Video Screencast Help

RMAN Crosscheck and DELETE EXPIRED taking a long time for large catalog entries

Created: 14 Jan 2013 • Updated: 17 Mar 2013 | 6 comments
This issue has been solved. See solution.

For our larger databases with many 100's or 1000s of backup pieces the admin operations of RMAN are very slow when interacting with Netbackup. Are there tools on the netbackup side to diagnose this? We have verified with Oracle that RMAN catalog is not the issue. Its taking nearly 15minutes per delete backup piece operation. Any thoughts on this?

Is there an index or catalog operation for the media server catalogs that we should have the netbackup admins perform?

 

Thanks for your help.

Discussion Filed Under:

Comments 6 CommentsJump to latest comment

allknowing's picture

Oh .. before someone posts about the format of the file names .. they follow this format

"FORMAT 'bk_%d_%s_%p_%t' "

Will Restore's picture

Article URL http://www.symantec.com/docs/TECH49868

old technote probably still applies

 

NetBackup for Oracle Agent designed to reduce the chance that an Oracle backup could not be found for restore, due to complexities that can be introduced by certain Recovery Manager (RMAN) procedures and conventions.  Under normal circumstances, this allows an available backup to be found sooner than before these changes.  If the backup has expired in the NetBackup catalog, however, significant additional checking is done before RMAN is informed that NetBackup is not able to locate the backup image.  This results in longer execution time for RMAN "crosscheck" and "delete expired" operations.

...

A better approach to manage the catalog synchronization is to use the Oracle and Symantec recommended method.  This method is to set the backup retention as an RMAN attribute, rather than a NetBackup attribute, and having RMAN delete obsolete, but unexpired, backups from NetBackup

Will Restore -- where there is a Will there is a way

allknowing's picture

Yes . . we are following those best practices.. Having the admins look into this note:

 

Indexing the catalog for faster access to backups

If the NetBackup environment contains a large number of backups, consider

indexing the catalogs to reduce the time that is required to restore files.

To index the catalog means to create indexes of the files that are recorded in the

NetBackup image catalog. NetBackup uses the indexes to go directly to the catalog

entry for a file. Without indexing, NetBackup must start searching for a file at

the beginning of the catalog entries.

To index image header files, run the following command:

bpimage -create_image_list [-client name]

Run this command to create the following index files in each client image directory:

IMAGE_FILES

IMAGE_INFO

IMAGE_LIST

To stop image header indexing for a client, remove these files.

Additional information is available about bpimage in the NetBackup Commands

Reference Guide.

D Thomas's picture

Any update on this? We are seeing similar issues. There are mentions of indexing the catalog, but I have not seen any mention of it until this post.

-create_image_list is an unrecognized option in 7.5.0.4

DT

allknowing's picture

Indexing didnt make a difference. Promises of improvements in the move to mysql (?) repository.

The real fix for us was to not bother with the crosscheck .. simple tell RMAN that the images are gone rahter then crosscheck and delete expired.

With a 90 day netbackup retention, we just run this command.
DELETE FORCE BACKUP COMPLETED BEFORE 'SYSDATE-90';

Daryl.

SOLUTION
allknowing's picture

Others have used this as well..

CHANGE BACKUPSET COMPLETED BEFORE 'SYSDATE-90' UNAVAILABLE;
CHANGE BACKUPSET COMPLETED BEFORE 'SYSDATE-90' DELETE;