Cache and home
An AFM fileset can be enabled on a GPFS-independent fileset. Each fileset has a distinct set of AFM attributes. An IBM Storage Scale cluster that contains AFM filesets is called a cache cluster. A cache cluster has a relationship with another remote site called the home, where either the cache or the home can be the data source or destination.
AFM constantly maintains an active relationship between the cache and the home. Changes are managed per fileset results in modular, scalable architecture capable of supporting billions of files and petabytes of data. Each AFM-enabled fileset is associated with a single home path.
AFM uses an NFSv3 or NSD (GPFS multi-cluster) protocol for the communication between the home and cache sites. A home export path is either an NFSv3 exported path or a multi-cluster or remote file system, which is mounted at the IBM Storage Scale cache cluster. This path is used by an AFM or AFM-DR fileset as a target path for data synchronization between sites. For AFM RO-mode filesets, the target path at the home NFS server can be exported as 'Read-Only' or 'Read/Write'. However, for AFM LU/SW/IW and AFM-DR mode filesets, the target NFS export path must be 'Read/Write'.
- Export as Read-Only for an AFM RO-mode
fileset.
/ibm/gpfs0/homeDataSource GatwayIP/*(ro,nohide,insecure,no_subtree_check,sync,no_root_squash,fsid=1001)
- Export as Read/Write for an AFM-LU/SW/IW and AFM-DR mode
fileset.
/ibm/gpfs0/homeDataSource GatwayIP/*(rw,nohide,insecure,no_subtree_check,sync,no_root_squash,fsid=1001)
- If you do not run the mmafmconfig enable command and configure an AFM relationship, ACLs and extended attributes are not supported. Sparse file information is not maintained. The following error message appears in the mmfs.log file: AFM: AFM is not enabled for fileset cachefs in filesystem gpfs. Some features will not be supported, see documentation for enabling AFM and unsupported features.
Any cluster can be a home cluster, a cache cluster, or both. In typical setup, a home is in an IBM Storage Scale cluster and a cache is defined in another IBM Storage Scale cluster. Multiple AFM-enabled filesets can be defined in one cache cluster, and each cache has a relationship with targets with a home, or different cluster.
In IW, RO, and LU modes, multiple caches might point to the same home. But in SW mode, only one-to-one relationship between cache and home is supported. AFM can also be configured as a subscription service, where home is the data feed and all caches can subscribe to this data feed.
Within a single cache cluster, application nodes experience POSIX semantics. File locking across cache and home is not supported.
When you perform operations on AFM filesets, ensure that the operations are supported on home over the chosen protocol. Because the operations that are done from cache are replayed on the remote as normal file system operations. When you use the NSD protocol with UID remapping, operations such as chown (change ownership) are not supported.