iolog: Added option read_iolog_chunked. Used to avoid reading large iologs at once.
[fio.git] / fio.1
CommitLineData
523bad63 1.TH fio 1 "August 2017" "User Manual"
d60e92d1
AC
2.SH NAME
3fio \- flexible I/O tester
4.SH SYNOPSIS
5.B fio
6[\fIoptions\fR] [\fIjobfile\fR]...
7.SH DESCRIPTION
8.B fio
9is a tool that will spawn a number of threads or processes doing a
10particular type of I/O action as specified by the user.
11The typical use of fio is to write a job file matching the I/O load
12one wants to simulate.
13.SH OPTIONS
14.TP
49da1240 15.BI \-\-debug \fR=\fPtype
7db7a5a0
TK
16Enable verbose tracing \fItype\fR of various fio actions. May be `all' for all \fItype\fRs
17or individual types separated by a comma (e.g. `\-\-debug=file,mem' will enable
bdd88be3
TK
18file and memory debugging). `help' will list all available tracing options.
19.TP
7db7a5a0 20.BI \-\-parse\-only
bdd88be3 21Parse options only, don't start any I/O.
49da1240 22.TP
d60e92d1
AC
23.BI \-\-output \fR=\fPfilename
24Write output to \fIfilename\fR.
25.TP
7db7a5a0
TK
26.BI \-\-output\-format \fR=\fPformat
27Set the reporting \fIformat\fR to `normal', `terse', `json', or
28`json+'. Multiple formats can be selected, separate by a comma. `terse'
29is a CSV based format. `json+' is like `json', except it adds a full
513e37ee 30dump of the latency buckets.
e28ee21d 31.TP
7db7a5a0 32.BI \-\-bandwidth\-log
d23ae827 33Generate aggregate bandwidth logs.
d60e92d1 34.TP
7db7a5a0
TK
35.BI \-\-minimal
36Print statistics in a terse, semicolon\-delimited format.
d60e92d1 37.TP
7db7a5a0
TK
38.BI \-\-append\-terse
39Print statistics in selected mode AND terse, semicolon\-delimited format.
40\fBDeprecated\fR, use \fB\-\-output\-format\fR instead to select multiple formats.
f6a7df53 41.TP
065248bf 42.BI \-\-terse\-version \fR=\fPversion
7db7a5a0 43Set terse \fIversion\fR output format (default `3', or `2', `4', `5').
49da1240 44.TP
7db7a5a0 45.BI \-\-version
bdd88be3
TK
46Print version information and exit.
47.TP
7db7a5a0 48.BI \-\-help
bdd88be3 49Print a summary of the command line options and exit.
49da1240 50.TP
7db7a5a0 51.BI \-\-cpuclock\-test
bdd88be3 52Perform test and validation of internal CPU clock.
fec0f21c 53.TP
bdd88be3 54.BI \-\-crctest \fR=\fP[test]
7db7a5a0 55Test the speed of the built\-in checksumming functions. If no argument is given,
bdd88be3 56all of them are tested. Alternatively, a comma separated list can be passed, in which
fec0f21c
JA
57case the given ones are tested.
58.TP
49da1240 59.BI \-\-cmdhelp \fR=\fPcommand
bdd88be3 60Print help information for \fIcommand\fR. May be `all' for all commands.
49da1240 61.TP
7db7a5a0
TK
62.BI \-\-enghelp \fR=\fP[ioengine[,command]]
63List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR
64defined by \fIioengine\fR. If no \fIioengine\fR is given, list all
65available ioengines.
de890a1e 66.TP
d60e92d1 67.BI \-\-showcmd \fR=\fPjobfile
7db7a5a0 68Convert \fIjobfile\fR to a set of command\-line options.
d60e92d1 69.TP
bdd88be3 70.BI \-\-readonly
4027b2a1 71Turn on safety read\-only checks, preventing writes and trims. The \fB\-\-readonly\fR
bdd88be3 72option is an extra safety guard to prevent users from accidentally starting
4027b2a1
VF
73a write or trim workload when that is not desired. Fio will only modify the
74device under test if `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite'
75is given. This safety net can be used as an extra precaution.
bdd88be3 76.TP
d60e92d1 77.BI \-\-eta \fR=\fPwhen
7db7a5a0 78Specifies when real\-time ETA estimate should be printed. \fIwhen\fR may
db37d890
JA
79be `always', `never' or `auto'. `auto' is the default, it prints ETA when
80requested if the output is a TTY. `always' disregards the output type, and
81prints ETA when requested. `never' never prints ETA.
82.TP
83.BI \-\-eta\-interval \fR=\fPtime
84By default, fio requests client ETA status roughly every second. With this
85option, the interval is configurable. Fio imposes a minimum allowed time to
86avoid flooding the console, less than 250 msec is not supported.
d60e92d1 87.TP
30b5d57f 88.BI \-\-eta\-newline \fR=\fPtime
bdd88be3
TK
89Force a new line for every \fItime\fR period passed. When the unit is omitted,
90the value is interpreted in seconds.
30b5d57f
JA
91.TP
92.BI \-\-status\-interval \fR=\fPtime
aa6cb459
VF
93Force a full status dump of cumulative (from job start) values at \fItime\fR
94intervals. This option does *not* provide per-period measurements. So
95values such as bandwidth are running averages. When the time unit is omitted,
96\fItime\fR is interpreted in seconds.
bdd88be3
TK
97.TP
98.BI \-\-section \fR=\fPname
99Only run specified section \fIname\fR in job file. Multiple sections can be specified.
7db7a5a0 100The \fB\-\-section\fR option allows one to combine related jobs into one file.
bdd88be3 101E.g. one job file could define light, moderate, and heavy sections. Tell
7db7a5a0 102fio to run only the "heavy" section by giving `\-\-section=heavy'
bdd88be3 103command line option. One can also specify the "write" operations in one
7db7a5a0 104section and "verify" operation in another section. The \fB\-\-section\fR option
bdd88be3
TK
105only applies to job sections. The reserved *global* section is always
106parsed and used.
c0a5d35e 107.TP
49da1240 108.BI \-\-alloc\-size \fR=\fPkb
7db7a5a0
TK
109Set the internal smalloc pool size to \fIkb\fR in KiB. The
110\fB\-\-alloc\-size\fR switch allows one to use a larger pool size for smalloc.
bdd88be3
TK
111If running large jobs with randommap enabled, fio can run out of memory.
112Smalloc is an internal allocator for shared structures from a fixed size
113memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
7db7a5a0
TK
114NOTE: While running `.fio_smalloc.*' backing store files are visible
115in `/tmp'.
d60e92d1 116.TP
49da1240
JA
117.BI \-\-warnings\-fatal
118All fio parser warnings are fatal, causing fio to exit with an error.
9183788d 119.TP
49da1240 120.BI \-\-max\-jobs \fR=\fPnr
7db7a5a0 121Set the maximum number of threads/processes to support to \fInr\fR.
7f4811bb
RNS
122NOTE: On Linux, it may be necessary to increase the shared-memory limit
123(`/proc/sys/kernel/shmmax') if fio runs into errors while creating jobs.
d60e92d1 124.TP
49da1240 125.BI \-\-server \fR=\fPargs
7db7a5a0
TK
126Start a backend server, with \fIargs\fR specifying what to listen to.
127See \fBCLIENT/SERVER\fR section.
f57a9c59 128.TP
49da1240 129.BI \-\-daemonize \fR=\fPpidfile
7db7a5a0 130Background a fio server, writing the pid to the given \fIpidfile\fR file.
49da1240 131.TP
bdd88be3 132.BI \-\-client \fR=\fPhostname
7db7a5a0
TK
133Instead of running the jobs locally, send and run them on the given \fIhostname\fR
134or set of \fIhostname\fRs. See \fBCLIENT/SERVER\fR section.
bdd88be3 135.TP
7db7a5a0
TK
136.BI \-\-remote\-config \fR=\fPfile
137Tell fio server to load this local \fIfile\fR.
f2a2ce0e
HL
138.TP
139.BI \-\-idle\-prof \fR=\fPoption
7db7a5a0 140Report CPU idleness. \fIoption\fR is one of the following:
bdd88be3
TK
141.RS
142.RS
143.TP
144.B calibrate
145Run unit work calibration only and exit.
146.TP
147.B system
148Show aggregate system idleness and unit work.
149.TP
150.B percpu
7db7a5a0 151As \fBsystem\fR but also show per CPU idleness.
bdd88be3
TK
152.RE
153.RE
154.TP
7db7a5a0
TK
155.BI \-\-inflate\-log \fR=\fPlog
156Inflate and output compressed \fIlog\fR.
bdd88be3 157.TP
7db7a5a0
TK
158.BI \-\-trigger\-file \fR=\fPfile
159Execute trigger command when \fIfile\fR exists.
bdd88be3 160.TP
7db7a5a0
TK
161.BI \-\-trigger\-timeout \fR=\fPtime
162Execute trigger at this \fItime\fR.
bdd88be3 163.TP
7db7a5a0
TK
164.BI \-\-trigger \fR=\fPcommand
165Set this \fIcommand\fR as local trigger.
bdd88be3 166.TP
7db7a5a0
TK
167.BI \-\-trigger\-remote \fR=\fPcommand
168Set this \fIcommand\fR as remote trigger.
bdd88be3 169.TP
7db7a5a0
TK
170.BI \-\-aux\-path \fR=\fPpath
171Use this \fIpath\fR for fio state generated files.
d60e92d1 172.SH "JOB FILE FORMAT"
7a14cf18
TK
173Any parameters following the options will be assumed to be job files, unless
174they match a job file parameter. Multiple job files can be listed and each job
7db7a5a0 175file will be regarded as a separate group. Fio will \fBstonewall\fR execution
7a14cf18
TK
176between each group.
177
178Fio accepts one or more job files describing what it is
179supposed to do. The job file format is the classic ini file, where the names
180enclosed in [] brackets define the job name. You are free to use any ASCII name
181you want, except *global* which has special meaning. Following the job name is
182a sequence of zero or more parameters, one per line, that define the behavior of
183the job. If the first character in a line is a ';' or a '#', the entire line is
184discarded as a comment.
185
186A *global* section sets defaults for the jobs described in that file. A job may
187override a *global* section parameter, and a job file may even have several
188*global* sections if so desired. A job is only affected by a *global* section
189residing above it.
190
7db7a5a0
TK
191The \fB\-\-cmdhelp\fR option also lists all options. If used with an \fIcommand\fR
192argument, \fB\-\-cmdhelp\fR will detail the given \fIcommand\fR.
7a14cf18 193
7db7a5a0
TK
194See the `examples/' directory for inspiration on how to write job files. Note
195the copyright and license requirements currently apply to
196`examples/' files.
54eb4569
TK
197.SH "JOB FILE PARAMETERS"
198Some parameters take an option of a given type, such as an integer or a
199string. Anywhere a numeric value is required, an arithmetic expression may be
200used, provided it is surrounded by parentheses. Supported operators are:
d59aa780 201.RS
7db7a5a0 202.P
d59aa780 203.B addition (+)
7db7a5a0
TK
204.P
205.B subtraction (\-)
206.P
d59aa780 207.B multiplication (*)
7db7a5a0 208.P
d59aa780 209.B division (/)
7db7a5a0 210.P
d59aa780 211.B modulus (%)
7db7a5a0 212.P
d59aa780
JA
213.B exponentiation (^)
214.RE
d59aa780
JA
215.P
216For time values in expressions, units are microseconds by default. This is
217different than for time values not in expressions (not enclosed in
54eb4569
TK
218parentheses).
219.SH "PARAMETER TYPES"
220The following parameter types are used.
d60e92d1
AC
221.TP
222.I str
6b86fc18
TK
223String. A sequence of alphanumeric characters.
224.TP
225.I time
226Integer with possible time suffix. Without a unit value is interpreted as
227seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
228hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and 'us'
229(or 'usec') for microseconds. For example, use 10m for 10 minutes.
d60e92d1
AC
230.TP
231.I int
6d500c2e
RE
232Integer. A whole number value, which may contain an integer prefix
233and an integer suffix.
0b43a833
TK
234.RS
235.RS
236.P
6b86fc18 237[*integer prefix*] **number** [*integer suffix*]
0b43a833
TK
238.RE
239.P
6b86fc18
TK
240The optional *integer prefix* specifies the number's base. The default
241is decimal. *0x* specifies hexadecimal.
0b43a833 242.P
6b86fc18
TK
243The optional *integer suffix* specifies the number's units, and includes an
244optional unit prefix and an optional unit. For quantities of data, the
245default unit is bytes. For quantities of time, the default unit is seconds
246unless otherwise specified.
0b43a833
TK
247.P
248With `kb_base=1000', fio follows international standards for unit
7db7a5a0 249prefixes. To specify power\-of\-10 decimal values defined in the
6b86fc18 250International System of Units (SI):
0b43a833
TK
251.RS
252.P
7db7a5a0 253.PD 0
eccce61a 254K means kilo (K) or 1000
7db7a5a0 255.P
eccce61a 256M means mega (M) or 1000**2
7db7a5a0 257.P
eccce61a 258G means giga (G) or 1000**3
7db7a5a0 259.P
eccce61a 260T means tera (T) or 1000**4
7db7a5a0 261.P
eccce61a 262P means peta (P) or 1000**5
7db7a5a0 263.PD
0b43a833
TK
264.RE
265.P
7db7a5a0 266To specify power\-of\-2 binary values defined in IEC 80000\-13:
0b43a833
TK
267.RS
268.P
7db7a5a0 269.PD 0
eccce61a 270Ki means kibi (Ki) or 1024
7db7a5a0 271.P
eccce61a 272Mi means mebi (Mi) or 1024**2
7db7a5a0 273.P
eccce61a 274Gi means gibi (Gi) or 1024**3
7db7a5a0 275.P
eccce61a 276Ti means tebi (Ti) or 1024**4
7db7a5a0 277.P
eccce61a 278Pi means pebi (Pi) or 1024**5
7db7a5a0 279.PD
0b43a833
TK
280.RE
281.P
282With `kb_base=1024' (the default), the unit prefixes are opposite
7db7a5a0 283from those specified in the SI and IEC 80000\-13 standards to provide
6b86fc18 284compatibility with old scripts. For example, 4k means 4096.
0b43a833 285.P
6b86fc18
TK
286For quantities of data, an optional unit of 'B' may be included
287(e.g., 'kB' is the same as 'k').
0b43a833 288.P
6b86fc18
TK
289The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
290not milli). 'b' and 'B' both mean byte, not bit.
0b43a833
TK
291.P
292Examples with `kb_base=1000':
293.RS
294.P
7db7a5a0 295.PD 0
6d500c2e 2964 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
7db7a5a0 297.P
6d500c2e 2981 MiB: 1048576, 1m, 1024k
7db7a5a0 299.P
6d500c2e 3001 MB: 1000000, 1mi, 1000ki
7db7a5a0 301.P
6d500c2e 3021 TiB: 1073741824, 1t, 1024m, 1048576k
7db7a5a0 303.P
6d500c2e 3041 TB: 1000000000, 1ti, 1000mi, 1000000ki
7db7a5a0 305.PD
0b43a833
TK
306.RE
307.P
308Examples with `kb_base=1024' (default):
309.RS
310.P
7db7a5a0 311.PD 0
6d500c2e 3124 KiB: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
7db7a5a0 313.P
6d500c2e 3141 MiB: 1048576, 1m, 1024k
7db7a5a0 315.P
6d500c2e 3161 MB: 1000000, 1mi, 1000ki
7db7a5a0 317.P
6d500c2e 3181 TiB: 1073741824, 1t, 1024m, 1048576k
7db7a5a0 319.P
6d500c2e 3201 TB: 1000000000, 1ti, 1000mi, 1000000ki
7db7a5a0 321.PD
0b43a833
TK
322.RE
323.P
6d500c2e 324To specify times (units are not case sensitive):
0b43a833
TK
325.RS
326.P
7db7a5a0 327.PD 0
6d500c2e 328D means days
7db7a5a0 329.P
6d500c2e 330H means hours
7db7a5a0 331.P
6d500c2e 332M mean minutes
7db7a5a0 333.P
6d500c2e 334s or sec means seconds (default)
7db7a5a0 335.P
6d500c2e 336ms or msec means milliseconds
7db7a5a0 337.P
6d500c2e 338us or usec means microseconds
7db7a5a0 339.PD
0b43a833
TK
340.RE
341.P
6b86fc18 342If the option accepts an upper and lower range, use a colon ':' or
7db7a5a0 343minus '\-' to separate such values. See \fIirange\fR parameter type.
