diskx - disk exerciser program
/usr/field/diskx [option(s)] [parameter(s)]
The following is a description of the diskx options: Help.
Displays a help message describing test options and functionality.
Performance test. Read and write transfers
will be timed to measure device throughput. No data validation
is performed as part of this test. Testing will be
done using a range of transfer sizes if the -F option is
not specified. The transfer size used for testing a range
of record sizes will start at the minimum value and be
incremented by the reciprocal of the number of specified
"splits". For example if the number of splits is set to
10, the transfer size will start at the minimum value, the
next transfer size will be the minimum value added to
1/10th of the range of values, similarly the next transfer
size will increase by 1/10th of the range during each
testing interval. If a specific number of transfers are
not specified, the transfer count will be set to allow the
entire partition to be read or written; this number will
vary depending on the transfer size and the partition
size. The performance test will run until completed or
interrupted and is not time limited by the -minutes parameter.
This test may take a long time to complete depending
on the test parameters.
To achieve maximum throughput, specify the -S
option to cause sequential transfers. If the
sequential attribute is not specified, transfers
will be done to random locations which may slow
down the throughput due to associated head seeks on
the device. Read-only test. Reads from the specified
partitions.
This test is useful for generating system I/O
activity. Because it is a read-only test, it is
possible to run more than one instance of the exerciser
on the same disk. Use the -num_xfer option
with the -r option. Write test. The purpose of
this test is to verify that data can be written to
the disk and read back for validation. Seeks are
also done as part of this test. This test provides
the most comprehensive coverage of disk transfer
functionality due to the usage of reads, write, and
seeks. The test also combines sequential and random
access patterns.
The test performs the following operations using a
range of transfer sizes. A single transfer size
will be utilized if the -F attribute is specified.
The first step is to sequentially write the entire
test partition (unless the number of transfers has
been specified using -num_xfer). Next the test
partition is sequentially read. The data read from
disk is examined to insure that it is the same as
what was originally written. At this point if random
transfer testing has not been disabled (using
the -S attribute) then writes will be issued to
random locations on the partition. After completion
of the random writes, reads will be issued to
random locations on the partition. The data read
in from random locations will be examined for
validity.
The following options are testing attributes which modify
how tests are to be run. (Stop on error.) This causes
testing to halt upon detection of all errors except data
validation errors and the accumulated test results will be
displayed. This option is associated with the -p, -r, and
-w tests. (Stop on Data Validation error.) This causes
testing to halt upon detection of a data validation error.
The accumulated test results will be displayed. The failing
LBA and 10 lines of the data expected and the data
returned will also be displayed. This option is associated
with the -w test. Perform fixed size transfers. If
this option is not specified transfers will be done using
random sizes. This attribute is associated with the following
tests: -p, -r, -w. Interactive mode. Under this
mode the user will be prompted for various test parameters.
Typical parameters include the transfer size and
the number of transfers. The following scaling factors are
allowed: k or K -- kilobyte (1024 * n) b or B -- block
(512 * n) m or M -- megabyte (1024 * 1024 * n)
For example 10k would specify 10240 bytes. Perform
the specified test once. If this option is specified
with the sequential transfer attribute, the
test will halt after the entire disk has been read
or written. This attribute is associated with the
-r and -w tests. Do not perform performance analysis
of read transfers. This will cause only write
performance testing to be performed. To perform
only read testing and to skip the write performance
tests the -R attribute must be specified. This
attribute is associated with the following tests:
-p. Opens the disk in read-only mode. This
attribute is associated with all tests. Performs
sequential transfers. In order to achieve maximum
throughput the -S attribute should be specified on
the command line. Transfers will be performed to
sequential disk locations. This may slow down the
observed throughput due to associated head seeks on
the device. If this option is not specified transfers
will be done to random disk locations. This
attribute is associated with the following tests:
-p, -r, -w. Directs output to the terminal. This
attribute is useful when output is directed to a
logfile using the -o option. By also specifying
this parameter after the -o filename options will
cause output to be directed to both the terminal
and the log file. This attribute is associated
with all tests. Does not check for disk labels or
partition overlaps. This option should be used
when testing DRD (Distributed Raw Disk) or LSM
(Logical Storage Manager) devices. See the -f and
-x options also. If any of the selected tests
write to the disk, the disk will be examined for
any existing file systems. If it appears that file
systems exist, the exerciser will prompt for confirmation
before proceeding. When this attribute
is specified the exerciser will NOT prompt for confirmation
before proceeding.
