About restoring a Greenplum database
NetBackup for Greenplum supports incremental restores from the incremental timestamps.
The following limitations apply to Greenplum database restores:
The NetBackup for Greenplum plug-in does not support the restore_subset option. It should stay set to off in the
nbgpbackup_config.yamlfile.The plug-in does not support the gprestore command with the --incremental option.
When you use the optimize_restore option, then NetBackup for Greenplum uses a cache directory. The agent makes every attempt to clean up the cache directory. However, if errors occur during a restore then you may need to manually clean up the
cachedirectory on all segment hosts.
For the best restore performance set incremental_optimize_restore: on in the nbgpbackup_config.yaml file. This setting starts a job for each image that is needed for the restore. By default, this setting is turned off and each restore_data request is a separate job. For a large number of tables, in the media server host properties update the to a higher value. For example, .
Table: Effect of restore options on number of jobs
Non-optimized | optimize_restore | incremental_optimize_restore |
|---|---|---|
Job for each time that nbgpbackup is called. This number can be thousands of jobs. | A job for each image that is used during the database restore. | A job for each image that is used during the database restore. |
Table: Restore options and use of the disk cache on segment hosts
Non-optimized | optimize_restore | incremental_optimize_restore |
|---|---|---|
No | Yes. Each image that is requested is restored to the | No |
Table: Effect of restore options on speed
Non-optimized | optimize_restore | incremental_optimize_restore |
|---|---|---|
Slowest | Fastest for any restores from a full backup. | Fastest for any restores from incremental backups. |
Table: Restore options and use of shared memory
Non-optimized | optimize_restore | incremental_optimize_restore |
|---|---|---|
No | No | Yes |