6b86fc18
TK
344If the lower value specified happens to be larger than the upper value
345the two values are swapped.
0b43a833 346.RE
d60e92d1
AC
347.TP
348.I bool
6b86fc18
TK
349Boolean. Usually parsed as an integer, however only defined for
350true and false (1 and 0).
d60e92d1
AC
351.TP
352.I irange
6b86fc18 353Integer range with suffix. Allows value range to be given, such as
7db7a5a0 3541024\-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
6b86fc18 355option allows two sets of ranges, they can be specified with a ',' or '/'
7db7a5a0 356delimiter: 1k\-4k/8k\-32k. Also see \fIint\fR parameter type.
83349190
YH
357.TP
358.I float_list
6b86fc18 359A list of floating point numbers, separated by a ':' character.
523bad63 360.SH "JOB PARAMETERS"
54eb4569 361With the above in mind, here follows the complete list of fio job parameters.
523bad63 362.SS "Units"
d60e92d1 363.TP
523bad63
TK
364.BI kb_base \fR=\fPint
365Select the interpretation of unit prefixes in input parameters.
366.RS
367.RS
d60e92d1 368.TP
523bad63
TK
369.B 1000
370Inputs comply with IEC 80000\-13 and the International
371System of Units (SI). Use:
372.RS
373.P
374.PD 0
375\- power\-of\-2 values with IEC prefixes (e.g., KiB)
376.P
377\- power\-of\-10 values with SI prefixes (e.g., kB)
378.PD
379.RE
380.TP
381.B 1024
382Compatibility mode (default). To avoid breaking old scripts:
383.P
384.RS
385.PD 0
386\- power\-of\-2 values with SI prefixes
387.P
388\- power\-of\-10 values with IEC prefixes
389.PD
390.RE
391.RE
392.P
393See \fBbs\fR for more details on input parameters.
394.P
395Outputs always use correct prefixes. Most outputs include both
396side\-by\-side, like:
397.P
398.RS
399bw=2383.3kB/s (2327.4KiB/s)
400.RE
401.P
402If only one value is reported, then kb_base selects the one to use:
403.P
404.RS
405.PD 0
4061000 \-\- SI prefixes
407.P
4081024 \-\- IEC prefixes
409.PD
410.RE
411.RE
412.TP
413.BI unit_base \fR=\fPint
414Base unit for reporting. Allowed values are:
415.RS
416.RS
417.TP
418.B 0
419Use auto\-detection (default).
420.TP
421.B 8
422Byte based.
423.TP
424.B 1
425Bit based.
426.RE
427.RE
428.SS "Job description"
429.TP
430.BI name \fR=\fPstr
431ASCII name of the job. This may be used to override the name printed by fio
432for this job. Otherwise the job name is used. On the command line this
433parameter has the special purpose of also signaling the start of a new job.
9cc8cb91 434.TP
d60e92d1 435.BI description \fR=\fPstr
523bad63
TK
436Text description of the job. Doesn't do anything except dump this text
437description when this job is run. It's not parsed.
438.TP
439.BI loops \fR=\fPint
440Run the specified number of iterations of this job. Used to repeat the same
441workload a given number of times. Defaults to 1.
442.TP
443.BI numjobs \fR=\fPint
444Create the specified number of clones of this job. Each clone of job
445is spawned as an independent thread or process. May be used to setup a
446larger number of threads/processes doing the same thing. Each thread is
447reported separately; to see statistics for all clones as a whole, use
448\fBgroup_reporting\fR in conjunction with \fBnew_group\fR.
449See \fB\-\-max\-jobs\fR. Default: 1.
450.SS "Time related parameters"
451.TP
452.BI runtime \fR=\fPtime
453Tell fio to terminate processing after the specified period of time. It
454can be quite hard to determine for how long a specified job will run, so
455this parameter is handy to cap the total runtime to a given time. When
f1dd3fb1 456the unit is omitted, the value is interpreted in seconds.
523bad63
TK
457.TP
458.BI time_based
459If set, fio will run for the duration of the \fBruntime\fR specified
460even if the file(s) are completely read or written. It will simply loop over
461the same workload as many times as the \fBruntime\fR allows.
462.TP
463.BI startdelay \fR=\fPirange(int)
464Delay the start of job for the specified amount of time. Can be a single
465value or a range. When given as a range, each thread will choose a value
466randomly from within the range. Value is in seconds if a unit is omitted.
467.TP
468.BI ramp_time \fR=\fPtime
469If set, fio will run the specified workload for this amount of time before
470logging any performance numbers. Useful for letting performance settle
471before logging results, thus minimizing the runtime required for stable
472results. Note that the \fBramp_time\fR is considered lead in time for a job,
473thus it will increase the total runtime if a special timeout or
474\fBruntime\fR is specified. When the unit is omitted, the value is
475given in seconds.
476.TP
477.BI clocksource \fR=\fPstr
478Use the given clocksource as the base of timing. The supported options are:
479.RS
480.RS
481.TP
482.B gettimeofday
483\fBgettimeofday\fR\|(2)
484.TP
485.B clock_gettime
486\fBclock_gettime\fR\|(2)
487.TP
488.B cpu
489Internal CPU clock source
490.RE
491.P
492\fBcpu\fR is the preferred clocksource if it is reliable, as it is very fast (and
493fio is heavy on time calls). Fio will automatically use this clocksource if
494it's supported and considered reliable on the system it is running on,
495unless another clocksource is specifically set. For x86/x86\-64 CPUs, this
496means supporting TSC Invariant.
497.RE
498.TP
499.BI gtod_reduce \fR=\fPbool
500Enable all of the \fBgettimeofday\fR\|(2) reducing options
501(\fBdisable_clat\fR, \fBdisable_slat\fR, \fBdisable_bw_measurement\fR) plus
502reduce precision of the timeout somewhat to really shrink the
503\fBgettimeofday\fR\|(2) call count. With this option enabled, we only do
504about 0.4% of the \fBgettimeofday\fR\|(2) calls we would have done if all
505time keeping was enabled.
506.TP
507.BI gtod_cpu \fR=\fPint
508Sometimes it's cheaper to dedicate a single thread of execution to just
509getting the current time. Fio (and databases, for instance) are very
510intensive on \fBgettimeofday\fR\|(2) calls. With this option, you can set
511one CPU aside for doing nothing but logging current time to a shared memory
512location. Then the other threads/processes that run I/O workloads need only
513copy that segment, instead of entering the kernel with a
514\fBgettimeofday\fR\|(2) call. The CPU set aside for doing these time
515calls will be excluded from other uses. Fio will manually clear it from the
516CPU mask of other jobs.
517.SS "Target file/device"
d60e92d1
AC
518.TP
519.BI directory \fR=\fPstr
523bad63
TK
520Prefix \fBfilename\fRs with this directory. Used to place files in a different
521location than `./'. You can specify a number of directories by
522separating the names with a ':' character. These directories will be
523assigned equally distributed to job clones created by \fBnumjobs\fR as
524long as they are using generated filenames. If specific \fBfilename\fR(s) are
525set fio will use the first listed directory, and thereby matching the
526\fBfilename\fR semantic which generates a file each clone if not specified, but
527let all clones use the same if set.
528.RS
529.P
530See the \fBfilename\fR option for information on how to escape ':' and '\'
531characters within the directory path itself.
532.RE
d60e92d1
AC
533.TP
534.BI filename \fR=\fPstr
523bad63
TK
535Fio normally makes up a \fBfilename\fR based on the job name, thread number, and
536file number (see \fBfilename_format\fR). If you want to share files
537between threads in a job or several
538jobs with fixed file paths, specify a \fBfilename\fR for each of them to override
539the default. If the ioengine is file based, you can specify a number of files
540by separating the names with a ':' colon. So if you wanted a job to open
541`/dev/sda' and `/dev/sdb' as the two working files, you would use
542`filename=/dev/sda:/dev/sdb'. This also means that whenever this option is
543specified, \fBnrfiles\fR is ignored. The size of regular files specified
544by this option will be \fBsize\fR divided by number of files unless an
545explicit size is specified by \fBfilesize\fR.
546.RS
547.P
548Each colon and backslash in the wanted path must be escaped with a '\'
549character. For instance, if the path is `/dev/dsk/foo@3,0:c' then you
550would use `filename=/dev/dsk/foo@3,0\\:c' and if the path is
551`F:\\\\filename' then you would use `filename=F\\:\\\\filename'.
552.P
553On Windows, disk devices are accessed as `\\\\\\\\.\\\\PhysicalDrive0' for
554the first device, `\\\\\\\\.\\\\PhysicalDrive1' for the second etc.
555Note: Windows and FreeBSD prevent write access to areas
556of the disk containing in\-use data (e.g. filesystems).
557.P
558The filename `\-' is a reserved name, meaning *stdin* or *stdout*. Which
559of the two depends on the read/write direction set.
560.RE
d60e92d1 561.TP
de98bd30 562.BI filename_format \fR=\fPstr
523bad63
TK
563If sharing multiple files between jobs, it is usually necessary to have fio
564generate the exact names that you want. By default, fio will name a file
de98bd30 565based on the default file format specification of
523bad63 566`jobname.jobnumber.filenumber'. With this option, that can be
de98bd30
JA
567customized. Fio will recognize and replace the following keywords in this
568string:
569.RS
570.RS
571.TP
572.B $jobname
573The name of the worker thread or process.
574.TP
575.B $jobnum
576The incremental number of the worker thread or process.
577.TP
578.B $filenum
579The incremental number of the file for that worker thread or process.
580.RE
581.P
523bad63
TK
582To have dependent jobs share a set of files, this option can be set to have
583fio generate filenames that are shared between the two. For instance, if
584`testfiles.$filenum' is specified, file number 4 for any job will be
585named `testfiles.4'. The default of `$jobname.$jobnum.$filenum'
de98bd30 586will be used if no other format specifier is given.
645943c0
JB
587.P
588If you specify a path then the directories will be created up to the main
589directory for the file. So for example if you specify `a/b/c/$jobnum` then the
590directories a/b/c will be created before the file setup part of the job. If you
591specify \fBdirectory\fR then the path will be relative that directory, otherwise
592it is treated as the absolute path.
de98bd30 593.RE
de98bd30 594.TP
922a5be8 595.BI unique_filename \fR=\fPbool
523bad63
TK
596To avoid collisions between networked clients, fio defaults to prefixing any
597generated filenames (with a directory specified) with the source of the
598client connecting. To disable this behavior, set this option to 0.
599.TP
600.BI opendir \fR=\fPstr
601Recursively open any files below directory \fIstr\fR.
922a5be8 602.TP
3ce9dcaf 603.BI lockfile \fR=\fPstr
523bad63
TK
604Fio defaults to not locking any files before it does I/O to them. If a file
605or file descriptor is shared, fio can serialize I/O to that file to make the
606end result consistent. This is usual for emulating real workloads that share
607files. The lock modes are:
3ce9dcaf
JA
608.RS
609.RS
610.TP
611.B none
523bad63 612No locking. The default.
3ce9dcaf
JA
613.TP
614.B exclusive
523bad63 615Only one thread or process may do I/O at a time, excluding all others.
3ce9dcaf
JA
616.TP
617.B readwrite
523bad63
TK
618Read\-write locking on the file. Many readers may
619access the file at the same time, but writes get exclusive access.
3ce9dcaf 620.RE
ce594fbe 621.RE
523bad63
TK
622.TP
623.BI nrfiles \fR=\fPint
624Number of files to use for this job. Defaults to 1. The size of files
625will be \fBsize\fR divided by this unless explicit size is specified by
626\fBfilesize\fR. Files are created for each thread separately, and each
627file will have a file number within its name by default, as explained in
628\fBfilename\fR section.
629.TP
630.BI openfiles \fR=\fPint
631Number of files to keep open at the same time. Defaults to the same as
632\fBnrfiles\fR, can be set smaller to limit the number simultaneous
633opens.
634.TP
635.BI file_service_type \fR=\fPstr
636Defines how fio decides which file from a job to service next. The following
637types are defined:
638.RS
639.RS
640.TP
641.B random
642Choose a file at random.
643.TP
644.B roundrobin
645Round robin over opened files. This is the default.
646.TP
647.B sequential
648Finish one file before moving on to the next. Multiple files can
649still be open depending on \fBopenfiles\fR.
650.TP
651.B zipf
652Use a Zipf distribution to decide what file to access.
653.TP
654.B pareto
655Use a Pareto distribution to decide what file to access.
656.TP
657.B normal
658Use a Gaussian (normal) distribution to decide what file to access.
659.TP
660.B gauss
661Alias for normal.
662.RE
3ce9dcaf 663.P
523bad63
TK
664For \fBrandom\fR, \fBroundrobin\fR, and \fBsequential\fR, a postfix can be appended to
665tell fio how many I/Os to issue before switching to a new file. For example,
666specifying `file_service_type=random:8' would cause fio to issue
6678 I/Os before selecting a new file at random. For the non\-uniform
668distributions, a floating point postfix can be given to influence how the
669distribution is skewed. See \fBrandom_distribution\fR for a description
670of how that would work.
671.RE
672.TP
673.BI ioscheduler \fR=\fPstr
674Attempt to switch the device hosting the file to the specified I/O scheduler
675before running.
676.TP
677.BI create_serialize \fR=\fPbool
678If true, serialize the file creation for the jobs. This may be handy to
679avoid interleaving of data files, which may greatly depend on the filesystem
680used and even the number of processors in the system. Default: true.
681.TP
682.BI create_fsync \fR=\fPbool
683\fBfsync\fR\|(2) the data file after creation. This is the default.
684.TP
685.BI create_on_open \fR=\fPbool
686If true, don't pre\-create files but allow the job's open() to create a file
687when it's time to do I/O. Default: false \-\- pre\-create all necessary files
688when the job starts.
689.TP
690.BI create_only \fR=\fPbool
691If true, fio will only run the setup phase of the job. If files need to be
692laid out or updated on disk, only that will be done \-\- the actual job contents
693are not executed. Default: false.
694.TP
695.BI allow_file_create \fR=\fPbool
696If true, fio is permitted to create files as part of its workload. If this
697option is false, then fio will error out if
698the files it needs to use don't already exist. Default: true.
699.TP
700.BI allow_mounted_write \fR=\fPbool
701If this isn't set, fio will abort jobs that are destructive (e.g. that write)
702to what appears to be a mounted device or partition. This should help catch
703creating inadvertently destructive tests, not realizing that the test will
704destroy data on the mounted file system. Note that some platforms don't allow
705writing against a mounted device regardless of this option. Default: false.
706.TP
707.BI pre_read \fR=\fPbool
708If this is given, files will be pre\-read into memory before starting the
709given I/O operation. This will also clear the \fBinvalidate\fR flag,
710since it is pointless to pre\-read and then drop the cache. This will only
711work for I/O engines that are seek\-able, since they allow you to read the
712same data multiple times. Thus it will not work on non\-seekable I/O engines
713(e.g. network, splice). Default: false.
714.TP
715.BI unlink \fR=\fPbool
716Unlink the job files when done. Not the default, as repeated runs of that
717job would then waste time recreating the file set again and again. Default:
718false.
719.TP
720.BI unlink_each_loop \fR=\fPbool
721Unlink job files after each iteration or loop. Default: false.
722.TP
723.BI zonesize \fR=\fPint
724Divide a file into zones of the specified size. See \fBzoneskip\fR.
725.TP
726.BI zonerange \fR=\fPint
727Give size of an I/O zone. See \fBzoneskip\fR.
728.TP
729.BI zoneskip \fR=\fPint
730Skip the specified number of bytes when \fBzonesize\fR data has been
731read. The two zone options can be used to only do I/O on zones of a file.
732.SS "I/O type"
733.TP
734.BI direct \fR=\fPbool
735If value is true, use non\-buffered I/O. This is usually O_DIRECT. Note that
8e889110 736OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
523bad63
TK
737ioengines don't support direct I/O. Default: false.
