies. Done. No need to "recover using backup controlfile."

    You could lose all copies simultaneously if the file system was not planned correctly. I just hate seeing systems they take all the disks and RAID them into one volume (drive letter.) Then some day a sys admin working the night shift drops that volume by accident. It's real easy to unmount the D drive, if you know what I mean.

    The only time I lost all control files was when I was replicating (at the operating system level) using an unsupported technology that was sending blocks in random order. When we cut the network cable, the control file data blocks had been sent but their header blocks had not been transferred yet, so all of the control files were considered corrupt by Oracle. So, I just did a recovery from backup control file. No biggie. Well, except you lose the contents of your redo logs, so that's why I keep my redo logs at 24 - 32 MB, and never make a redo bigger than 48 MB. Ask the boss one day "how many MB of your critical business data can we afford to lose?" Sometimes its better to suffer from over checkpointing that to put your business at risk.
    MarkRem
    Author, Oracle Database 10g: From Nuts to Soup
    http://www.remidata.com/book_nuts2soup.htm

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •