107 lines
3.3 KiB
Groff
107 lines
3.3 KiB
Groff
'\"macro stdmacro
|
|
'\" t
|
|
.TH cachefs 4
|
|
.SH NAME
|
|
cachefs \- Cache File System Description
|
|
.SH DESCRIPTION
|
|
.B Cachefs
|
|
is a file system layered above other standard IRIX file systems. It
|
|
performs automatic local caching of data from one file system (typically
|
|
remote) on a local file system.
|
|
Any of the following file systems may be used for the cached data:
|
|
.I xfs
|
|
and
|
|
.IR efs .
|
|
Any of the following file systems may be used as the source of data
|
|
to be cached:
|
|
.IR nfs ,
|
|
.IR nfs3 ,
|
|
.IR iso9660 ,
|
|
.IR cdfs ,
|
|
.IR hfs ,
|
|
.IR kfs ,
|
|
and
|
|
.IR dos .
|
|
.P
|
|
Data is cached when it is first accessed. Subsequent accesses will retrieve
|
|
the data from the cache. Consistency will be maintained
|
|
in a manner similar to that used for NFS.
|
|
.SH TERMINOLOGY
|
|
.RS 0i
|
|
.IP "\f3front file system\f1"
|
|
This is the local file system on which cached data is stored.
|
|
.IP "\f3back file system\f1"
|
|
This is the file system which contains the definitive copy of the data.
|
|
Data is retrieved from this file system and cached on the front file
|
|
system.
|
|
.IP \f3cnode\f1
|
|
This is the internal data structure used by
|
|
.B cachefs
|
|
to manage its files. There is one
|
|
.I cnode
|
|
for each open file. Each
|
|
.I cnode
|
|
occupies 128 bytes on a 32 bit system and 256 bytes on a 64 bit system.
|
|
In addition, each
|
|
.I cnode
|
|
points to a dynamically allocated area of 512 bytes containing metadata
|
|
and attributes.
|
|
.RE
|
|
.SH TUNING
|
|
.B Cachefs
|
|
has a number of different tuning parameters, some of which are supplied
|
|
on the
|
|
.IR mount (1M)
|
|
command (see
|
|
.IR fstab (4)),
|
|
some of which are supplied via
|
|
.IR cfsadmin (1M),
|
|
and some of which are kernel tuanbles adjustable with
|
|
.IR systune (1M).
|
|
.P
|
|
The kernel tunables are adjustable on a running kernel and control
|
|
readahead, asynchronous operations, and
|
|
.I cnode
|
|
caching. The kernel tunables are described below.
|
|
.RS .25i
|
|
.IP \f3cachefs_readahead\f1
|
|
This controls the number of blocks to read ahead of the
|
|
current block being read. These will be read asynchronously.
|
|
The size of a block is whatever the preferred I/O size is for the
|
|
front file system.
|
|
.IP \f3cachefs_max_threads\f1
|
|
This is the maximum number of asynchronous I/O daemons allowed to be
|
|
running per
|
|
.B cachefs
|
|
mounted file system.
|
|
.IP \f3fileheader_cache_size\f1
|
|
This is the size in 512-byte units of the in-memory cache of file header
|
|
information (cached attributes and other metadata). This number
|
|
indicates the number of such headers retained after other system
|
|
internal data (e.g., the vnode) for a file has been released.
|
|
.IP \f3replacement_timeout\f1
|
|
This controls the time between reconstructions of the replacement list
|
|
by the replacement daemon (cachefs_replacement). If no replacement
|
|
requests are made by the kernel within this time limit, the daemon
|
|
will time out and reconstruct the list.
|
|
.SH CONFIGURATION FILES
|
|
Mounts of cachefs file systems performed at system startup use options
|
|
supplied in
|
|
.I /etc/config/cachefs.options.
|
|
This file may contain any options for
|
|
.IR mount (1M)
|
|
desired. Of particular interest is the "-m" option. This option
|
|
allows the administrator to limit the number of parallel mount
|
|
processes mounting cachefs file systems. This will be required on
|
|
systems where some cachefs file systems fail to mount due to a lack
|
|
of swap space.
|
|
.RE
|
|
.SH LIMITATIONS
|
|
Mandatory file locking is not supported on
|
|
.BR cachefs .
|
|
.SH SEE ALSO
|
|
.IR cfsadmin (1M),
|
|
.IR fstab (4),
|
|
.IR systune (1M),
|
|
.IR mount (1M)
|