738.TP
739.BI atomic \fR=\fPbool
740If value is true, attempt to use atomic direct I/O. Atomic writes are
741guaranteed to be stable once acknowledged by the operating system. Only
742Linux supports O_ATOMIC right now.
743.TP
744.BI buffered \fR=\fPbool
745If value is true, use buffered I/O. This is the opposite of the
746\fBdirect\fR option. Defaults to true.
d60e92d1
AC
747.TP
748.BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
523bad63 749Type of I/O pattern. Accepted values are:
d60e92d1
AC
750.RS
751.RS
752.TP
753.B read
d1429b5c 754Sequential reads.
d60e92d1
AC
755.TP
756.B write
d1429b5c 757Sequential writes.
d60e92d1 758.TP
fa769d44 759.B trim
169c098d 760Sequential trims (Linux block devices only).
fa769d44 761.TP
d60e92d1 762.B randread
d1429b5c 763Random reads.
d60e92d1
AC
764.TP
765.B randwrite
d1429b5c 766Random writes.
d60e92d1 767.TP
fa769d44 768.B randtrim
169c098d 769Random trims (Linux block devices only).
fa769d44 770.TP
523bad63
TK
771.B rw,readwrite
772Sequential mixed reads and writes.
d60e92d1 773.TP
ff6bb260 774.B randrw
523bad63 775Random mixed reads and writes.
82a90686
JA
776.TP
777.B trimwrite
523bad63
TK
778Sequential trim+write sequences. Blocks will be trimmed first,
779then the same blocks will be written to.
d60e92d1
AC
780.RE
781.P
523bad63
TK
782Fio defaults to read if the option is not specified. For the mixed I/O
783types, the default is to split them 50/50. For certain types of I/O the
784result may still be skewed a bit, since the speed may be different.
785.P
786It is possible to specify the number of I/Os to do before getting a new
787offset by appending `:<nr>' to the end of the string given. For a
788random read, it would look like `rw=randread:8' for passing in an offset
789modifier with a value of 8. If the suffix is used with a sequential I/O
790pattern, then the `<nr>' value specified will be added to the generated
791offset for each I/O turning sequential I/O into sequential I/O with holes.
792For instance, using `rw=write:4k' will skip 4k for every write. Also see
793the \fBrw_sequencer\fR option.
d60e92d1
AC
794.RE
795.TP
38dad62d 796.BI rw_sequencer \fR=\fPstr
523bad63
TK
797If an offset modifier is given by appending a number to the `rw=\fIstr\fR'
798line, then this option controls how that number modifies the I/O offset
799being generated. Accepted values are:
38dad62d
JA
800.RS
801.RS
802.TP
803.B sequential
523bad63 804Generate sequential offset.
38dad62d
JA
805.TP
806.B identical
523bad63 807Generate the same offset.
38dad62d
JA
808.RE
809.P
523bad63
TK
810\fBsequential\fR is only useful for random I/O, where fio would normally
811generate a new random offset for every I/O. If you append e.g. 8 to randread,
812you would get a new random offset for every 8 I/Os. The result would be a
813seek for only every 8 I/Os, instead of for every I/O. Use `rw=randread:8'
814to specify that. As sequential I/O is already sequential, setting
815\fBsequential\fR for that would not result in any differences. \fBidentical\fR
816behaves in a similar fashion, except it sends the same offset 8 number of
817times before generating a new offset.
38dad62d 818.RE
90fef2d1 819.TP
771e58be
JA
820.BI unified_rw_reporting \fR=\fPbool
821Fio normally reports statistics on a per data direction basis, meaning that
523bad63
TK
822reads, writes, and trims are accounted and reported separately. If this
823option is set fio sums the results and report them as "mixed" instead.
771e58be 824.TP
d60e92d1 825.BI randrepeat \fR=\fPbool
523bad63
TK
826Seed the random number generator used for random I/O patterns in a
827predictable way so the pattern is repeatable across runs. Default: true.
56e2a5fc
CE
828.TP
829.BI allrandrepeat \fR=\fPbool
830Seed all random number generators in a predictable way so results are
523bad63 831repeatable across runs. Default: false.
d60e92d1 832.TP
04778baf
JA
833.BI randseed \fR=\fPint
834Seed the random number generators based on this seed value, to be able to
835control what sequence of output is being generated. If not set, the random
836sequence depends on the \fBrandrepeat\fR setting.
837.TP
a596f047 838.BI fallocate \fR=\fPstr
523bad63
TK
839Whether pre\-allocation is performed when laying down files.
840Accepted values are:
a596f047
EG
841.RS
842.RS
843.TP
844.B none
523bad63 845Do not pre\-allocate space.
a596f047 846.TP
2c3e17be 847.B native
523bad63
TK
848Use a platform's native pre\-allocation call but fall back to
849\fBnone\fR behavior if it fails/is not implemented.
2c3e17be 850.TP
a596f047 851.B posix
523bad63 852Pre\-allocate via \fBposix_fallocate\fR\|(3).
a596f047
EG
853.TP
854.B keep
523bad63
TK
855Pre\-allocate via \fBfallocate\fR\|(2) with
856FALLOC_FL_KEEP_SIZE set.
a596f047
EG
857.TP
858.B 0
523bad63 859Backward\-compatible alias for \fBnone\fR.
a596f047
EG
860.TP
861.B 1
523bad63 862Backward\-compatible alias for \fBposix\fR.
a596f047
EG
863.RE
864.P
523bad63
TK
865May not be available on all supported platforms. \fBkeep\fR is only available
866on Linux. If using ZFS on Solaris this cannot be set to \fBposix\fR
867because ZFS doesn't support pre\-allocation. Default: \fBnative\fR if any
868pre\-allocation methods are available, \fBnone\fR if not.
a596f047 869.RE
7bc8c2cf 870.TP
ecb2083d 871.BI fadvise_hint \fR=\fPstr
c712c97a
JA
872Use \fBposix_fadvise\fR\|(2) or \fBposix_madvise\fR\|(2) to advise the kernel
873what I/O patterns are likely to be issued. Accepted values are:
ecb2083d
JA
874.RS
875.RS
876.TP
877.B 0
878Backwards compatible hint for "no hint".
879.TP
880.B 1
881Backwards compatible hint for "advise with fio workload type". This
523bad63 882uses FADV_RANDOM for a random workload, and FADV_SEQUENTIAL
ecb2083d
JA
883for a sequential workload.
884.TP
885.B sequential
523bad63 886Advise using FADV_SEQUENTIAL.
ecb2083d
JA
887.TP
888.B random
523bad63 889Advise using FADV_RANDOM.
ecb2083d
JA
890.RE
891.RE
d60e92d1 892.TP
8f4b9f24 893.BI write_hint \fR=\fPstr
523bad63
TK
894Use \fBfcntl\fR\|(2) to advise the kernel what life time to expect
895from a write. Only supported on Linux, as of version 4.13. Accepted
8f4b9f24
JA
896values are:
897.RS
898.RS
899.TP
900.B none
901No particular life time associated with this file.
902.TP
903.B short
904Data written to this file has a short life time.
905.TP
906.B medium
907Data written to this file has a medium life time.
908.TP
909.B long
910Data written to this file has a long life time.
911.TP
912.B extreme
913Data written to this file has a very long life time.
914.RE
523bad63
TK
915.P
916The values are all relative to each other, and no absolute meaning
917should be associated with them.
8f4b9f24 918.RE
37659335 919.TP
523bad63
TK
920.BI offset \fR=\fPint
921Start I/O at the provided offset in the file, given as either a fixed size in
83c8b093
JF
922bytes or a percentage. If a percentage is given, the generated offset will be
923aligned to the minimum \fBblocksize\fR or to the value of \fBoffset_align\fR if
924provided. Data before the given offset will not be touched. This
523bad63
TK
925effectively caps the file size at `real_size \- offset'. Can be combined with
926\fBsize\fR to constrain the start and end range of the I/O workload.
927A percentage can be specified by a number between 1 and 100 followed by '%',
928for example, `offset=20%' to specify 20%.
6d500c2e 929.TP
83c8b093
JF
930.BI offset_align \fR=\fPint
931If set to non-zero value, the byte offset generated by a percentage \fBoffset\fR
932is aligned upwards to this value. Defaults to 0 meaning that a percentage
933offset is aligned to the minimum block size.
934.TP
523bad63
TK
935.BI offset_increment \fR=\fPint
936If this is provided, then the real offset becomes `\fBoffset\fR + \fBoffset_increment\fR
937* thread_number', where the thread number is a counter that starts at 0 and
938is incremented for each sub\-job (i.e. when \fBnumjobs\fR option is
939specified). This option is useful if there are several jobs which are
940intended to operate on a file in parallel disjoint segments, with even
941spacing between the starting points.
6d500c2e 942.TP
523bad63
TK
943.BI number_ios \fR=\fPint
944Fio will normally perform I/Os until it has exhausted the size of the region
945set by \fBsize\fR, or if it exhaust the allocated time (or hits an error
946condition). With this setting, the range/size can be set independently of
947the number of I/Os to perform. When fio reaches this number, it will exit
948normally and report status. Note that this does not extend the amount of I/O
949that will be done, it will only stop fio if this condition is met before
950other end\-of\-job criteria.
d60e92d1 951.TP
523bad63
TK
952.BI fsync \fR=\fPint
953If writing to a file, issue an \fBfsync\fR\|(2) (or its equivalent) of
954the dirty data for every number of blocks given. For example, if you give 32
955as a parameter, fio will sync the file after every 32 writes issued. If fio is
956using non\-buffered I/O, we may not sync the file. The exception is the sg
957I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
958means fio does not periodically issue and wait for a sync to complete. Also
959see \fBend_fsync\fR and \fBfsync_on_close\fR.
6d500c2e 960.TP
523bad63
TK
961.BI fdatasync \fR=\fPint
962Like \fBfsync\fR but uses \fBfdatasync\fR\|(2) to only sync data and
963not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
964\fBfdatasync\fR\|(2) so this falls back to using \fBfsync\fR\|(2).
965Defaults to 0, which means fio does not periodically issue and wait for a
966data\-only sync to complete.
d60e92d1 967.TP
523bad63
TK
968.BI write_barrier \fR=\fPint
969Make every N\-th write a barrier write.
901bb994 970.TP
523bad63
TK
971.BI sync_file_range \fR=\fPstr:int
972Use \fBsync_file_range\fR\|(2) for every \fIint\fR number of write
973operations. Fio will track range of writes that have happened since the last
974\fBsync_file_range\fR\|(2) call. \fIstr\fR can currently be one or more of:
975.RS
976.RS
fd68418e 977.TP
523bad63
TK
978.B wait_before
979SYNC_FILE_RANGE_WAIT_BEFORE
c5751c62 980.TP
523bad63
TK
981.B write
982SYNC_FILE_RANGE_WRITE
c5751c62 983.TP
523bad63
TK
984.B wait_after
985SYNC_FILE_RANGE_WRITE_AFTER
2fa5a241 986.RE
523bad63
TK
987.P
988So if you do `sync_file_range=wait_before,write:8', fio would use
989`SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE' for every 8
990writes. Also see the \fBsync_file_range\fR\|(2) man page. This option is
991Linux specific.
2fa5a241 992.RE
ce35b1ec 993.TP
523bad63
TK
994.BI overwrite \fR=\fPbool
995If true, writes to a file will always overwrite existing data. If the file
996doesn't already exist, it will be created before the write phase begins. If
997the file exists and is large enough for the specified write phase, nothing
998will be done. Default: false.
5c94b008 999.TP
523bad63
TK
1000.BI end_fsync \fR=\fPbool
1001If true, \fBfsync\fR\|(2) file contents when a write stage has completed.
1002Default: false.
d60e92d1 1003.TP
523bad63
TK
1004.BI fsync_on_close \fR=\fPbool
1005If true, fio will \fBfsync\fR\|(2) a dirty file on close. This differs
1006from \fBend_fsync\fR in that it will happen on every file close, not
1007just at the end of the job. Default: false.
d60e92d1 1008.TP
523bad63
TK
1009.BI rwmixread \fR=\fPint
1010Percentage of a mixed workload that should be reads. Default: 50.
1011.TP
1012.BI rwmixwrite \fR=\fPint
1013Percentage of a mixed workload that should be writes. If both
1014\fBrwmixread\fR and \fBrwmixwrite\fR is given and the values do not
1015add up to 100%, the latter of the two will be used to override the
1016first. This may interfere with a given rate setting, if fio is asked to
1017limit reads or writes to a certain rate. If that is the case, then the
1018distribution may be skewed. Default: 50.
1019.TP
1020.BI random_distribution \fR=\fPstr:float[,str:float][,str:float]
1021By default, fio will use a completely uniform random distribution when asked
1022to perform random I/O. Sometimes it is useful to skew the distribution in
1023specific ways, ensuring that some parts of the data is more hot than others.
1024fio includes the following distribution models:
d60e92d1
AC
1025.RS
1026.RS
1027.TP
1028.B random
523bad63 1029Uniform random distribution
8c07860d
JA
1030.TP
1031.B zipf
523bad63 1032Zipf distribution
8c07860d
JA
1033.TP
1034.B pareto
523bad63 1035Pareto distribution
8c07860d 1036.TP
dd3503d3 1037.B normal
523bad63 1038Normal (Gaussian) distribution
dd3503d3 1039.TP
523bad63
TK
1040.B zoned
1041Zoned random distribution
59466396
JA
1042.B zoned_abs
1043Zoned absolute random distribution
d60e92d1
AC
1044.RE
1045.P
523bad63
TK
1046When using a \fBzipf\fR or \fBpareto\fR distribution, an input value is also
1047needed to define the access pattern. For \fBzipf\fR, this is the `Zipf theta'.
1048For \fBpareto\fR, it's the `Pareto power'. Fio includes a test
1049program, \fBfio\-genzipf\fR, that can be used visualize what the given input
1050values will yield in terms of hit rates. If you wanted to use \fBzipf\fR with
1051a `theta' of 1.2, you would use `random_distribution=zipf:1.2' as the
1052option. If a non\-uniform model is used, fio will disable use of the random
1053map. For the \fBnormal\fR distribution, a normal (Gaussian) deviation is
1054supplied as a value between 0 and 100.
1055.P
1056For a \fBzoned\fR distribution, fio supports specifying percentages of I/O
1057access that should fall within what range of the file or device. For
1058example, given a criteria of:
d60e92d1 1059.RS
523bad63
TK
1060.P
1061.PD 0
106260% of accesses should be to the first 10%
1063.P
106430% of accesses should be to the next 20%
1065.P
10668% of accesses should be to the next 30%
1067.P
10682% of accesses should be to the next 40%
1069.PD
1070.RE
1071.P
1072we can define that through zoning of the random accesses. For the above
1073example, the user would do:
1074.RS
1075.P
1076random_distribution=zoned:60/10:30/20:8/30:2/40
1077.RE
1078.P
59466396
JA
1079A \fBzoned_abs\fR distribution works exactly like the\fBzoned\fR, except that
1080it takes absolute sizes. For example, let's say you wanted to define access
1081according to the following criteria:
1082.RS
1083.P
1084.PD 0
108560% of accesses should be to the first 20G
1086.P
108730% of accesses should be to the next 100G
1088.P
108910% of accesses should be to the next 500G
1090.PD
1091.RE
1092.P
1093we can define an absolute zoning distribution with:
1094.RS
1095.P
1096random_distribution=zoned:60/10:30/20:8/30:2/40
1097.RE
1098.P
6a16ece8
JA
1099For both \fBzoned\fR and \fBzoned_abs\fR, fio supports defining up to 256
1100separate zones.
1101.P
59466396 1102Similarly to how \fBbssplit\fR works for setting ranges and percentages
523bad63
TK
1103of block sizes. Like \fBbssplit\fR, it's possible to specify separate
1104zones for reads, writes, and trims. If just one set is given, it'll apply to
1105all of them.
1106.RE
1107.TP
1108.BI percentage_random \fR=\fPint[,int][,int]
1109For a random workload, set how big a percentage should be random. This
1110defaults to 100%, in which case the workload is fully random. It can be set
1111from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1112sequential. Any setting in between will result in a random mix of sequential
1113and random I/O, at the given percentages. Comma\-separated values may be
1114specified for reads, writes, and trims as described in \fBblocksize\fR.
