IBM Hub/Switch Installation Manual page 456

Chapter 7 HPSS User Interface Configuration
Fsid specifies the file system id for this aggregate. The value is defined by dfstab.
Option specifies how the filesets on the aggregate will be managed by HPSS. The parameter may
be either archive/delete, archive/rename, or mirror. If mirror is selected, the name and data space
will be mirrored by HPSS, and the end user can access the name and data space from either DFS or
HPSS. Otherwise, only files are archived by HPSS, and the files can only be accessed from DFS. If
archive/delete is selected, any files deleted from DFS will also be deleted from HPSS. If archive/
rename is selected, any file deleted from DFS is not deleted from HPSS, but renamed instead. This
allows for the possibility of restoring a file later, if it was accidentally deleted.
MigratePolicy and PurgePolicy specify the name of the policies that will be used to migrate and
purge files on this aggregate. The name of the migrate policy must appear in the MigratePolicy
section of policy.dat, and the name for the purge policy must appear in the PurgePolicy section of
the file. In the example above, the names of the policies are wait and run, but these names have no
special meaning to HDM. Section 7.6.3.3.4 discusses policy.dat in detail.
StageType specifies the type of staging HDM will use when it is necessary to stage a file from HPSS
to DFS. Legal choices are whole and partial. If whole is specified, the whole file will be staged;
otherwise only that part of the file necessary to satisfy a request will be staged. With partial, the
amount of data staged will be at least the size of the data access, or if the data access is small and
the file is large, a 16MB chunk of the file surrounding the data being accessed will be staged.
Following is a description of the parameters for a DFS fileset:
Ftname specifies the name of the fileset. The name should be the same as the name for the DFS
fileset, which is also the name of the HPSS fileset.
Global specifies the global mount point for the fileset. This name will be a DFS style path name. For
example, /:/hpss/mirror.
Local specifies the local mount point for the fileset. This name will be a UNIX style path name.
Typically, the mount point will be in the directory, /var/hpss/hdm/hdm<id>/aggr/<fileset name>,
where fileset name will be the same as the fileset name. For example, hpss.mirror.
Ftid specifies the fileset Id. The parameter is specified in the form <high>.<low>, where <high> and
<low> are numbers representing the high and low 32 bits for the fileset Id. This Id should be the
same as the DFS fileset Id, which is also the fileset Id of the HPSS fileset.
If a fileset uses one of the archive options, the global and local mount points will typically not be
set and filesys.dat will show NO_MOUNT_POINT for these parameters. If a mirrored fileset will
only be accessed by users local to the cell, the global mount point can also be specified as
NO_MOUNT_POINT.
When a fileset is only partially configured, the global and local mount points are each represented
by a '?'. While this condition exists, DFS users cannot access the fileset. Typically, this happens only
for a short period of time while the administrator is setting up the HPSS fileset. To complete the
configuration, an administrator will use SSM to create the HPSS fileset.
An administrator must not edit filesys.dat to "fix" the "?"s! This data will be assigned when the
HPSS fileset is created.
456 September 2002 HPSS Installation Guide
Release 4.5, Revision 2