presto(8) CLIX presto(8)
NAME
presto - Controls and monitors the Prestoserve NFS accelerator system
SYNOPSIS
presto [-FlLpRUv] [-h hostname] [-u|d [special device...] ]
FLAGS
-d Disables acceleration on selected file systems, flushing any
dirty Presto buffers to their disks. If no file systems are
specified, then all accelerated file systems will be affected,
and the state will be set to DOWN. This option does not reset
statistics. This option takes effect before the -u or -R
options.
-F Flushes back any currently dirty buffers back to disk without
disabling.
-l Lists the file systems and unmounted block devices that have
been accelerated.
-L Acts much like -l, but any unusual state for each file system is
shown after its mount point. Unusual per file system state
includes the following: disabled - acceleration is not enabled
on this mounted file system, enabled -accelertion enabled on
this mounted file system, and BAD MARK - mismatch between ID
mark on board and the ID mark for the device; buffers could not
be flushed.
-p Prints information. The information is grouped into two areas:
current Prestoserve board state, and statistics for write, read,
and total operations.
The current board state accounts for all buffers. A dirty
buffer is one that contains a disk block image that has yet to
be flushed to disk. A clean buffer contains a valid disk block
image that has been flushed to disk. An invalid buffer does not
presently contain a disk block image. An active buffer is one
that is currently in transition to the actual disk; for example,
a write operation has started, but has not completed on that
buffer. The buffer state transition diagram is roughly as
follows:
invalid -> dirty -> active -> clean -> dirty.
For each read or write, the driver increments one of the
following: clean hits, dirty hits, allocations, or passes; the
sum of these four counters is represented by the count field.
The hit rate field is computed from the ratio of clean and dirty
2/94 - Intergraph Corporation 1
presto(8) CLIX presto(8)
hits to the total count. Clean hits and dirty hits give the
number of cache hits on clean and dirty buffers. Allocations
means a new buffer had to be allocated for a disk block image.
Passes means that the driver passed the I/O operation directly
through to the actual device driver.
Write cache efficiency is computed from the ratio of write dirty
hits to the number of writes copied into the cache, for example:
efficiency=100*writedirtyhits/(writecount-writepasses)
-u Sets Prestoserve board state to UP and enables acceleration on
selected file systems. If DOWN, then this activates
functionality and resets all statistics and buffers to their
initial values. If special devices (for example,
/dev/dsk/s0u0p7.3) are specified, only those file systems will
have acceleration enabled. If no file systems are specified,
the local, writable file systems listed in both /etc/fstab and
/etc/mnttab will have acceleration enabled. Any file systems
previously accelerated remain accelerated. If was in the ERROR
state, then this command will force to write back any blocks
still in the cache to see if the disk error condition still
exists.
-U Activates the prestoserve board and accelerates the requested
file systems in /etc/presto.conf. Causes Prestoserve to be
brought into the UP state.
-R Sets state to DOWN, invalidates all buffers, and resets the
write and read statistics. Unlike the -d option, -R will
discard dirty buffers that could not be flushed to their
intended disks. This option is useful when dirty data is no
longer useful and must be purged, for example, after a disk head
crash. The -R option takes effect before the -u option.
-v Indicates verbose mode.
DESCRIPTION
The presto command allows users to get information about the driver, and
allows the superuser to control the device. Running the presto command
with no options tells the state of the device (``UP,'' ``DOWN,'' or
``ERROR''), how many bytes of non-volatile memory the cache is using, how
long the cache has been enabled, the write cache efficiency (defined
later), and battery status. All information is printed to standard
output.
When the state is UP, the driver improves I/O performance to accelerated
file systems by caching synchronous disk write operations to the battery
backed up, non-volatile memory. When the state is DOWN, the driver simply
passes all I/O requests through to their intended devices.
2 Intergraph Corporation - 2/94
presto(8) CLIX presto(8)
The state after a machine reboots is DOWN; therefore, the presto command
is usually included in the /etc/init.d/prestosrv file to enable file
system acceleration when the machine reboots.
When the driver detects a disk error during a write back, it enters the
ERROR state. In this state, it effectively disables itself, but continues
to maintain the integrity of cached data that cannot be written back.
Possible disk errors could be the disk drive's being write protected or
offline, a cable problem, or a bad disk block. After remedying the disk
error, use presto -U to verify the fix and to re-enable if there are no
more disk errors. If the disk error is caused by some uncorrectable
problem causing any cached data to no longer be useful (such as a head
crash), refer to the -R option.
By default, all local writable file systems listed in both /etc/fstab and
/etc/mnttab (in case something was mounted and not listed in /etc/fstab)
are accelerated when the Prestoserve board is brought up.
FILES
/dev/presto Generic control device.
/etc/fstab
/etc/mnttab
/etc/presto.conf
Files in which accelerated file systems are specified.
/etc/prestosrv File in which a presto -U command is added to accelerate
file system I/O after a machine reboots.
CAUTIONS
The -l and -L options ignore file systems mounted though a process like
automount (8), since there is no general way to map one of these mount
points to the (possibly non-existent) physical block device on the server.
The -R option will erase cached disk data. Use it only in extreme cases.
RELATED INFORMATION
Files: fstab (4), presto.conf (4)
2/94 - Intergraph Corporation 3