1115.TP
1116.BI norandommap
1117Normally fio will cover every block of the file when doing random I/O. If
1118this option is given, fio will just get a new random offset without looking
1119at past I/O history. This means that some blocks may not be read or written,
1120and that some blocks may be read/written more than once. If this option is
1121used with \fBverify\fR and multiple blocksizes (via \fBbsrange\fR),
1122only intact blocks are verified, i.e., partially\-overwritten blocks are
47e6a6e5
SW
1123ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1124the same block to be overwritten, which can cause verification errors. Either
1125do not use norandommap in this case, or also use the lfsr random generator.
523bad63
TK
1126.TP
1127.BI softrandommap \fR=\fPbool
1128See \fBnorandommap\fR. If fio runs with the random block map enabled and
1129it fails to allocate the map, if this option is set it will continue without
1130a random block map. As coverage will not be as complete as with random maps,
1131this option is disabled by default.
1132.TP
1133.BI random_generator \fR=\fPstr
1134Fio supports the following engines for generating I/O offsets for random I/O:
1135.RS
1136.RS
1137.TP
1138.B tausworthe
1139Strong 2^88 cycle random number generator.
1140.TP
1141.B lfsr
1142Linear feedback shift register generator.
1143.TP
1144.B tausworthe64
1145Strong 64\-bit 2^258 cycle random number generator.
1146.RE
1147.P
1148\fBtausworthe\fR is a strong random number generator, but it requires tracking
1149on the side if we want to ensure that blocks are only read or written
1150once. \fBlfsr\fR guarantees that we never generate the same offset twice, and
1151it's also less computationally expensive. It's not a true random generator,
1152however, though for I/O purposes it's typically good enough. \fBlfsr\fR only
1153works with single block sizes, not with workloads that use multiple block
1154sizes. If used with such a workload, fio may read or write some blocks
1155multiple times. The default value is \fBtausworthe\fR, unless the required
1156space exceeds 2^32 blocks. If it does, then \fBtausworthe64\fR is
1157selected automatically.
1158.RE
1159.SS "Block size"
1160.TP
1161.BI blocksize \fR=\fPint[,int][,int] "\fR,\fB bs" \fR=\fPint[,int][,int]
1162The block size in bytes used for I/O units. Default: 4096. A single value
1163applies to reads, writes, and trims. Comma\-separated values may be
1164specified for reads, writes, and trims. A value not terminated in a comma
1165applies to subsequent types. Examples:
1166.RS
1167.RS
1168.P
1169.PD 0
1170bs=256k means 256k for reads, writes and trims.
1171.P
1172bs=8k,32k means 8k for reads, 32k for writes and trims.
1173.P
1174bs=8k,32k, means 8k for reads, 32k for writes, and default for trims.
1175.P
1176bs=,8k means default for reads, 8k for writes and trims.
1177.P
1178bs=,8k, means default for reads, 8k for writes, and default for trims.
1179.PD
1180.RE
1181.RE
1182.TP
1183.BI blocksize_range \fR=\fPirange[,irange][,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange][,irange]
1184A range of block sizes in bytes for I/O units. The issued I/O unit will
1185always be a multiple of the minimum size, unless
1186\fBblocksize_unaligned\fR is set.
1187Comma\-separated ranges may be specified for reads, writes, and trims as
1188described in \fBblocksize\fR. Example:
1189.RS
1190.RS
1191.P
1192bsrange=1k\-4k,2k\-8k
1193.RE
1194.RE
1195.TP
1196.BI bssplit \fR=\fPstr[,str][,str]
1197Sometimes you want even finer grained control of the block sizes issued, not
1198just an even split between them. This option allows you to weight various
1199block sizes, so that you are able to define a specific amount of block sizes
1200issued. The format for this option is:
1201.RS
1202.RS
1203.P
1204bssplit=blocksize/percentage:blocksize/percentage
1205.RE
1206.P
1207for as many block sizes as needed. So if you want to define a workload that
1208has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would write:
1209.RS
1210.P
1211bssplit=4k/10:64k/50:32k/40
1212.RE
1213.P
1214Ordering does not matter. If the percentage is left blank, fio will fill in
1215the remaining values evenly. So a bssplit option like this one:
1216.RS
1217.P
1218bssplit=4k/50:1k/:32k/
1219.RE
1220.P
1221would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always add up
1222to 100, if bssplit is given a range that adds up to more, it will error out.
1223.P
1224Comma\-separated values may be specified for reads, writes, and trims as
1225described in \fBblocksize\fR.
1226.P
1227If you want a workload that has 50% 2k reads and 50% 4k reads, while having
122890% 4k writes and 10% 8k writes, you would specify:
1229.RS
1230.P
cf04b906 1231bssplit=2k/50:4k/50,4k/90:8k/10
523bad63 1232.RE
6a16ece8
JA
1233.P
1234Fio supports defining up to 64 different weights for each data direction.
523bad63
TK
1235.RE
1236.TP
1237.BI blocksize_unaligned "\fR,\fB bs_unaligned"
1238If set, fio will issue I/O units with any size within
1239\fBblocksize_range\fR, not just multiples of the minimum size. This
1240typically won't work with direct I/O, as that normally requires sector
1241alignment.
1242.TP
1243.BI bs_is_seq_rand \fR=\fPbool
1244If this option is set, fio will use the normal read,write blocksize settings
1245as sequential,random blocksize settings instead. Any random read or write
1246will use the WRITE blocksize settings, and any sequential read or write will
1247use the READ blocksize settings.
1248.TP
1249.BI blockalign \fR=\fPint[,int][,int] "\fR,\fB ba" \fR=\fPint[,int][,int]
1250Boundary to which fio will align random I/O units. Default:
1251\fBblocksize\fR. Minimum alignment is typically 512b for using direct
1252I/O, though it usually depends on the hardware block size. This option is
1253mutually exclusive with using a random map for files, so it will turn off
1254that option. Comma\-separated values may be specified for reads, writes, and
1255trims as described in \fBblocksize\fR.
1256.SS "Buffers and memory"
1257.TP
1258.BI zero_buffers
1259Initialize buffers with all zeros. Default: fill buffers with random data.
1260.TP
1261.BI refill_buffers
1262If this option is given, fio will refill the I/O buffers on every
1263submit. The default is to only fill it at init time and reuse that
1264data. Only makes sense if zero_buffers isn't specified, naturally. If data
1265verification is enabled, \fBrefill_buffers\fR is also automatically enabled.
1266.TP
1267.BI scramble_buffers \fR=\fPbool
1268If \fBrefill_buffers\fR is too costly and the target is using data
1269deduplication, then setting this option will slightly modify the I/O buffer
1270contents to defeat normal de\-dupe attempts. This is not enough to defeat
1271more clever block compression attempts, but it will stop naive dedupe of
1272blocks. Default: true.
1273.TP
1274.BI buffer_compress_percentage \fR=\fPint
72592780
SW
1275If this is set, then fio will attempt to provide I/O buffer content
1276(on WRITEs) that compresses to the specified level. Fio does this by
1277providing a mix of random data followed by fixed pattern data. The
1278fixed pattern is either zeros, or the pattern specified by
1279\fBbuffer_pattern\fR. If the \fBbuffer_pattern\fR option is used, it
1280might skew the compression ratio slightly. Setting
1281\fBbuffer_compress_percentage\fR to a value other than 100 will also
1282enable \fBrefill_buffers\fR in order to reduce the likelihood that
1283adjacent blocks are so similar that they over compress when seen
1284together. See \fBbuffer_compress_chunk\fR for how to set a finer or
1285coarser granularity of the random/fixed data regions. Defaults to unset
1286i.e., buffer data will not adhere to any compression level.
523bad63
TK
1287.TP
1288.BI buffer_compress_chunk \fR=\fPint
72592780
SW
1289This setting allows fio to manage how big the random/fixed data region
1290is when using \fBbuffer_compress_percentage\fR. When
1291\fBbuffer_compress_chunk\fR is set to some non-zero value smaller than the
1292block size, fio can repeat the random/fixed region throughout the I/O
1293buffer at the specified interval (which particularly useful when
1294bigger block sizes are used for a job). When set to 0, fio will use a
1295chunk size that matches the block size resulting in a single
1296random/fixed region within the I/O buffer. Defaults to 512. When the
1297unit is omitted, the value is interpreted in bytes.
523bad63
TK
1298.TP
1299.BI buffer_pattern \fR=\fPstr
1300If set, fio will fill the I/O buffers with this pattern or with the contents
1301of a file. If not set, the contents of I/O buffers are defined by the other
1302options related to buffer contents. The setting can be any pattern of bytes,
1303and can be prefixed with 0x for hex values. It may also be a string, where
1304the string must then be wrapped with "". Or it may also be a filename,
1305where the filename must be wrapped with '' in which case the file is
1306opened and read. Note that not all the file contents will be read if that
1307would cause the buffers to overflow. So, for example:
1308.RS
1309.RS
1310.P
1311.PD 0
1312buffer_pattern='filename'
1313.P
1314or:
1315.P
1316buffer_pattern="abcd"
1317.P
1318or:
1319.P
1320buffer_pattern=\-12
1321.P
1322or:
1323.P
1324buffer_pattern=0xdeadface
1325.PD
1326.RE
1327.P
1328Also you can combine everything together in any order:
1329.RS
1330.P
1331buffer_pattern=0xdeadface"abcd"\-12'filename'
1332.RE
1333.RE
1334.TP
1335.BI dedupe_percentage \fR=\fPint
1336If set, fio will generate this percentage of identical buffers when
1337writing. These buffers will be naturally dedupable. The contents of the
1338buffers depend on what other buffer compression settings have been set. It's
1339possible to have the individual buffers either fully compressible, or not at
72592780
SW
1340all \-\- this option only controls the distribution of unique buffers. Setting
1341this option will also enable \fBrefill_buffers\fR to prevent every buffer
1342being identical.
523bad63
TK
1343.TP
1344.BI invalidate \fR=\fPbool
1345Invalidate the buffer/page cache parts of the files to be used prior to
1346starting I/O if the platform and file type support it. Defaults to true.
1347This will be ignored if \fBpre_read\fR is also specified for the
1348same job.
1349.TP
1350.BI sync \fR=\fPbool
1351Use synchronous I/O for buffered writes. For the majority of I/O engines,
1352this means using O_SYNC. Default: false.
1353.TP
1354.BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
1355Fio can use various types of memory as the I/O unit buffer. The allowed
1356values are:
1357.RS
1358.RS
1359.TP
1360.B malloc
1361Use memory from \fBmalloc\fR\|(3) as the buffers. Default memory type.
1362.TP
1363.B shm
1364Use shared memory as the buffers. Allocated through \fBshmget\fR\|(2).
1365.TP
1366.B shmhuge
1367Same as \fBshm\fR, but use huge pages as backing.
1368.TP
1369.B mmap
1370Use \fBmmap\fR\|(2) to allocate buffers. May either be anonymous memory, or can
1371be file backed if a filename is given after the option. The format
1372is `mem=mmap:/path/to/file'.
1373.TP
1374.B mmaphuge
1375Use a memory mapped huge file as the buffer backing. Append filename
1376after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file'.
1377.TP
1378.B mmapshared
1379Same as \fBmmap\fR, but use a MMAP_SHARED mapping.
1380.TP
1381.B cudamalloc
1382Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1383The \fBioengine\fR must be \fBrdma\fR.
1384.RE
1385.P
1386The area allocated is a function of the maximum allowed bs size for the job,
1387multiplied by the I/O depth given. Note that for \fBshmhuge\fR and
1388\fBmmaphuge\fR to work, the system must have free huge pages allocated. This
1389can normally be checked and set by reading/writing
1390`/proc/sys/vm/nr_hugepages' on a Linux system. Fio assumes a huge page
1391is 4MiB in size. So to calculate the number of huge pages you need for a
1392given job file, add up the I/O depth of all jobs (normally one unless
1393\fBiodepth\fR is used) and multiply by the maximum bs set. Then divide
1394that number by the huge page size. You can see the size of the huge pages in
1395`/proc/meminfo'. If no huge pages are allocated by having a non\-zero
1396number in `nr_hugepages', using \fBmmaphuge\fR or \fBshmhuge\fR will fail. Also
1397see \fBhugepage\-size\fR.
1398.P
1399\fBmmaphuge\fR also needs to have hugetlbfs mounted and the file location
1400should point there. So if it's mounted in `/huge', you would use
1401`mem=mmaphuge:/huge/somefile'.
1402.RE
1403.TP
1404.BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
1405This indicates the memory alignment of the I/O memory buffers. Note that
1406the given alignment is applied to the first I/O unit buffer, if using
1407\fBiodepth\fR the alignment of the following buffers are given by the
1408\fBbs\fR used. In other words, if using a \fBbs\fR that is a
1409multiple of the page sized in the system, all buffers will be aligned to
1410this value. If using a \fBbs\fR that is not page aligned, the alignment
1411of subsequent I/O memory buffers is the sum of the \fBiomem_align\fR and
1412\fBbs\fR used.
1413.TP
1414.BI hugepage\-size \fR=\fPint
1415Defines the size of a huge page. Must at least be equal to the system
1416setting, see `/proc/meminfo'. Defaults to 4MiB. Should probably
1417always be a multiple of megabytes, so using `hugepage\-size=Xm' is the
1418preferred way to set this to avoid setting a non\-pow\-2 bad value.
1419.TP
1420.BI lockmem \fR=\fPint
1421Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
1422simulate a smaller amount of memory. The amount specified is per worker.
1423.SS "I/O size"
1424.TP
1425.BI size \fR=\fPint
1426The total size of file I/O for each thread of this job. Fio will run until
1427this many bytes has been transferred, unless runtime is limited by other options
1428(such as \fBruntime\fR, for instance, or increased/decreased by \fBio_size\fR).
1429Fio will divide this size between the available files determined by options
1430such as \fBnrfiles\fR, \fBfilename\fR, unless \fBfilesize\fR is
1431specified by the job. If the result of division happens to be 0, the size is
1432set to the physical size of the given files or devices if they exist.
1433If this option is not specified, fio will use the full size of the given
1434files or devices. If the files do not exist, size must be given. It is also
1435possible to give size as a percentage between 1 and 100. If `size=20%' is
1436given, fio will use 20% of the full size of the given files or devices.
1437Can be combined with \fBoffset\fR to constrain the start and end range
1438that I/O will be done within.
1439.TP
1440.BI io_size \fR=\fPint "\fR,\fB io_limit" \fR=\fPint
1441Normally fio operates within the region set by \fBsize\fR, which means
1442that the \fBsize\fR option sets both the region and size of I/O to be
1443performed. Sometimes that is not what you want. With this option, it is
1444possible to define just the amount of I/O that fio should do. For instance,
1445if \fBsize\fR is set to 20GiB and \fBio_size\fR is set to 5GiB, fio
1446will perform I/O within the first 20GiB but exit when 5GiB have been
1447done. The opposite is also possible \-\- if \fBsize\fR is set to 20GiB,
1448and \fBio_size\fR is set to 40GiB, then fio will do 40GiB of I/O within
1449the 0..20GiB region.
1450.TP
1451.BI filesize \fR=\fPirange(int)
1452Individual file sizes. May be a range, in which case fio will select sizes
1453for files at random within the given range and limited to \fBsize\fR in
1454total (if that is given). If not given, each created file is the same size.
1455This option overrides \fBsize\fR in terms of file size, which means
1456this value is used as a fixed size or possible range of each file.
1457.TP
1458.BI file_append \fR=\fPbool
1459Perform I/O after the end of the file. Normally fio will operate within the
1460size of a file. If this option is set, then fio will append to the file
1461instead. This has identical behavior to setting \fBoffset\fR to the size
1462of a file. This option is ignored on non\-regular files.
1463.TP
1464.BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
1465Sets size to something really large and waits for ENOSPC (no space left on
1466device) as the terminating condition. Only makes sense with sequential
1467write. For a read workload, the mount point will be filled first then I/O
1468started on the result. This option doesn't make sense if operating on a raw
1469device node, since the size of that is already known by the file system.