The following options are used to specify test parameters.
These options are followed by an associated parameter
specification. Test parameters may also be modified in an
interactive manner. Refer to the description of the -i
test attribute for details. To specify a numerical value,
type the parameter name followed by a space and then the
number. For example -perf_min 512
The following scaling factors are allowed: k or K -- kilobyte
(1024 * n) b or B -- block (512 * n) m or M --
megabyte (1024 * 1024 * n)
To illustrate this, -perf_min 10K, causes transfers to be
done in sizes of 10240 bytes. Specifies a code to be
imbedded in the high byte of a long word pattern. This
option is only valid with 64 bit patterns, -pattern 8 or
-pattern 99. See also the -pattern option. This parameter
is associated only with the -w test option. Specifies the
level of diagnostic output to display. The higher this
number is, a greater volume of output will be produced
describing the operations the exerciser is performing.
This parameter is associated with all tests. Specifies
the maximum number of error messages may be produced as a
result of an individual test. Limits on error output is
done to prevent a flooding of diagnostic messages in the
event of persistent errors. This parameter is associated
with all tests. Specifies which device special file to
test. The devname parameter is the name associated with
either a block or character special file which represents
the disk to be tested. The device special files are
located in /dev/disk and named dskNx, where N is the number
of the device and x is a letter representing the partition.
For example, /dev/disk/dsk0a refers to partition
a of disk 0. If no partition is specified, it is assumed
that testing is to be done to all partitions. For example
if the specified devname is /dev/disk/dsk0 then testing
will be done to all partitions. For raw device names, you
must specify the device special files in the /dev/rdisk
directory, which have the same dskNx format. For example,
/dev/rdisk/dsk0a. This parameter must be specified and is
associated with all tests.
On a cluster system using DRD devices, or when
using LSM, use the -x option with the -f option to
specify the pathname. You must also use the -X
option and specify a value for the the -numb_blocks
option when using -x. (See the EXAMPLES section).
Specifies how many minutes to allow testing to continue.
This parameter is associated with the following
tests: -r, -w. Specifies the maximum transfer
size to be performed. When transfers are to be
done using random sizes, the sizes will be within
the range specified by the -max_xfer and -min_xfer
parameters. If fixed size transfers are specified
(see the -F test attribute) then transfers will be
done in a size specified by the -min_xfer parameter.
Transfer sizes to the character special file
should be specified in multiples of 512 bytes. If
the specified transfer size is not an even multiple
the value will be rounded down to the nearest 512
bytes. This parameter is associated with the following
tests: -r, -w. Specifies the minimum transfer
size to be performed. This parameter is associated
with the following tests: -r, -w. Specifies
the range of disk blocks that diskx should use.
Use this option when testing DRD devices or when
using LSM. Specifies how many transfers to perform
before changing the partition that is currently
being tested. This parameter is only useful when
more than one partition is being tested. If this
parameter is not specified then the number of
transfers will be set to be enough to completely
cover a partition. This parameter is associated
with the following tests: -r, -w. Sends output to
the specified filename. The default is to not create
an output file and send output to the terminal.
This parameter is associated with all tests. This
option specifies the write pattern to be used when
exercising a disk. 1 indicates a byte pattern (8
bits) 2 indicates a short word pattern (16 bits) 4
indicates a int word pattern (32 bits) 8 indicates
a long word pattern (64 bits) 99 indicates a
repeating pattern (64 bits)
The default is a long word pattern (64 bits). This
parameter is associated only with the -w test
option. See also the -code option. Specifies the
maximum transfer size to be performed. When transfers
are to be done using random sizes, the sizes
will be within the range specified by the -perf_min
and -perf_max parameters. If fixed size transfers
are specified (see the -F test attribute) then
transfers will be done in a size specified by the
-perf_min parameter. This parameter is associated
with the following tests: -p. Specifies the minimum
transfer size to be performed. This parameter
is associated with the following tests: -p. Specifies
how the transfer size will be changed when
testing a range of transfer sizes. The range of
transfer sizes is divided by perf_splits to obtain
a transfer size increment. For example if
perf_splits is set to 10, tests will be run by
starting with the minimum transfer size and
increasing the size by 1/10th of the range of values
for each test iteration. The last transfer
size will be set to the specified maximum transfer
size. This parameter is associated with the following
tests: -p. Specifies the number of transfers
to be performed in performance analysis. If this
value is not explicitly specified the number of
transfers will be set equal to the number required
to read the entire partition. This parameter is
associated with the following tests: -p. Specifies
not to apply partition logic & checks. This flag
is most likely to be used when pseudo device
drivers such as DRD (Distributed Raw Disk) or LSM
(Logical Storage Manager) are being used and the
device type has no partitions. In such cases, the
device special file as specified on the command
line is used rather than a constructed name that
appends a letter signifying the partition. When
using -x with the read, write, and performance
tests you must specify the number of blocks on the
device using the -num_blocks option. (Without any
partition logic it is impossible for disks to automatically
size the partition.)
