Backup procedure with SOBAR

This section provides a detailed example of the backup procedure used with SOBAR.

Throughout these procedures, the sample file system used is called smallfs. Where appropriate, replace this value with your file system name.

1. Backup the cluster configuration information.

   The cluster configuration must be backed up by the administrator. The minimum cluster configuration information needed is: IP addresses, node names, roles, quorum and server roles, cluster-wide configuration settings from mmchconfig, cluster manager node roles, remote shell configuration, mutual ssh and rsh authentication setup, and the cluster UID. More complete configuration information can be found in the mmsdrfs file and CCR.

2. Preserve disk configuration information.
   Disk configuration must also be preserved in order to recover a file system. The basic disk configuration information needed, for a backup intended for disaster recovery, is the number of disk volumes that were previously available and the sizes of those volumes. In order to recover from a complete file system loss, at least as much disk space as was previously available will be needed for restoration. It is only feasible to restore the image of a file system onto replacement disks if the disk volumes available are of similar enough sizes to the originals that all data can be restored to the new disks. At a minimum, the following disk configuration information is needed:

   • Disk device names
   • Disk device sizes
   • The number of disk volumes
   • NSD server configuration
   • Disk RAID configurations
   • Failure group designations
   • The mmsdrfs file contents
3. Backup the GPFS™ file system configuration information.
   In addition to the disks, the file system built on those volumes has configuration information that can be captured using the mmbackupconfig command. This information includes block size, replication factors, number and size of disks, storage pool layout, filesets and junction points, policy rules, quota information, and a number of other file system attributes. The file system configuration information can be backed up into a single file using a command similar to the following:

   mmbackupconfig smallfs -o /tmp/smallfs.bkpcfg.out925

4. Pre-migrate all newer file data into secondary storage.
   File contents in a space-managed GPFS will reside in secondary storage managed by the HSM. In the case of IBM Spectrum Protect™ HSM, disk and tape pools will typically hold the offline images of migrated files. HSM can also be used to pre-migrate all newer file data into secondary storage, so that all files will have either a migrated or pre-migrated status (XATTR) recorded, and their current contents are copied or updated into the secondary storage. The IBM Spectrum Protect command dsmmigrate can be used as follows:

   dsmmigrate -Premigrate -Recursive /smallfs

   To optionally check the status of the files that were pre-migrated with the previous command, use the following command:

   dsmls /smallfs/*

5. Create a global snapshot of the live file system, to provide a quiescent image for image backup, using a command similar to the following:

   mmcrsnapshot smallfs smallfssnap

6. Choose a staging area in which to save the GPFS metadata image files.

   The image backup process stores each piece of the partial file system image backup in its own file in the shared work directory typically used by policy runs. These files can become quite large depending on the number of files in the file system. Also, because the file system holding this shared directory must be accessible to every node participating in the parallel backup task, it might also be a GPFS file system. It is imperative that the staging directory chosen be accessible to both the tsapolicy archiver process and the IBM Spectrum Protect Backup-Archive client. This staging directory is specified with the -g option of the mmimgbackup command.

7. Backup the file system image.
   The following command will back up an image of the GPFS metadata from the file system using a parallel policy run with the default IBM Spectrum Protect backup client to backup the file system metadata image:

   mmimgbackup smallfs -S smallfssnap -g /u/user/backup -N aixnodes

   The metadata of the file system, the directories, inodes, attributes, symlinks, and so on are all captured in parallel by using the archive module extension feature of the mmapplypolicy command. After completing the parallel execution of the policy-driven archiving process, a collection of image files in this format will remain. These image files are gathered by the mmimgbackup command and archived to IBM Spectrum Protect automatically.

   If using the -N nodes option, it is recommended that the same operating system be used when running mmimgbackup. Note the directory created with -g GlobalWorkDirectory to store the image files.

8. After the image backup is complete, delete the snapshot used for backup with the following command:

   mmdelsnapshot smallfs smallfssnap