System metadata backup and restore

SAN File System manages data and metadata separately. When you back up a file, only the file's data and attributes are backed up. For disaster recovery purposes, you must also back up system metadata (which includes information about fileset attachment points, storage pools, volumes, and policies) separately.

You can create a file that contains a backup copy of system metadata either from the SAN File System console or from the administrative command-line interface using the mkdrfile command . The file, which is stored in the /usr/tank/server/DR directory on the master metadata server's local disk, contains everything that is required to recreate the system metadata. When needed, you can use the contents of this file (along with normal restore processes for file data) to recreate the state of the cluster.

To restore system metadata, you process the information that is contained in the system metadata backup file using the builddrscript command. This command creates several scripts that you must first review in order to obtain a restore scenario, and then run to recreate the SAN File System configuration. After the system configuration information is restored, you can then restore the user data files from the SAN File System clients.

You should run the mkdrfile and builddrscript commands often to ensure that any configuration changes are reflected in the output of these commands. You should store copies of the output of the mkdrfile and builddrscript commands in an easily recoverable location on backup media where critical system and application files are kept for backup and restore purposes.

Note: To assist in protecting against the corruption of metadata and other metadata failures, you can check the metadata from the SAN File System console or by using the startmetadatacheck command. This command performs a consistency check on the system metadata, and optionally repairs any problems it finds. It allows you to check file metadata for one or more filesets, the system metadata, or both. There is also an option to check only the metadata structure, or to check the metadata structure and its contents.
There are three cases when you might need to perform a consistency check or repair operation:
  • As part of a regularly scheduled cycle of preventive maintenance
  • In response to an alert that recommends that this operation be performed (extra detail might be supplied that specifies the restore option that you must use in order to salvage the metadata)
  • If metadata corruption (or any other SAN File System corruption) is suspected

If the check-metadata operation cannot resolve the problem, you must perform a full restore of SAN File System, beginning with restoring the metadata. It is critical that mkdrfile is run in order to recover from such a situation.

Parent topic: Backup and restore

Parent topic: Planning the backup and restore strategy

Terms of use | Feedback
(C) Copyright IBM Corporation 2003, 2004. All Rights Reserved.