1470Additionally, writing beyond end\-of\-device will not return ENOSPC there.
1471.SS "I/O engine"
1472.TP
1473.BI ioengine \fR=\fPstr
1474Defines how the job issues I/O to the file. The following types are defined:
1475.RS
1476.RS
1477.TP
1478.B sync
1479Basic \fBread\fR\|(2) or \fBwrite\fR\|(2)
1480I/O. \fBlseek\fR\|(2) is used to position the I/O location.
1481See \fBfsync\fR and \fBfdatasync\fR for syncing write I/Os.
1482.TP
1483.B psync
1484Basic \fBpread\fR\|(2) or \fBpwrite\fR\|(2) I/O. Default on
1485all supported operating systems except for Windows.
1486.TP
1487.B vsync
1488Basic \fBreadv\fR\|(2) or \fBwritev\fR\|(2) I/O. Will emulate
1489queuing by coalescing adjacent I/Os into a single submission.
1490.TP
1491.B pvsync
1492Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
a46c5e01 1493.TP
2cafffbe
JA
1494.B pvsync2
1495Basic \fBpreadv2\fR\|(2) or \fBpwritev2\fR\|(2) I/O.
1496.TP
d60e92d1 1497.B libaio
523bad63
TK
1498Linux native asynchronous I/O. Note that Linux may only support
1499queued behavior with non\-buffered I/O (set `direct=1' or
1500`buffered=0').
1501This engine defines engine specific options.
d60e92d1
AC
1502.TP
1503.B posixaio
523bad63
TK
1504POSIX asynchronous I/O using \fBaio_read\fR\|(3) and
1505\fBaio_write\fR\|(3).
03e20d68
BC
1506.TP
1507.B solarisaio
1508Solaris native asynchronous I/O.
1509.TP
1510.B windowsaio
38f8c318 1511Windows native asynchronous I/O. Default on Windows.
d60e92d1
AC
1512.TP
1513.B mmap
523bad63
TK
1514File is memory mapped with \fBmmap\fR\|(2) and data copied
1515to/from using \fBmemcpy\fR\|(3).
d60e92d1
AC
1516.TP
1517.B splice
523bad63
TK
1518\fBsplice\fR\|(2) is used to transfer the data and
1519\fBvmsplice\fR\|(2) to transfer data from user space to the
1520kernel.
d60e92d1 1521.TP
d60e92d1 1522.B sg
523bad63
TK
1523SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1524ioctl, or if the target is an sg character device we use
1525\fBread\fR\|(2) and \fBwrite\fR\|(2) for asynchronous
1526I/O. Requires \fBfilename\fR option to specify either block or
52b81b7c 1527character devices. The sg engine includes engine specific options.
d60e92d1
AC
1528.TP
1529.B null
523bad63
TK
1530Doesn't transfer any data, just pretends to. This is mainly used to
1531exercise fio itself and for debugging/testing purposes.
d60e92d1
AC
1532.TP
1533.B net
523bad63
TK
1534Transfer over the network to given `host:port'. Depending on the
1535\fBprotocol\fR used, the \fBhostname\fR, \fBport\fR,
1536\fBlisten\fR and \fBfilename\fR options are used to specify
1537what sort of connection to make, while the \fBprotocol\fR option
1538determines which protocol will be used. This engine defines engine
1539specific options.
d60e92d1
AC
1540.TP
1541.B netsplice
523bad63
TK
1542Like \fBnet\fR, but uses \fBsplice\fR\|(2) and
1543\fBvmsplice\fR\|(2) to map data and send/receive.
1544This engine defines engine specific options.
d60e92d1 1545.TP
53aec0a4 1546.B cpuio
523bad63
TK
1547Doesn't transfer any data, but burns CPU cycles according to the
1548\fBcpuload\fR and \fBcpuchunks\fR options. Setting
1549\fBcpuload\fR\=85 will cause that job to do nothing but burn 85%
1550of the CPU. In case of SMP machines, use `numjobs=<nr_of_cpu>'
1551to get desired CPU usage, as the cpuload only loads a
1552single CPU at the desired rate. A job never finishes unless there is
1553at least one non\-cpuio job.
d60e92d1
AC
1554.TP
1555.B guasi
f1dd3fb1 1556The GUASI I/O engine is the Generic Userspace Asynchronous Syscall
523bad63
TK
1557Interface approach to async I/O. See \fIhttp://www.xmailserver.org/guasi\-lib.html\fR
1558for more info on GUASI.
d60e92d1 1559.TP
21b8aee8 1560.B rdma
523bad63
TK
1561The RDMA I/O engine supports both RDMA memory semantics
1562(RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
609ac152
SB
1563InfiniBand, RoCE and iWARP protocols. This engine defines engine
1564specific options.
d54fce84
DM
1565.TP
1566.B falloc
523bad63
TK
1567I/O engine that does regular fallocate to simulate data transfer as
1568fio ioengine.
1569.RS
1570.P
1571.PD 0
1572DDIR_READ does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1573.P
1574DIR_WRITE does fallocate(,mode = 0).
1575.P
1576DDIR_TRIM does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1577.PD
1578.RE
1579.TP
1580.B ftruncate
1581I/O engine that sends \fBftruncate\fR\|(2) operations in response
1582to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1583size to the current block offset. \fBblocksize\fR is ignored.
d54fce84
DM
1584.TP
1585.B e4defrag
523bad63
TK
1586I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1587defragment activity in request to DDIR_WRITE event.
0d978694 1588.TP
d5f9b0ea
IF
1589.B rados
1590I/O engine supporting direct access to Ceph Reliable Autonomic Distributed
1591Object Store (RADOS) via librados. This ioengine defines engine specific
1592options.
1593.TP
0d978694 1594.B rbd
523bad63
TK
1595I/O engine supporting direct access to Ceph Rados Block Devices
1596(RBD) via librbd without the need to use the kernel rbd driver. This
1597ioengine defines engine specific options.
a7c386f4 1598.TP
1599.B gfapi
523bad63
TK
1600Using GlusterFS libgfapi sync interface to direct access to
1601GlusterFS volumes without having to go through FUSE. This ioengine
1602defines engine specific options.
cc47f094 1603.TP
1604.B gfapi_async
523bad63
TK
1605Using GlusterFS libgfapi async interface to direct access to
1606GlusterFS volumes without having to go through FUSE. This ioengine
1607defines engine specific options.
1b10477b 1608.TP
b74e419e 1609.B libhdfs
523bad63
TK
1610Read and write through Hadoop (HDFS). The \fBfilename\fR option
1611is used to specify host,port of the hdfs name\-node to connect. This
1612engine interprets offsets a little differently. In HDFS, files once
1613created cannot be modified so random writes are not possible. To
1614imitate this the libhdfs engine expects a bunch of small files to be
1615created over HDFS and will randomly pick a file from them
1616based on the offset generated by fio backend (see the example
1617job file to create such files, use `rw=write' option). Please
1618note, it may be necessary to set environment variables to work
1619with HDFS/libhdfs properly. Each job uses its own connection to
1620HDFS.
65fa28ca
DE
1621.TP
1622.B mtd
523bad63
TK
1623Read, write and erase an MTD character device (e.g.,
1624`/dev/mtd0'). Discards are treated as erases. Depending on the
1625underlying device type, the I/O may have to go in a certain pattern,
1626e.g., on NAND, writing sequentially to erase blocks and discarding
1627before overwriting. The \fBtrimwrite\fR mode works well for this
65fa28ca 1628constraint.
5c4ef02e
JA
1629.TP
1630.B pmemblk
523bad63 1631Read and write using filesystem DAX to a file on a filesystem
363a5f65 1632mounted with DAX on a persistent memory device through the PMDK
523bad63 1633libpmemblk library.
104ee4de 1634.TP
523bad63
TK
1635.B dev\-dax
1636Read and write using device DAX to a persistent memory device (e.g.,
363a5f65 1637/dev/dax0.0) through the PMDK libpmem library.
d60e92d1 1638.TP
523bad63
TK
1639.B external
1640Prefix to specify loading an external I/O engine object file. Append
1641the engine filename, e.g. `ioengine=external:/tmp/foo.o' to load
d243fd6d
TK
1642ioengine `foo.o' in `/tmp'. The path can be either
1643absolute or relative. See `engines/skeleton_external.c' in the fio source for
1644details of writing an external I/O engine.
1216cc5a
JB
1645.TP
1646.B filecreate
b71968b1
SW
1647Simply create the files and do no I/O to them. You still need to set
1648\fBfilesize\fR so that all the accounting still occurs, but no actual I/O will be
1649done other than creating the file.
ae0db592
TI
1650.TP
1651.B libpmem
1652Read and write using mmap I/O to a file on a filesystem
363a5f65 1653mounted with DAX on a persistent memory device through the PMDK
ae0db592 1654libpmem library.
523bad63
TK
1655.SS "I/O engine specific parameters"
1656In addition, there are some parameters which are only valid when a specific
1657\fBioengine\fR is in use. These are used identically to normal parameters,
1658with the caveat that when used on the command line, they must come after the
1659\fBioengine\fR that defines them is selected.
d60e92d1 1660.TP
523bad63
TK
1661.BI (libaio)userspace_reap
1662Normally, with the libaio engine in use, fio will use the
1663\fBio_getevents\fR\|(3) system call to reap newly returned events. With
1664this flag turned on, the AIO ring will be read directly from user\-space to
1665reap events. The reaping mode is only enabled when polling for a minimum of
16660 events (e.g. when `iodepth_batch_complete=0').
3ce9dcaf 1667.TP
523bad63
TK
1668.BI (pvsync2)hipri
1669Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1670than normal.
82407585 1671.TP
523bad63
TK
1672.BI (pvsync2)hipri_percentage
1673When hipri is set this determines the probability of a pvsync2 I/O being high
1674priority. The default is 100%.
d60e92d1 1675.TP
523bad63
TK
1676.BI (cpuio)cpuload \fR=\fPint
1677Attempt to use the specified percentage of CPU cycles. This is a mandatory
1678option when using cpuio I/O engine.
997b5680 1679.TP
523bad63
TK
1680.BI (cpuio)cpuchunks \fR=\fPint
1681Split the load into cycles of the given time. In microseconds.
1ad01bd1 1682.TP
523bad63
TK
1683.BI (cpuio)exit_on_io_done \fR=\fPbool
1684Detect when I/O threads are done, then exit.
d60e92d1 1685.TP
523bad63
TK
1686.BI (libhdfs)namenode \fR=\fPstr
1687The hostname or IP address of a HDFS cluster namenode to contact.
d01612f3 1688.TP
523bad63
TK
1689.BI (libhdfs)port
1690The listening port of the HFDS cluster namenode.
d60e92d1 1691.TP
523bad63
TK
1692.BI (netsplice,net)port
1693The TCP or UDP port to bind to or connect to. If this is used with
1694\fBnumjobs\fR to spawn multiple instances of the same job type, then
1695this will be the starting port number since fio will use a range of
1696ports.
d60e92d1 1697.TP
609ac152
SB
1698.BI (rdma)port
1699The port to use for RDMA-CM communication. This should be the same
1700value on the client and the server side.
1701.TP
1702.BI (netsplice,net, rdma)hostname \fR=\fPstr
1703The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O.
1704If the job is a TCP listener or UDP reader, the hostname is not used
1705and must be omitted unless it is a valid UDP multicast address.
591e9e06 1706.TP
523bad63
TK
1707.BI (netsplice,net)interface \fR=\fPstr
1708The IP address of the network interface used to send or receive UDP
1709multicast.
ddf24e42 1710.TP
523bad63
TK
1711.BI (netsplice,net)ttl \fR=\fPint
1712Time\-to\-live value for outgoing UDP multicast packets. Default: 1.
d60e92d1 1713.TP
523bad63
TK
1714.BI (netsplice,net)nodelay \fR=\fPbool
1715Set TCP_NODELAY on TCP connections.
fa769d44 1716.TP
523bad63
TK
1717.BI (netsplice,net)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1718The network protocol to use. Accepted values are:
1719.RS
e76b1da4
JA
1720.RS
1721.TP
523bad63
TK
1722.B tcp
1723Transmission control protocol.
e76b1da4 1724.TP
523bad63
TK
1725.B tcpv6
1726Transmission control protocol V6.
e76b1da4 1727.TP
523bad63
TK
1728.B udp
1729User datagram protocol.
1730.TP
1731.B udpv6
1732User datagram protocol V6.
e76b1da4 1733.TP
523bad63
TK
1734.B unix
1735UNIX domain socket.
e76b1da4
JA
1736.RE
1737.P
523bad63
TK
1738When the protocol is TCP or UDP, the port must also be given, as well as the
1739hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1740normal \fBfilename\fR option should be used and the port is invalid.
1741.RE
1742.TP
1743.BI (netsplice,net)listen
1744For TCP network connections, tell fio to listen for incoming connections
1745rather than initiating an outgoing connection. The \fBhostname\fR must
1746be omitted if this option is used.
1747.TP
1748.BI (netsplice,net)pingpong
1749Normally a network writer will just continue writing data, and a network
1750reader will just consume packages. If `pingpong=1' is set, a writer will
1751send its normal payload to the reader, then wait for the reader to send the
1752same payload back. This allows fio to measure network latencies. The
1753submission and completion latencies then measure local time spent sending or
1754receiving, and the completion latency measures how long it took for the
1755other end to receive and send back. For UDP multicast traffic
1756`pingpong=1' should only be set for a single reader when multiple readers
1757are listening to the same address.
1758.TP
1759.BI (netsplice,net)window_size \fR=\fPint
1760Set the desired socket buffer size for the connection.
e76b1da4 1761.TP
523bad63
TK
1762.BI (netsplice,net)mss \fR=\fPint
1763Set the TCP maximum segment size (TCP_MAXSEG).
d60e92d1 1764.TP
523bad63
TK
1765.BI (e4defrag)donorname \fR=\fPstr
1766File will be used as a block donor (swap extents between files).
d60e92d1 1767.TP
523bad63
TK
1768.BI (e4defrag)inplace \fR=\fPint
1769Configure donor file blocks allocation strategy:
1770.RS
1771.RS
d60e92d1 1772.TP
523bad63
TK
1773.B 0
1774Default. Preallocate donor's file on init.
d60e92d1 1775.TP
523bad63
TK
1776.B 1
1777Allocate space immediately inside defragment event, and free right
1778after event.
1779.RE
1780.RE
d60e92d1 1781.TP
d5f9b0ea 1782.BI (rbd,rados)clustername \fR=\fPstr
523bad63 1783Specifies the name of the Ceph cluster.
92d42d69 1784.TP
523bad63
TK
1785.BI (rbd)rbdname \fR=\fPstr
1786Specifies the name of the RBD.
92d42d69 1787.TP
d5f9b0ea
IF
1788.BI (rbd,rados)pool \fR=\fPstr
1789Specifies the name of the Ceph pool containing RBD or RADOS data.
92d42d69 1790.TP
d5f9b0ea 1791.BI (rbd,rados)clientname \fR=\fPstr
523bad63
TK
1792Specifies the username (without the 'client.' prefix) used to access the
1793Ceph cluster. If the \fBclustername\fR is specified, the \fBclientname\fR shall be
1794the full *type.id* string. If no type. prefix is given, fio will add 'client.'
1795by default.
92d42d69 1796.TP
d5f9b0ea
IF
1797.BI (rbd,rados)busy_poll \fR=\fPbool
1798Poll store instead of waiting for completion. Usually this provides better
1799throughput at cost of higher(up to 100%) CPU utilization.
1800.TP
523bad63
TK
1801.BI (mtd)skip_bad \fR=\fPbool
1802Skip operations against known bad blocks.
8116fd24 1803.TP
523bad63
TK
1804.BI (libhdfs)hdfsdirectory
1805libhdfs will create chunk in this HDFS directory.
e0a04ac1 1806.TP
523bad63
TK
1807.BI (libhdfs)chunk_size
1808The size of the chunk to use for each file.
609ac152
SB
1809.TP
1810.BI (rdma)verb \fR=\fPstr
1811The RDMA verb to use on this side of the RDMA ioengine
1812connection. Valid values are write, read, send and recv. These
1813correspond to the equivalent RDMA verbs (e.g. write = rdma_write
1814etc.). Note that this only needs to be specified on the client side of
1815the connection. See the examples folder.
