Open System Services Management and Operations Guide (G06.30+, H06.08+, J06.03+)

SECURITY-PRV-ADMINISTRATOR (SPA) security group and a member of the super group
(255, nnn), but is not the super ID (255, 255) or a member of the Safeguard
SECURITY-OSS-ADMINISTRATOR (SOA) group.
For more information about restricted-access filesets, see “Using Restricted-Access Filesets and
File Privileges” (page 228).
• You can use the ALTER FILESET command on a fileset that is not in the STOPPED state. However,
the changes do not take effect until the fileset is stopped and restarted.
• Assigning a new storage-pool file to a fileset has no effect on existing OSS files in the fileset.
New OSS files will be created only on disk volumes listed in the new storage-pool file. Disk
volumes with existing OSS files in the fileset remain a part of the fileset’s storage pool even
when they are not listed in the storage-pool file (are not in the creation pool).
• HP recommends that you choose a large and equal value for both INODECACHE and
LINKCACHE attributes to minimize disk I/O and to increase performance. However, choosing
a large value for both INODECACHE and LINKCACHE attributes results in high memory usage
by the Name Server. A typical INODECACHE cache entry is 300 bytes and a typical
LINKCACHE cache entry is 120 bytes. As an example, setting the caches at 64000 results
in a total memory requirement for the two caches of approximately 26,880,000 bytes (25.6
MB).
• If a fileset is in the STARTED state, you cannot change its OSS name server process.
• The pathname specified by the MNTPOINT option must be an existing directory, but the OSS
Monitor does not validate this until an attempt is made to start the fileset. Once validated, a
normalized version of the pathname is used for the mount point for sorting purposes, so the
apparent pathname for the mount point in an INFO FILESET command might not appear to
be the same as the specified pathname.
• The MAXDIRTYINODETIME option is meaningful only for filesets that use the BUFFERED LOG
option.
A fileset with an updated (flushed) inode cache is considered “clean” instead of “dirty” and
does not need recovery after a failure. The more often the inode cache is flushed, the less
likely a fileset is to be corrupted by a failure and to need recovery after the failure.
The larger the value specified for seconds2, the more likely that a fileset recovery is needed
after a failure, but the faster fileset access becomes because fewer disk writes occur to update
the cache from memory. The smaller the value for seconds2, the less likely that a fileset
recovery is needed after a failure, but the slower fileset access becomes because more disk
writes occur to update the cache from memory.
Fileset recovery delays subsequent availability of the fileset, so the tradeoff for slightly faster
current access is increased probability of delayed access after a failure.
• The MAXINODES value specifies an upper bound on the number of inodes that can be created
in the fileset. It does not guarantee that MAXINODES number of inodes will be created in the
fileset. Specifying a large MAXINODES value increases the fileset recovery time in the case
of an outage. HP recommends that you specify a MAXINODES value less than or equal to
1000000.
• FTIOMODE must have a setting equal to or higher than the setting of NORMALIOMODE.
NORMALIOMODE settings are ranked, from highest to lowest:
UNBUFFEREDCP
DP2BUFFEREDCP
OSSBUFFEREDCP
DP2BUFFERED
OSSBUFFERED
• If an FTIOMODE setting of OSSBUFFEREDCP or a NORMALIOMODE setting of
OSSBUFFEREDCP or OSSBUFFERED is used, the OSS filesystem buffers the data unless there
OSS Monitor SCF Command Reference Information 277