RMAN issues when backing up Physical Standbys

In releases prior to 11g each database has an associated DBID when using the RMAN utility.  It just so happens that all physical standby databases have the SAME DBID as the primary so the RMAN catalog gets confused when doing any command against that DBID. 

Recommendation to tag your backups with the NODENAME to uniquely identify this HOST/SID/DBID combination:

backup tag db_level_${RLEV}_${ORACLE_SID}_${NODENAME}
 
Excerpt from SR that was given to me by Kenney M. Snell, an attendee at IOUG COLLABORATE 08 , this helps with orphaned records left in the RMAN catalog due to the same DBID issue.  Contact Kenney at ksnell@ups.com

Problem:

 Recovery Manager complete.
Database…. backed up successfully…
Failed to validate restore of database.
ACTION : Please check the logfile for further details.

 

I have an understanding of the bug and the workaround, I have these final questions. I maybe asking some questions you have already answered, but I want to bring closure to this SR.

UPDATE

=======

1. Will this be fixed in 10g or is this fixed in 11g?

It is marked as fixed in 11g release 2.

2. We are looking at issuing full RMAN backups (datafiles and logs) on both the Primary and the Standby. Our RMAN backup script has a ‘LIKE’ command for the Archive logs on both sides. I reviewed the use of TAGs to make sure the RMAN knows which backup came from which server. Do we need to change our RMAN script and add TAGS on both the datafiles and the archive log backups?

No, the known problems with backup of standby/primary is associated with archivelogs only. RMAN catalog sometimes gets confused with on which server the archivelog files were backed up. These issues have been addressed in 11g.

 3. As you are aware, we have the soft links present on both sides (primary, standby) servers, which prevents the bug that is documented for this SR, to occur. Now if we are performing restores (ex, data file) for a Standby that has a full backup performed on a Standby,

3a. Is the same soft link that prevented the bug on restore database validate, going to cause an invalid restore or prevent a successful restore? If so, what is the workaround?

 The link should not cause an issue with the restore. The workaround would be to explicitly use ‘set newname’ based on which server you are doing the restore.

 3b. If we used a TAG statement for our backups, would this resolve the issue?

Ex. Bbackup database include current controlfile tag='<server name used in BACKUP command

>’;Ex. RESTORE DATABASE FROM TAG ‘<server name used in BACKUP command>’;

 No, because the issue was not in finding the backup, it is in realizing which location to place the datafiles.

 4. I am getting some resistance on implementing your work around from a previous SR where you recommended the ‘LIKE’ command for archivelog backups. Ex. backup archivelog like ‘$ORACLE_BACKUP_DIR/$ORACLE_SID/archlog_backup/%’ delete input;

If we continue to use the archivelog backup using the (backup archivelog all delete input;), do we need to use TAGs for each backup site, in order to perform (ex. datafile) restores on the Standby?

 No, again, the use of ‘like’ is resolve some of the maintenance issues associated with archivelog files in a standby/primary environment. When the restore occurs the files are in backupsets and RMAN will simply restore to the location based on the server on which the restore is being performed.

 You may be interested in:

Note 357759.1 Known RMAN – Data Guard Problems

 
 

 

Advertisements

About April C Sims

Oracle DBA for over a decade...enough said.
This entry was posted in RMAN and tagged . Bookmark the permalink.

5 Responses to RMAN issues when backing up Physical Standbys

  1. Teddy says:

    I am having a similar issue:

    I am using 10.2.0.4 on Windows Crapware and am trying to delete the ARCHIVELOGs on the Standby using RMAN DELETE ARCHIVELOG command.

    I have one physical db and multiple standbys. Right now, RMAN is getting all confused about the different standby databases because they have the same DBID.

    Question: Are there any other steps to prevent RMAN from getting confused about the multiple Standby DBs with the same DBID so that it can delete each individual Standby database’s ARCHIVELOGS?

    • aprilcsims says:

      There are archivelog deletion policies… CONFIGURE ARCHIVELOG DELETION POLICY TO [CLEAR | NONE | APPLIED ON STANDBY];
      There is an attached article to the document below that outlines all of the different scenarios…
      Using RMAN Effectively In A Dataguard Environment. [ID 848716.1]
      It is also recommend to “tag” the backups on your production with the nodename (see RMAN docs on how to tag backups).
      backup current controlfile for standby tag=’NODENAME’;
      Not enough information to help much further…look at all the docs suggested.

      RMAN backups in Max Performance/Max Availability Data Guard Environment [ID 331924.1]…the following is quoted from the document, not sure if this applies in your situation. I am assuming you are running in max performance mode. Also see doc Maintenance Of Archivelogs On Standby Databases [ID 464668.1]
      The 10g RMAN-Data Guard paper states that ‘archive log deletion policy applied on standby’ only checks that a log has been applied on all mandatory standby databases.
      However, Data Guard best practice for Max Performance and Max Availability modes is to *not* use mandatory standby (so that primary does not hang in the event of online logs filling up). Therefore, the following parameter must be set:

      _log_deletion_policy=’ALL’

      This will allow RMAN to delete logs from the Flash Recovery Area on primary database host that have been applied to all standby databases (whether mandatory or optional). Database will require bounce after setting this parameter.

  2. Bino says:

    Hi Aprilcsims
    I would like to know whether I can take rman backups (both level 0 & incremental) in active data guard with real time apply on without a catalog database.
    Regards,
    Bino

    • Bino says:

      Sorry the version is 11.2.0.3.2

    • April C Sims says:

      I don’t see why not….shouldn’t be an issue….but using a catalog database would allow you to use those backups to restore to any of the other physical standby databases in the dataguard configuration.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s