The diskx program is intended to test various aspects of
disk driver functionality. These tests provide more comprehensive
functional coverage than the dskx utility.
Main functional areas which are tested include: Read testing.
Write testing. Seek testing. Performance analysis.
Some of the tests involve writing to the disk. For this
reason the exerciser should be used with caution on disks
which may contain useful data which could be over-written.
Tests which write to the disk will first check for the
existence of file systems on the test partitions and partitions
which overlap the test partitions. If a file system
is found on these partitions, a prompt will appear
asking if testing should continue.
There are a number of options that diskx accepts. These
options control which tests are performed and also specify
parameters of test operation.
The following are descriptions of some of the tables which
are produced by the disk exerciser.
This is the header used to describe the results of the
transfer tests followed by a description of each column:
Disk Transfer Statistics
Part Seeks Seek_Er Writes Writ_Er MB_Write Reads Read_Er
MB_Read Data_Er
A letter used to represent the disk partition. The number
of seek system calls that were issued. The number of seek
system calls returning error status. The number of write
system calls issued. The number of write system calls
returning error status. The number of megabytes of data
written. The number of read system calls issued. The
number of read system calls returning error status. The
number of megabytes of data read. The number of transfers
which had data validation errors. This does not specify
the number of bytes that were in error; rather it specifies
that the transfer had at least one byte in error.
This header is used to describe the results of the performance
tests.
Performance test results:
Part- Transfer Count of Read Write
Transfer ition Size Transfers Rate
Rate Errors
A letter used to represent the disk partition. This is
the size of the read or write system call. Specifies the
number of read or write system calls. The observed disk
throughput obtained by timing the specified number of read
system calls. The observed disk throughput obtained by
timing the specified number of write system calls. Shows
how many read and write system calls resulted in a return
value of error status. If this value is nonzero then the
values displayed in the Read and Write Rate columns may
not accurately reflect correct device performance.
The following are example command lines with a description
of what the resulting test action will be. The example
below will perform read-only testing on the character
device special file that dsk0 represents. Since no partition
is specified, reading will be done from all partitions.
The default range of transfer sizes will be used.
Output from the exerciser program will be displayed on the
terminal. diskx -f /dev/rdisk/dsk0 -r The following example
performs read-only testing on an LSM volume. diskx -x
-X -f /dev/vol/vol1 -r -num_blocks 100000 Performance
tests will be run in the "a" partition of dsk0. Program
output will be logged to the file diskx.out. The -S option
will cause sequential transfers for best test results.
Testing will be done over the default range of transfer
sizes. diskx -f /dev/disk/dsk0a -o diskx.out -p -n -S
This command will run the read test on all partitions of
the disks. The disk exerciser is being invoked here as 3
separate processes. In this manner the 3 processes will
be generating a lot of system I/O activity. This may be
useful for system stress testing purposes. diskx -f
/dev/rdisk/dsk0 -r &; diskx -f /dev/rdisk/dsk1 -r \ &;
diskx -f /dev/rdisk/dsk2 -r & This command will run the
write test on partition "b" of disk 17. diskx -f
/dev/rdisk/dsk17b -w -pattern 99
The repeating long word pattern will write the following
test pattern to the disk: resulting test
pattern written: <address> <quad word> <quad
word> ... 0000020 0007060504030201
0007060504030201 0000030 0007060504030201
0007060504030201 0000040 0007060504030201
0007060504030201 ... This command will run the
write test on partition "b" of disk 17. diskx -f
/dev/rdisk/dsk17b -w -pattern 8 -code 255
The specified pattern and code will write the following
test pattern to the disk: resulting test
pattern written: <address> <quad word> <quad
word> .... 0000020 ff00000000000004
ff00000000000005 0000030 ff00000000000006
ff00000000000007 0000040 ff00000000000008
ff00000000000009 ...
Commands: cmx(8), fsx(8), memx(8), shmx(8), tapex(8)
Functions: getdiskbyname(3)
Files: disktab(4)
Cluster Administration
diskx(8)
[ Back ] |
