Support of linking GPFS dependent fileset

Active File Management (AFM) supports to create and link a dependent fileset inside an independent fileset.

AFM can be enabled only on a GPFS independent fileset. However, the user can create a GPFS dependent fileset by using inode space of an AFM or AFM-DR fileset and link this dependent fileset under the AFM or AFM-DR fileset for the data synchronization. All data and metadata operations that are generated inside the dependent fileset, which is linked inside AFM fileset, work seamlessly. The user can create data inside the AFM or AFM-DR fileset path and the linked dependent fileset path. AFM synchronizes the created data to the home or secondary fileset under the same path. AFM queues operations that come under the root path of the AFM fileset and the linked path of the dependent fileset and flushes the operations to the home. Also, during the prefetch, AFM can bring data from home dependent fileset path to the AFM fileset that has the dependent fileset that is linked at same path. The path location must be the same at the cache or primary and the home or secondary site.

All I/O requests that are generated under the linked dependent fileset use the same gateway node that is assigned to the AFM or AFM-DR fileset. The gateway node maintains the directory structure during the data synchronization and keeps the home or secondary in sync.

To define an AFM relationship between a cache or primary and a home or secondary, stop the specific AFM or AFM-DR fileset, and then link the dependent fileset at the same junction path under the AFM or AFM-DR fileset. That is, the relative path under the AFM or AFM-DR fileset at both sides must be the same so that all operations generated at one site sync to the appropriate path at the other site. When the dependent fileset is linked inside the AFM or AFM-DR fileset, the user must issue Start command to start the replication.

Prerequisite for linking dependent fileset under an AFM or AFM-DR fileset:
  1. Set up a dependent fileset at both cache or primary and home or secondary sites and then start replication. Because AFM does not replicate dependent fileset information between cache or primary and home or secondary sites.
  2. Ensure that the dependent fileset link path (junction path) is the same relative path to the AFM or AFM-DR fileset at both sites.

    That is, the linking of dependent fileset inside AFM or AFM-DR fileset must be identical at both sites.

  3. Create and link dependent fileset at a new location to perform failover or changeSecondary operation on the AFM or AFM-DR fileset, which has the linked dependent fileset.
    Note: Ensure that the failover or change secondary site has an identical path to the cache or primary for the data synchronization before you perform the failover or changeSecondary operation.
  4. Run the mmafmconfig enable command on each dependent fileset that is linked inside an independent fileset at the home/secondary.
To link a dependent fileset inside an AFM or AFM-DR fileset, do the following steps:
  1. Create an AFM or AFM-DR fileset and set up an AFM relationship.
  2. Create a dependent fileset by using inode space of the created AFM or AFM-DR fileset under which you want to link the dependent fileset at both sites.
  3. Stop the specific AFM or AFM-DR fileset at the cache or primary.
  4. Link the dependent fileset inside the AFM or AFM-DR fileset at the same relative path at both sites so that the directory structure is the same across the sites.
  5. Start the AFM or AFM-DR fileset.
    Note: To unlink the dependent fileset, an AFM or AFM-DR fileset must be stopped first.
Example
  1. Create an AFM or AFM-DR fileset and set up an AFM relationship.
    # mmcrfileset <fs> <AFM/AFM-DR fileset> -p afmMode=<mode>,afmTarget=<AFMTarget> --inode-space New
    
    # mmlinkfileset <fs> <AFM/AFM-DR fileset> -J /< GPFS_PATH >/<AFM|AFM-DR link path>
    For more information, see Administering AFM.
  2. Create a dependent fileset by using inode space of the created AFM or AFM-DR fileset of the created AFM or AFM-DR fileset.
    # mmcrfileset <fs> <dep-fileset> --inode-space <AFM/AFM-DR Fileset>
  3. Stop the AFM or AFM-DR fileset at the cache.
    # mmafmctl <fs> stop -j <AFM/ADR Fileset>
    For more information, see Stop and start replication on a fileset.
  4. Link the dependent fileset at the relative junction path at both sites.
    # mmlinkfileset <fs> <dep-fileset> -J <GPFS_PATH/AFM|AFM-DR Path/Dependent Fileset>
    
  5. Start the AFM or AFM-DR fileset at the cache or primary.
    # mmafmctl <fs> start -j <AFM/ADR Fileset>
    For more information, see Stop and start replication on a fileset.
Note: This feature supports linking only the GPFS dependent fileset that is under the AFM or AFM-DR fileset root path. This feature does not allow linking the nested levels of the dependent fileset under the AFM or AFM-DR fileset.