1816.TP
1817.BI (rdma)bindname \fR=\fPstr
1818The name to use to bind the local RDMA-CM connection to a local RDMA
1819device. This could be a hostname or an IPv4 or IPv6 address. On the
1820server side this will be passed into the rdma_bind_addr() function and
1821on the client site it will be used in the rdma_resolve_add()
1822function. This can be useful when multiple paths exist between the
1823client and the server or in certain loopback configurations.
52b81b7c
KD
1824.TP
1825.BI (sg)readfua \fR=\fPbool
1826With readfua option set to 1, read operations include the force
1827unit access (fua) flag. Default: 0.
1828.TP
1829.BI (sg)writefua \fR=\fPbool
1830With writefua option set to 1, write operations include the force
1831unit access (fua) flag. Default: 0.
2c3a9150
VF
1832.TP
1833.BI (sg)sg_write_mode \fR=\fPstr
1834Specify the type of write commands to issue. This option can take three
1835values:
1836.RS
1837.RS
1838.TP
1839.B write (default)
1840Write opcodes are issued as usual
1841.TP
1842.B verify
1843Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
1844directs the device to carry out a medium verification with no data
1845comparison. The writefua option is ignored with this selection.
1846.TP
1847.B same
1848Issue WRITE SAME commands. This transfers a single block to the device
1849and writes this same block of data to a contiguous sequence of LBAs
1850beginning at the specified offset. fio's block size parameter
1851specifies the amount of data written with each command. However, the
1852amount of data actually transferred to the device is equal to the
1853device's block (sector) size. For a device with 512 byte sectors,
1854blocksize=8k will write 16 sectors with each command. fio will still
1855generate 8k of data for each command butonly the first 512 bytes will
1856be used and transferred to the device. The writefua option is ignored
1857with this selection.
1858
523bad63
TK
1859.SS "I/O depth"
1860.TP
1861.BI iodepth \fR=\fPint
1862Number of I/O units to keep in flight against the file. Note that
1863increasing \fBiodepth\fR beyond 1 will not affect synchronous ioengines (except
1864for small degrees when \fBverify_async\fR is in use). Even async
1865engines may impose OS restrictions causing the desired depth not to be
1866achieved. This may happen on Linux when using libaio and not setting
1867`direct=1', since buffered I/O is not async on that OS. Keep an
1868eye on the I/O depth distribution in the fio output to verify that the
1869achieved depth is as expected. Default: 1.
1870.TP
1871.BI iodepth_batch_submit \fR=\fPint "\fR,\fP iodepth_batch" \fR=\fPint
1872This defines how many pieces of I/O to submit at once. It defaults to 1
1873which means that we submit each I/O as soon as it is available, but can be
1874raised to submit bigger batches of I/O at the time. If it is set to 0 the
1875\fBiodepth\fR value will be used.
1876.TP
1877.BI iodepth_batch_complete_min \fR=\fPint "\fR,\fP iodepth_batch_complete" \fR=\fPint
1878This defines how many pieces of I/O to retrieve at once. It defaults to 1
1879which means that we'll ask for a minimum of 1 I/O in the retrieval process
1880from the kernel. The I/O retrieval will go on until we hit the limit set by
1881\fBiodepth_low\fR. If this variable is set to 0, then fio will always
1882check for completed events before queuing more I/O. This helps reduce I/O
1883latency, at the cost of more retrieval system calls.
1884.TP
1885.BI iodepth_batch_complete_max \fR=\fPint
1886This defines maximum pieces of I/O to retrieve at once. This variable should
1887be used along with \fBiodepth_batch_complete_min\fR=\fIint\fR variable,
1888specifying the range of min and max amount of I/O which should be
1889retrieved. By default it is equal to \fBiodepth_batch_complete_min\fR
1890value. Example #1:
e0a04ac1 1891.RS
e0a04ac1 1892.RS
e0a04ac1 1893.P
523bad63
TK
1894.PD 0
1895iodepth_batch_complete_min=1
e0a04ac1 1896.P
523bad63
TK
1897iodepth_batch_complete_max=<iodepth>
1898.PD
e0a04ac1
JA
1899.RE
1900.P
523bad63
TK
1901which means that we will retrieve at least 1 I/O and up to the whole
1902submitted queue depth. If none of I/O has been completed yet, we will wait.
1903Example #2:
e8b1961d 1904.RS
523bad63
TK
1905.P
1906.PD 0
1907iodepth_batch_complete_min=0
1908.P
1909iodepth_batch_complete_max=<iodepth>
1910.PD
e8b1961d
JA
1911.RE
1912.P
523bad63
TK
1913which means that we can retrieve up to the whole submitted queue depth, but
1914if none of I/O has been completed yet, we will NOT wait and immediately exit
1915the system call. In this example we simply do polling.
1916.RE
e8b1961d 1917.TP
523bad63
TK
1918.BI iodepth_low \fR=\fPint
1919The low water mark indicating when to start filling the queue
1920again. Defaults to the same as \fBiodepth\fR, meaning that fio will
1921attempt to keep the queue full at all times. If \fBiodepth\fR is set to
1922e.g. 16 and \fBiodepth_low\fR is set to 4, then after fio has filled the queue of
192316 requests, it will let the depth drain down to 4 before starting to fill
1924it again.
d60e92d1 1925.TP
523bad63
TK
1926.BI serialize_overlap \fR=\fPbool
1927Serialize in-flight I/Os that might otherwise cause or suffer from data races.
1928When two or more I/Os are submitted simultaneously, there is no guarantee that
1929the I/Os will be processed or completed in the submitted order. Further, if
1930two or more of those I/Os are writes, any overlapping region between them can
1931become indeterminate/undefined on certain storage. These issues can cause
1932verification to fail erratically when at least one of the racing I/Os is
1933changing data and the overlapping region has a non-zero size. Setting
1934\fBserialize_overlap\fR tells fio to avoid provoking this behavior by explicitly
1935serializing in-flight I/Os that have a non-zero overlap. Note that setting
1936this option can reduce both performance and the \fBiodepth\fR achieved.
1937Additionally this option does not work when \fBio_submit_mode\fR is set to
1938offload. Default: false.
d60e92d1 1939.TP
523bad63
TK
1940.BI io_submit_mode \fR=\fPstr
1941This option controls how fio submits the I/O to the I/O engine. The default
1942is `inline', which means that the fio job threads submit and reap I/O
1943directly. If set to `offload', the job threads will offload I/O submission
1944to a dedicated pool of I/O threads. This requires some coordination and thus
1945has a bit of extra overhead, especially for lower queue depth I/O where it
1946can increase latencies. The benefit is that fio can manage submission rates
1947independently of the device completion rates. This avoids skewed latency
1948reporting if I/O gets backed up on the device side (the coordinated omission
1949problem).
1950.SS "I/O rate"
d60e92d1 1951.TP
523bad63
TK
1952.BI thinktime \fR=\fPtime
1953Stall the job for the specified period of time after an I/O has completed before issuing the
1954next. May be used to simulate processing being done by an application.
1955When the unit is omitted, the value is interpreted in microseconds. See
1956\fBthinktime_blocks\fR and \fBthinktime_spin\fR.
d60e92d1 1957.TP
523bad63
TK
1958.BI thinktime_spin \fR=\fPtime
1959Only valid if \fBthinktime\fR is set \- pretend to spend CPU time doing
1960something with the data received, before falling back to sleeping for the
1961rest of the period specified by \fBthinktime\fR. When the unit is
1962omitted, the value is interpreted in microseconds.
d60e92d1
AC
1963.TP
1964.BI thinktime_blocks \fR=\fPint
523bad63
TK
1965Only valid if \fBthinktime\fR is set \- control how many blocks to issue,
1966before waiting \fBthinktime\fR usecs. If not set, defaults to 1 which will make
1967fio wait \fBthinktime\fR usecs after every block. This effectively makes any
1968queue depth setting redundant, since no more than 1 I/O will be queued
1969before we have to complete it and do our \fBthinktime\fR. In other words, this
1970setting effectively caps the queue depth if the latter is larger.
d60e92d1 1971.TP
6d500c2e 1972.BI rate \fR=\fPint[,int][,int]
523bad63
TK
1973Cap the bandwidth used by this job. The number is in bytes/sec, the normal
1974suffix rules apply. Comma\-separated values may be specified for reads,
1975writes, and trims as described in \fBblocksize\fR.
1976.RS
1977.P
1978For example, using `rate=1m,500k' would limit reads to 1MiB/sec and writes to
1979500KiB/sec. Capping only reads or writes can be done with `rate=,500k' or
1980`rate=500k,' where the former will only limit writes (to 500KiB/sec) and the
1981latter will only limit reads.
1982.RE
d60e92d1 1983.TP
6d500c2e 1984.BI rate_min \fR=\fPint[,int][,int]
523bad63
TK
1985Tell fio to do whatever it can to maintain at least this bandwidth. Failing
1986to meet this requirement will cause the job to exit. Comma\-separated values
1987may be specified for reads, writes, and trims as described in
1988\fBblocksize\fR.
d60e92d1 1989.TP
6d500c2e 1990.BI rate_iops \fR=\fPint[,int][,int]
523bad63
TK
1991Cap the bandwidth to this number of IOPS. Basically the same as
1992\fBrate\fR, just specified independently of bandwidth. If the job is
1993given a block size range instead of a fixed value, the smallest block size
1994is used as the metric. Comma\-separated values may be specified for reads,
1995writes, and trims as described in \fBblocksize\fR.
d60e92d1 1996.TP
6d500c2e 1997.BI rate_iops_min \fR=\fPint[,int][,int]
523bad63
TK
1998If fio doesn't meet this rate of I/O, it will cause the job to exit.
1999Comma\-separated values may be specified for reads, writes, and trims as
2000described in \fBblocksize\fR.
d60e92d1 2001.TP
6de65959 2002.BI rate_process \fR=\fPstr
523bad63
TK
2003This option controls how fio manages rated I/O submissions. The default is
2004`linear', which submits I/O in a linear fashion with fixed delays between
2005I/Os that gets adjusted based on I/O completion rates. If this is set to
2006`poisson', fio will submit I/O based on a more real world random request
6de65959 2007flow, known as the Poisson process
523bad63 2008(\fIhttps://en.wikipedia.org/wiki/Poisson_point_process\fR). The lambda will be
5d02b083 200910^6 / IOPS for the given workload.
1a9bf814
JA
2010.TP
2011.BI rate_ignore_thinktime \fR=\fPbool
2012By default, fio will attempt to catch up to the specified rate setting, if any
2013kind of thinktime setting was used. If this option is set, then fio will
2014ignore the thinktime and continue doing IO at the specified rate, instead of
2015entering a catch-up mode after thinktime is done.
523bad63 2016.SS "I/O latency"
ff6bb260 2017.TP
523bad63 2018.BI latency_target \fR=\fPtime
3e260a46 2019If set, fio will attempt to find the max performance point that the given
523bad63
TK
2020workload will run at while maintaining a latency below this target. When
2021the unit is omitted, the value is interpreted in microseconds. See
2022\fBlatency_window\fR and \fBlatency_percentile\fR.
3e260a46 2023.TP
523bad63 2024.BI latency_window \fR=\fPtime
3e260a46 2025Used with \fBlatency_target\fR to specify the sample window that the job
523bad63
TK
2026is run at varying queue depths to test the performance. When the unit is
2027omitted, the value is interpreted in microseconds.
3e260a46
JA
2028.TP
2029.BI latency_percentile \fR=\fPfloat
523bad63
TK
2030The percentage of I/Os that must fall within the criteria specified by
2031\fBlatency_target\fR and \fBlatency_window\fR. If not set, this
2032defaults to 100.0, meaning that all I/Os must be equal or below to the value
2033set by \fBlatency_target\fR.
2034.TP
2035.BI max_latency \fR=\fPtime
2036If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2037maximum latency. When the unit is omitted, the value is interpreted in
2038microseconds.
2039.TP
2040.BI rate_cycle \fR=\fPint
2041Average bandwidth for \fBrate\fR and \fBrate_min\fR over this number
2042of milliseconds. Defaults to 1000.
2043.SS "I/O replay"
2044.TP
2045.BI write_iolog \fR=\fPstr
2046Write the issued I/O patterns to the specified file. See
2047\fBread_iolog\fR. Specify a separate file for each job, otherwise the
2048iologs will be interspersed and the file may be corrupt.
2049.TP
2050.BI read_iolog \fR=\fPstr
2051Open an iolog with the specified filename and replay the I/O patterns it
2052contains. This can be used to store a workload and replay it sometime
2053later. The iolog given may also be a blktrace binary file, which allows fio
2054to replay a workload captured by blktrace. See
2055\fBblktrace\fR\|(8) for how to capture such logging data. For blktrace
2056replay, the file needs to be turned into a blkparse binary data file first
2057(`blkparse <device> \-o /dev/null \-d file_for_fio.bin').
3e260a46 2058.TP
98e7161c
AK
2059.BI read_iolog_chunked \fR=\fPbool
2060Determines how iolog is read. If false (default) entire \fBread_iolog\fR will
2061be read at once. If selected true, input from iolog will be read gradually.
2062Useful when iolog is very large, or it is generated.
2063.TP
523bad63
TK
2064.BI replay_no_stall \fR=\fPbool
2065When replaying I/O with \fBread_iolog\fR the default behavior is to
2066attempt to respect the timestamps within the log and replay them with the
2067appropriate delay between IOPS. By setting this variable fio will not
2068respect the timestamps and attempt to replay them as fast as possible while
2069still respecting ordering. The result is the same I/O pattern to a given
2070device, but different timings.
2071.TP
6dd7fa77
JA
2072.BI replay_time_scale \fR=\fPint
2073When replaying I/O with \fBread_iolog\fR, fio will honor the original timing
2074in the trace. With this option, it's possible to scale the time. It's a
2075percentage option, if set to 50 it means run at 50% the original IO rate in
2076the trace. If set to 200, run at twice the original IO rate. Defaults to 100.
2077.TP
523bad63
TK
2078.BI replay_redirect \fR=\fPstr
2079While replaying I/O patterns using \fBread_iolog\fR the default behavior
2080is to replay the IOPS onto the major/minor device that each IOP was recorded
2081from. This is sometimes undesirable because on a different machine those
2082major/minor numbers can map to a different device. Changing hardware on the
2083same system can also result in a different major/minor mapping.
2084\fBreplay_redirect\fR causes all I/Os to be replayed onto the single specified
2085device regardless of the device it was recorded
2086from. i.e. `replay_redirect=/dev/sdc' would cause all I/O
2087in the blktrace or iolog to be replayed onto `/dev/sdc'. This means
2088multiple devices will be replayed onto a single device, if the trace
2089contains multiple devices. If you want multiple devices to be replayed
2090concurrently to multiple redirected devices you must blkparse your trace
2091into separate traces and replay them with independent fio invocations.
2092Unfortunately this also breaks the strict time ordering between multiple
2093device accesses.
2094.TP
2095.BI replay_align \fR=\fPint
2096Force alignment of I/O offsets and lengths in a trace to this power of 2
2097value.
2098.TP
2099.BI replay_scale \fR=\fPint
2100Scale sector offsets down by this factor when replaying traces.
2101.SS "Threads, processes and job synchronization"
2102.TP
38f68906
JA
2103.BI replay_skip \fR=\fPstr
2104Sometimes it's useful to skip certain IO types in a replay trace. This could
2105be, for instance, eliminating the writes in the trace. Or not replaying the
2106trims/discards, if you are redirecting to a device that doesn't support them.
2107This option takes a comma separated list of read, write, trim, sync.
2108.TP
523bad63
TK
2109.BI thread
2110Fio defaults to creating jobs by using fork, however if this option is
2111given, fio will create jobs by using POSIX Threads' function
2112\fBpthread_create\fR\|(3) to create threads instead.
2113.TP
2114.BI wait_for \fR=\fPstr
2115If set, the current job won't be started until all workers of the specified
2116waitee job are done.
2117.\" ignore blank line here from HOWTO as it looks normal without it
2118\fBwait_for\fR operates on the job name basis, so there are a few
2119limitations. First, the waitee must be defined prior to the waiter job
2120(meaning no forward references). Second, if a job is being referenced as a
2121waitee, it must have a unique name (no duplicate waitees).
2122.TP
2123.BI nice \fR=\fPint
2124Run the job with the given nice value. See man \fBnice\fR\|(2).
2125.\" ignore blank line here from HOWTO as it looks normal without it
2126On Windows, values less than \-15 set the process class to "High"; \-1 through
2127\-15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2128priority class.
2129.TP
2130.BI prio \fR=\fPint
2131Set the I/O priority value of this job. Linux limits us to a positive value
2132between 0 and 7, with 0 being the highest. See man
2133\fBionice\fR\|(1). Refer to an appropriate manpage for other operating
2134systems since meaning of priority may differ.
2135.TP
2136.BI prioclass \fR=\fPint
2137Set the I/O priority class. See man \fBionice\fR\|(1).
15501535 2138.TP
d60e92d1 2139.BI cpus_allowed \fR=\fPstr
523bad63 2140Controls the same options as \fBcpumask\fR, but accepts a textual
b570e037
SW
2141specification of the permitted CPUs instead and CPUs are indexed from 0. So
2142to use CPUs 0 and 5 you would specify `cpus_allowed=0,5'. This option also
2143allows a range of CPUs to be specified \-\- say you wanted a binding to CPUs
21440, 5, and 8 to 15, you would set `cpus_allowed=0,5,8\-15'.
2145.RS
2146.P
2147On Windows, when `cpus_allowed' is unset only CPUs from fio's current
2148processor group will be used and affinity settings are inherited from the
2149system. An fio build configured to target Windows 7 makes options that set
2150CPUs processor group aware and values will set both the processor group
2151and a CPU from within that group. For example, on a system where processor
2152group 0 has 40 CPUs and processor group 1 has 32 CPUs, `cpus_allowed'
2153values between 0 and 39 will bind CPUs from processor group 0 and
2154`cpus_allowed' values between 40 and 71 will bind CPUs from processor
2155group 1. When using `cpus_allowed_policy=shared' all CPUs specified by a
2156single `cpus_allowed' option must be from the same processor group. For
2157Windows fio builds not built for Windows 7, CPUs will only be selected from
2158(and be relative to) whatever processor group fio happens to be running in
2159and CPUs from other processor groups cannot be used.
2160.RE
d60e92d1 2161.TP
c2acfbac 2162.BI cpus_allowed_policy \fR=\fPstr
523bad63
TK
2163Set the policy of how fio distributes the CPUs specified by
2164\fBcpus_allowed\fR or \fBcpumask\fR. Two policies are supported:
c2acfbac
JA
2165.RS
2166.RS
2167.TP
2168.B shared
2169All jobs will share the CPU set specified.
2170.TP
2171.B split
2172Each job will get a unique CPU from the CPU set.
2173.RE
2174.P
523bad63
TK
2175\fBshared\fR is the default behavior, if the option isn't specified. If
2176\fBsplit\fR is specified, then fio will will assign one cpu per job. If not
2177enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2178in the set.
c2acfbac 2179.RE
c2acfbac 2180.TP
b570e037
SW
2181.BI cpumask \fR=\fPint
2182Set the CPU affinity of this job. The parameter given is a bit mask of
2183allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2184and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2185\fBsched_setaffinity\fR\|(2). This may not work on all supported
2186operating systems or kernel versions. This option doesn't work well for a
2187higher CPU count than what you can store in an integer mask, so it can only
2188control cpus 1\-32. For boxes with larger CPU counts, use
2189\fBcpus_allowed\fR.
2190.TP
d0b937ed 2191.BI numa_cpu_nodes \fR=\fPstr
cecbfd47 2192Set this job running on specified NUMA nodes' CPUs. The arguments allow
523bad63
TK
2193comma delimited list of cpu numbers, A\-B ranges, or `all'. Note, to enable
2194NUMA options support, fio must be built on a system with libnuma\-dev(el)
2195installed.
d0b937ed
YR
2196.TP
2197.BI numa_mem_policy \fR=\fPstr
523bad63
TK
2198Set this job's memory policy and corresponding NUMA nodes. Format of the
2199arguments:
39c7a2ca
VF
2200.RS
2201.RS
523bad63
TK
2202.P
2203<mode>[:<nodelist>]
39c7a2ca 2204.RE
523bad63 2205.P
f1dd3fb1 2206`mode' is one of the following memory policies: `default', `prefer',
523bad63
TK
2207`bind', `interleave' or `local'. For `default' and `local' memory
2208policies, no node needs to be specified. For `prefer', only one node is
2209allowed. For `bind' and `interleave' the `nodelist' may be as
2210follows: a comma delimited list of numbers, A\-B ranges, or `all'.
39c7a2ca
VF
2211.RE
2212.TP
523bad63
TK
2213.BI cgroup \fR=\fPstr
2214Add job to this control group. If it doesn't exist, it will be created. The
2215system must have a mounted cgroup blkio mount point for this to work. If
2216your system doesn't have it mounted, you can do so with:
d60e92d1
AC
2217.RS
2218.RS
d60e92d1 2219.P
523bad63
TK
2220# mount \-t cgroup \-o blkio none /cgroup
2221.RE
d60e92d1
AC
2222.RE
2223.TP
523bad63
TK
2224.BI cgroup_weight \fR=\fPint
2225Set the weight of the cgroup to this value. See the documentation that comes
2226with the kernel, allowed values are in the range of 100..1000.
d60e92d1 2227.TP
523bad63
TK
2228.BI cgroup_nodelete \fR=\fPbool
2229Normally fio will delete the cgroups it has created after the job
2230completion. To override this behavior and to leave cgroups around after the
2231job completion, set `cgroup_nodelete=1'. This can be useful if one wants
2232to inspect various cgroup files after job completion. Default: false.
c8eeb9df 2233.TP
523bad63
TK
2234.BI flow_id \fR=\fPint
2235The ID of the flow. If not specified, it defaults to being a global
2236flow. See \fBflow\fR.
d60e92d1 2237.TP
523bad63
TK
2238.BI flow \fR=\fPint
2239Weight in token\-based flow control. If this value is used, then there is
2240a 'flow counter' which is used to regulate the proportion of activity between
2241two or more jobs. Fio attempts to keep this flow counter near zero. The
2242\fBflow\fR parameter stands for how much should be added or subtracted to the
2243flow counter on each iteration of the main I/O loop. That is, if one job has
2244`flow=8' and another job has `flow=\-1', then there will be a roughly 1:8
2245ratio in how much one runs vs the other.
d60e92d1 2246.TP
523bad63
TK
2247.BI flow_watermark \fR=\fPint
2248The maximum value that the absolute value of the flow counter is allowed to
2249reach before the job must wait for a lower value of the counter.
6b7f6851 2250.TP
523bad63
TK
2251.BI flow_sleep \fR=\fPint
2252The period of time, in microseconds, to wait after the flow watermark has
2253been exceeded before retrying operations.
25460cf6 2254.TP
523bad63
TK
2255.BI stonewall "\fR,\fB wait_for_previous"
2256Wait for preceding jobs in the job file to exit, before starting this
2257one. Can be used to insert serialization points in the job file. A stone
2258wall also implies starting a new reporting group, see
2259\fBgroup_reporting\fR.
2378826d 2260.TP
523bad63
TK
2261.BI exitall
2262By default, fio will continue running all other jobs when one job finishes
2263but sometimes this is not the desired action. Setting \fBexitall\fR will
2264instead make fio terminate all other jobs when one job finishes.
e81ecca3 2265.TP
523bad63
TK
2266.BI exec_prerun \fR=\fPstr
2267Before running this job, issue the command specified through
2268\fBsystem\fR\|(3). Output is redirected in a file called `jobname.prerun.txt'.
e9f48479 2269.TP
523bad63
TK
2270.BI exec_postrun \fR=\fPstr
2271After the job completes, issue the command specified though
2272\fBsystem\fR\|(3). Output is redirected in a file called `jobname.postrun.txt'.
d60e92d1 2273.TP
523bad63
TK
2274.BI uid \fR=\fPint
2275Instead of running as the invoking user, set the user ID to this value
2276before the thread/process does any work.
39c1c323 2277.TP
523bad63
TK
2278.BI gid \fR=\fPint
2279Set group ID, see \fBuid\fR.
2280.SS "Verification"
d60e92d1 2281.TP
589e88b7 2282.BI verify_only
523bad63 2283Do not perform specified workload, only verify data still matches previous
5e4c7118 2284invocation of this workload. This option allows one to check data multiple
523bad63
TK
2285times at a later date without overwriting it. This option makes sense only
2286for workloads that write data, and does not support workloads with the
5e4c7118
JA
2287\fBtime_based\fR option set.
2288.TP
d60e92d1 2289.BI do_verify \fR=\fPbool
523bad63
TK
2290Run the verify phase after a write phase. Only valid if \fBverify\fR is
2291set. Default: true.
d60e92d1
AC
2292.TP
2293.BI verify \fR=\fPstr
523bad63
TK
2294If writing to a file, fio can verify the file contents after each iteration
2295of the job. Each verification method also implies verification of special
2296header, which is written to the beginning of each block. This header also
2297includes meta information, like offset of the block, block number, timestamp
2298when block was written, etc. \fBverify\fR can be combined with
2299\fBverify_pattern\fR option. The allowed values are:
d60e92d1
AC
2300.RS
2301.RS
2302.TP
523bad63
TK
2303.B md5
2304Use an md5 sum of the data area and store it in the header of
2305each block.
2306.TP
2307.B crc64
2308Use an experimental crc64 sum of the data area and store it in the
2309header of each block.
2310.TP
2311.B crc32c
2312Use a crc32c sum of the data area and store it in the header of
2313each block. This will automatically use hardware acceleration
2314(e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
2315fall back to software crc32c if none is found. Generally the
f1dd3fb1 2316fastest checksum fio supports when hardware accelerated.
523bad63
TK
2317.TP
2318.B crc32c\-intel
2319Synonym for crc32c.
2320.TP
2321.B crc32
2322Use a crc32 sum of the data area and store it in the header of each
2323block.
2324.TP
2325.B crc16
2326Use a crc16 sum of the data area and store it in the header of each
2327block.
2328.TP
2329.B crc7
2330Use a crc7 sum of the data area and store it in the header of each
2331block.
2332.TP
2333.B xxhash
2334Use xxhash as the checksum function. Generally the fastest software
2335checksum that fio supports.
2336.TP
2337.B sha512
2338Use sha512 as the checksum function.
2339.TP
2340.B sha256
2341Use sha256 as the checksum function.
2342.TP
2343.B sha1
2344Use optimized sha1 as the checksum function.
2345.TP
2346.B sha3\-224
2347Use optimized sha3\-224 as the checksum function.
2348.TP
2349.B sha3\-256
2350Use optimized sha3\-256 as the checksum function.
2351.TP
2352.B sha3\-384
2353Use optimized sha3\-384 as the checksum function.
2354.TP
2355.B sha3\-512
2356Use optimized sha3\-512 as the checksum function.
d60e92d1
AC
2357.TP
2358.B meta
523bad63
TK
2359This option is deprecated, since now meta information is included in
2360generic verification header and meta verification happens by
2361default. For detailed information see the description of the
2362\fBverify\fR setting. This option is kept because of
2363compatibility's sake with old configurations. Do not use it.
d60e92d1 2364.TP
59245381 2365.B pattern
523bad63
TK
2366Verify a strict pattern. Normally fio includes a header with some
2367basic information and checksumming, but if this option is set, only
2368the specific pattern set with \fBverify_pattern\fR is verified.
59245381 2369.TP
d60e92d1 2370.B null
523bad63
TK
2371Only pretend to verify. Useful for testing internals with
2372`ioengine=null', not for much else.
d60e92d1 2373.RE
523bad63
TK
2374.P
2375This option can be used for repeated burn\-in tests of a system to make sure
2376that the written data is also correctly read back. If the data direction
2377given is a read or random read, fio will assume that it should verify a
2378previously written file. If the data direction includes any form of write,
2379the verify will be of the newly written data.
47e6a6e5
SW
2380.P
2381To avoid false verification errors, do not use the norandommap option when
2382verifying data with async I/O engines and I/O depths > 1. Or use the
2383norandommap and the lfsr random generator together to avoid writing to the
2384same offset with muliple outstanding I/Os.
d60e92d1
AC
2385.RE
2386.TP
f7fa2653 2387.BI verify_offset \fR=\fPint
d60e92d1 2388Swap the verification header with data somewhere else in the block before
523bad63 2389writing. It is swapped back before verifying.
d60e92d1 2390.TP
f7fa2653 2391.BI verify_interval \fR=\fPint
523bad63
TK
2392Write the verification header at a finer granularity than the
2393\fBblocksize\fR. It will be written for chunks the size of
2394\fBverify_interval\fR. \fBblocksize\fR should divide this evenly.
d60e92d1 2395.TP
996093bb 2396.BI verify_pattern \fR=\fPstr
523bad63
TK
2397If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2398filling with totally random bytes, but sometimes it's interesting to fill
2399with a known pattern for I/O verification purposes. Depending on the width
2400of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
2401be either a decimal or a hex number). The \fBverify_pattern\fR if larger than
2402a 32\-bit quantity has to be a hex number that starts with either "0x" or
2403"0X". Use with \fBverify\fR. Also, \fBverify_pattern\fR supports %o
2404format, which means that for each block offset will be written and then
2405verified back, e.g.:
2fa5a241
RP
2406.RS
2407.RS
523bad63
TK
2408.P
2409verify_pattern=%o
2fa5a241 2410.RE
523bad63 2411.P
2fa5a241 2412Or use combination of everything:
2fa5a241 2413.RS
523bad63
TK
2414.P
2415verify_pattern=0xff%o"abcd"\-12
2fa5a241
RP
2416.RE
2417.RE
996093bb 2418.TP
d60e92d1 2419.BI verify_fatal \fR=\fPbool
523bad63
TK
2420Normally fio will keep checking the entire contents before quitting on a
2421block verification failure. If this option is set, fio will exit the job on
2422the first observed failure. Default: false.
d60e92d1 2423.TP
b463e936 2424.BI verify_dump \fR=\fPbool
523bad63
TK
2425If set, dump the contents of both the original data block and the data block
2426we read off disk to files. This allows later analysis to inspect just what
2427kind of data corruption occurred. Off by default.
b463e936 2428.TP
e8462bd8 2429.BI verify_async \fR=\fPint
523bad63
TK
2430Fio will normally verify I/O inline from the submitting thread. This option
2431takes an integer describing how many async offload threads to create for I/O
2432verification instead, causing fio to offload the duty of verifying I/O
2433contents to one or more separate threads. If using this offload option, even
2434sync I/O engines can benefit from using an \fBiodepth\fR setting higher
2435than 1, as it allows them to have I/O in flight while verifies are running.
2436Defaults to 0 async threads, i.e. verification is not asynchronous.
e8462bd8
JA
2437.TP
2438.BI verify_async_cpus \fR=\fPstr
523bad63
TK
2439Tell fio to set the given CPU affinity on the async I/O verification
2440threads. See \fBcpus_allowed\fR for the format used.
e8462bd8 2441.TP
6f87418f
JA
2442.BI verify_backlog \fR=\fPint
2443Fio will normally verify the written contents of a job that utilizes verify
2444once that job has completed. In other words, everything is written then
2445everything is read back and verified. You may want to verify continually
523bad63
TK
2446instead for a variety of reasons. Fio stores the meta data associated with
2447an I/O block in memory, so for large verify workloads, quite a bit of memory
2448would be used up holding this meta data. If this option is enabled, fio will
2449write only N blocks before verifying these blocks.
6f87418f
JA
2450.TP
2451.BI verify_backlog_batch \fR=\fPint
523bad63
TK
2452Control how many blocks fio will verify if \fBverify_backlog\fR is
2453set. If not set, will default to the value of \fBverify_backlog\fR
2454(meaning the entire queue is read back and verified). If
2455\fBverify_backlog_batch\fR is less than \fBverify_backlog\fR then not all
2456blocks will be verified, if \fBverify_backlog_batch\fR is larger than
2457\fBverify_backlog\fR, some blocks will be verified more than once.
2458.TP
2459.BI verify_state_save \fR=\fPbool
2460When a job exits during the write phase of a verify workload, save its
2461current state. This allows fio to replay up until that point, if the verify
2462state is loaded for the verify read phase. The format of the filename is,
2463roughly:
2464.RS
2465.RS
2466.P
2467<type>\-<jobname>\-<jobindex>\-verify.state.
2468.RE
2469.P
2470<type> is "local" for a local run, "sock" for a client/server socket
2471connection, and "ip" (192.168.0.1, for instance) for a networked
2472client/server connection. Defaults to true.
2473.RE
2474.TP
2475.BI verify_state_load \fR=\fPbool
2476If a verify termination trigger was used, fio stores the current write state
2477of each thread. This can be used at verification time so that fio knows how
2478far it should verify. Without this information, fio will run a full
2479verification pass, according to the settings in the job file used. Default
2480false.
6f87418f 2481.TP
fa769d44
SW
2482.BI trim_percentage \fR=\fPint
2483Number of verify blocks to discard/trim.
2484.TP
2485.BI trim_verify_zero \fR=\fPbool
523bad63 2486Verify that trim/discarded blocks are returned as zeros.
fa769d44
SW
2487.TP
2488.BI trim_backlog \fR=\fPint
523bad63 2489Verify that trim/discarded blocks are returned as zeros.
fa769d44
SW
2490.TP
2491.BI trim_backlog_batch \fR=\fPint
523bad63 2492Trim this number of I/O blocks.
fa769d44
SW
2493.TP
2494.BI experimental_verify \fR=\fPbool
2495Enable experimental verification.
523bad63 2496.SS "Steady state"
fa769d44 2497.TP
523bad63
TK
2498.BI steadystate \fR=\fPstr:float "\fR,\fP ss" \fR=\fPstr:float
2499Define the criterion and limit for assessing steady state performance. The
2500first parameter designates the criterion whereas the second parameter sets
2501the threshold. When the criterion falls below the threshold for the
2502specified duration, the job will stop. For example, `iops_slope:0.1%' will
2503direct fio to terminate the job when the least squares regression slope
2504falls below 0.1% of the mean IOPS. If \fBgroup_reporting\fR is enabled
2505this will apply to all jobs in the group. Below is the list of available
2506steady state assessment criteria. All assessments are carried out using only
2507data from the rolling collection window. Threshold limits can be expressed
2508as a fixed value or as a percentage of the mean in the collection window.
2509.RS
2510.RS
d60e92d1 2511.TP
523bad63
TK
2512.B iops
2513Collect IOPS data. Stop the job if all individual IOPS measurements
2514are within the specified limit of the mean IOPS (e.g., `iops:2'
2515means that all individual IOPS values must be within 2 of the mean,
2516whereas `iops:0.2%' means that all individual IOPS values must be
2517within 0.2% of the mean IOPS to terminate the job).
d60e92d1 2518.TP
523bad63
TK
2519.B iops_slope
2520Collect IOPS data and calculate the least squares regression
2521slope. Stop the job if the slope falls below the specified limit.
d60e92d1 2522.TP
523bad63
TK
2523.B bw
2524Collect bandwidth data. Stop the job if all individual bandwidth
2525measurements are within the specified limit of the mean bandwidth.
64bbb865 2526.TP
523bad63
TK
2527.B bw_slope
2528Collect bandwidth data and calculate the least squares regression
2529slope. Stop the job if the slope falls below the specified limit.
2530.RE
2531.RE
d1c46c04 2532.TP
523bad63
TK
2533.BI steadystate_duration \fR=\fPtime "\fR,\fP ss_dur" \fR=\fPtime
2534A rolling window of this duration will be used to judge whether steady state
2535has been reached. Data will be collected once per second. The default is 0
2536which disables steady state detection. When the unit is omitted, the
2537value is interpreted in seconds.
0c63576e 2538.TP
523bad63
TK
2539.BI steadystate_ramp_time \fR=\fPtime "\fR,\fP ss_ramp" \fR=\fPtime
2540Allow the job to run for the specified duration before beginning data
2541collection for checking the steady state job termination criterion. The
2542default is 0. When the unit is omitted, the value is interpreted in seconds.
2543.SS "Measurements and reporting"
0c63576e 2544.TP
3a5db920
JA
2545.BI per_job_logs \fR=\fPbool
2546If set, this generates bw/clat/iops log with per file private filenames. If
523bad63
TK
2547not set, jobs with identical names will share the log filename. Default:
2548true.
2549.TP
2550.BI group_reporting
2551It may sometimes be interesting to display statistics for groups of jobs as
2552a whole instead of for each individual job. This is especially true if
2553\fBnumjobs\fR is used; looking at individual thread/process output
2554quickly becomes unwieldy. To see the final report per\-group instead of
2555per\-job, use \fBgroup_reporting\fR. Jobs in a file will be part of the
2556same reporting group, unless if separated by a \fBstonewall\fR, or by
2557using \fBnew_group\fR.
2558.TP
2559.BI new_group
2560Start a new reporting group. See: \fBgroup_reporting\fR. If not given,
2561all jobs in a file will be part of the same reporting group, unless
2562separated by a \fBstonewall\fR.
2563.TP
2564.BI stats \fR=\fPbool
2565By default, fio collects and shows final output results for all jobs
2566that run. If this option is set to 0, then fio will ignore it in
2567the final stat output.
3a5db920 2568.TP
836bad52 2569.BI write_bw_log \fR=\fPstr
523bad63 2570If given, write a bandwidth log for this job. Can be used to store data of
074f0817 2571the bandwidth of the jobs in their lifetime.
523bad63 2572.RS
074f0817
SW
2573.P
2574If no str argument is given, the default filename of
2575`jobname_type.x.log' is used. Even when the argument is given, fio
2576will still append the type of log. So if one specifies:
523bad63
TK
2577.RS
2578.P
074f0817 2579write_bw_log=foo
523bad63
TK
2580.RE
2581.P
074f0817
SW
2582The actual log name will be `foo_bw.x.log' where `x' is the index
2583of the job (1..N, where N is the number of jobs). If
2584\fBper_job_logs\fR is false, then the filename will not include the
2585`.x` job index.
2586.P
2587The included \fBfio_generate_plots\fR script uses gnuplot to turn these
2588text files into nice graphs. See the \fBLOG FILE FORMATS\fR section for how data is
2589structured within the file.
523bad63 2590.RE
901bb994 2591.TP
074f0817
SW
2592.BI write_lat_log \fR=\fPstr
2593Same as \fBwrite_bw_log\fR, except this option creates I/O
2594submission (e.g., `name_slat.x.log'), completion (e.g.,
2595`name_clat.x.log'), and total (e.g., `name_lat.x.log') latency
2596files instead. See \fBwrite_bw_log\fR for details about the
2597filename format and the \fBLOG FILE FORMATS\fR section for how data is structured
2598within the files.
2599.TP
1e613c9c 2600.BI write_hist_log \fR=\fPstr
074f0817
SW
2601Same as \fBwrite_bw_log\fR but writes an I/O completion latency
2602histogram file (e.g., `name_hist.x.log') instead. Note that this
2603file will be empty unless \fBlog_hist_msec\fR has also been set.
2604See \fBwrite_bw_log\fR for details about the filename format and
2605the \fBLOG FILE FORMATS\fR section for how data is structured
2606within the file.
1e613c9c 2607.TP
c8eeb9df 2608.BI write_iops_log \fR=\fPstr
074f0817
SW
2609Same as \fBwrite_bw_log\fR, but writes an IOPS file (e.g.
2610`name_iops.x.log') instead. See \fBwrite_bw_log\fR for
2611details about the filename format and the \fBLOG FILE FORMATS\fR section for how data
2612is structured within the file.
c8eeb9df 2613.TP
b8bc8cba
JA
2614.BI log_avg_msec \fR=\fPint
2615By default, fio will log an entry in the iops, latency, or bw log for every
523bad63 2616I/O that completes. When writing to the disk log, that can quickly grow to a
b8bc8cba 2617very large size. Setting this option makes fio average the each log entry
e6989e10 2618over the specified period of time, reducing the resolution of the log. See
523bad63
TK
2619\fBlog_max_value\fR as well. Defaults to 0, logging all entries.
2620Also see \fBLOG FILE FORMATS\fR section.
b8bc8cba 2621.TP
1e613c9c 2622.BI log_hist_msec \fR=\fPint
523bad63
TK
2623Same as \fBlog_avg_msec\fR, but logs entries for completion latency
2624histograms. Computing latency percentiles from averages of intervals using
2625\fBlog_avg_msec\fR is inaccurate. Setting this option makes fio log
2626histogram entries over the specified period of time, reducing log sizes for
2627high IOPS devices while retaining percentile accuracy. See
074f0817
SW
2628\fBlog_hist_coarseness\fR and \fBwrite_hist_log\fR as well.
2629Defaults to 0, meaning histogram logging is disabled.
1e613c9c
KC
2630.TP
2631.BI log_hist_coarseness \fR=\fPint
523bad63
TK
2632Integer ranging from 0 to 6, defining the coarseness of the resolution of
2633the histogram logs enabled with \fBlog_hist_msec\fR. For each increment
2634in coarseness, fio outputs half as many bins. Defaults to 0, for which
2635histogram logs contain 1216 latency bins. See \fBLOG FILE FORMATS\fR section.
2636.TP
2637.BI log_max_value \fR=\fPbool
2638If \fBlog_avg_msec\fR is set, fio logs the average over that window. If
2639you instead want to log the maximum value, set this option to 1. Defaults to
26400, meaning that averaged values are logged.
1e613c9c 2641.TP
ae588852 2642.BI log_offset \fR=\fPbool
523bad63
TK
2643If this is set, the iolog options will include the byte offset for the I/O
2644entry as well as the other data values. Defaults to 0 meaning that
2645offsets are not present in logs. Also see \fBLOG FILE FORMATS\fR section.
ae588852 2646.TP
aee2ab67 2647.BI log_compression \fR=\fPint
523bad63
TK
2648If this is set, fio will compress the I/O logs as it goes, to keep the
2649memory footprint lower. When a log reaches the specified size, that chunk is
2650removed and compressed in the background. Given that I/O logs are fairly
2651highly compressible, this yields a nice memory savings for longer runs. The
2652downside is that the compression will consume some background CPU cycles, so
2653it may impact the run. This, however, is also true if the logging ends up
2654consuming most of the system memory. So pick your poison. The I/O logs are
2655saved normally at the end of a run, by decompressing the chunks and storing
2656them in the specified log file. This feature depends on the availability of
2657zlib.
aee2ab67 2658.TP
c08f9fe2 2659.BI log_compression_cpus \fR=\fPstr
523bad63
TK
2660Define the set of CPUs that are allowed to handle online log compression for
2661the I/O jobs. This can provide better isolation between performance
0cf90a62
SW
2662sensitive jobs, and background compression work. See \fBcpus_allowed\fR for
2663the format used.
c08f9fe2 2664.TP
b26317c9 2665.BI log_store_compressed \fR=\fPbool
c08f9fe2 2666If set, fio will store the log files in a compressed format. They can be
523bad63
TK
2667decompressed with fio, using the \fB\-\-inflate\-log\fR command line
2668parameter. The files will be stored with a `.fz' suffix.
b26317c9 2669.TP
3aea75b1
KC
2670.BI log_unix_epoch \fR=\fPbool
2671If set, fio will log Unix timestamps to the log files produced by enabling
523bad63 2672write_type_log for each log type, instead of the default zero\-based
3aea75b1
KC
2673timestamps.
2674.TP
66347cfa 2675.BI block_error_percentiles \fR=\fPbool
523bad63
TK
2676If set, record errors in trim block\-sized units from writes and trims and
2677output a histogram of how many trims it took to get to errors, and what kind
2678of error was encountered.
d60e92d1 2679.TP
523bad63
TK
2680.BI bwavgtime \fR=\fPint
2681Average the calculated bandwidth over the given time. Value is specified in
2682milliseconds. If the job also does bandwidth logging through
2683\fBwrite_bw_log\fR, then the minimum of this option and
2684\fBlog_avg_msec\fR will be used. Default: 500ms.
d60e92d1 2685.TP
523bad63
TK
2686.BI iopsavgtime \fR=\fPint
2687Average the calculated IOPS over the given time. Value is specified in
2688milliseconds. If the job also does IOPS logging through
2689\fBwrite_iops_log\fR, then the minimum of this option and
2690\fBlog_avg_msec\fR will be used. Default: 500ms.
d60e92d1 2691.TP
d60e92d1 2692.BI disk_util \fR=\fPbool
523bad63
TK
2693Generate disk utilization statistics, if the platform supports it.
2694Default: true.
fa769d44 2695.TP
523bad63
TK
2696.BI disable_lat \fR=\fPbool
2697Disable measurements of total latency numbers. Useful only for cutting back
2698the number of calls to \fBgettimeofday\fR\|(2), as that does impact
2699performance at really high IOPS rates. Note that to really get rid of a
2700large amount of these calls, this option must be used with
2701\fBdisable_slat\fR and \fBdisable_bw_measurement\fR as well.
9e684a49 2702.TP
523bad63
TK
2703.BI disable_clat \fR=\fPbool
2704Disable measurements of completion latency numbers. See
2705\fBdisable_lat\fR.
9e684a49 2706.TP
523bad63
TK
2707.BI disable_slat \fR=\fPbool
2708Disable measurements of submission latency numbers. See
2709\fBdisable_lat\fR.
9e684a49 2710.TP
523bad63
TK
2711.BI disable_bw_measurement \fR=\fPbool "\fR,\fP disable_bw" \fR=\fPbool
2712Disable measurements of throughput/bandwidth numbers. See
2713\fBdisable_lat\fR.
9e684a49 2714.TP
83349190 2715.BI clat_percentiles \fR=\fPbool
b599759b
JA
2716Enable the reporting of percentiles of completion latencies. This option is
2717mutually exclusive with \fBlat_percentiles\fR.
2718.TP
2719.BI lat_percentiles \fR=\fPbool
b71968b1 2720Enable the reporting of percentiles of I/O latencies. This is similar to
b599759b
JA
2721\fBclat_percentiles\fR, except that this includes the submission latency.
2722This option is mutually exclusive with \fBclat_percentiles\fR.
83349190
YH
2723.TP
2724.BI percentile_list \fR=\fPfloat_list
66347cfa 2725Overwrite the default list of percentiles for completion latencies and the
523bad63
TK
2726block error histogram. Each number is a floating number in the range
2727(0,100], and the maximum length of the list is 20. Use ':' to separate the
2728numbers, and list the numbers in ascending order. For example,
2729`\-\-percentile_list=99.5:99.9' will cause fio to report the values of
2730completion latency below which 99.5% and 99.9% of the observed latencies
2731fell, respectively.
e883cb35
JF
2732.TP
2733.BI significant_figures \fR=\fPint
c32ba107
JA
2734If using \fB\-\-output\-format\fR of `normal', set the significant figures
2735to this value. Higher values will yield more precise IOPS and throughput
2736units, while lower values will round. Requires a minimum value of 1 and a
e883cb35 2737maximum value of 10. Defaults to 4.
523bad63 2738.SS "Error handling"
e4585935 2739.TP
523bad63
TK
2740.BI exitall_on_error
2741When one job finishes in error, terminate the rest. The default is to wait
2742for each job to finish.
e4585935 2743.TP
523bad63
TK
2744.BI continue_on_error \fR=\fPstr
2745Normally fio will exit the job on the first observed failure. If this option
2746is set, fio will continue the job when there is a 'non\-fatal error' (EIO or
2747EILSEQ) until the runtime is exceeded or the I/O size specified is
2748completed. If this option is used, there are two more stats that are
2749appended, the total error count and the first error. The error field given
2750in the stats is the first error that was hit during the run.
2751The allowed values are:
2752.RS
2753.RS
046395d7 2754.TP
523bad63
TK
2755.B none
2756Exit on any I/O or verify errors.
de890a1e 2757.TP
523bad63
TK
2758.B read
2759Continue on read errors, exit on all others.
2cafffbe 2760.TP
523bad63
TK
2761.B write
2762Continue on write errors, exit on all others.
a0679ce5 2763.TP
523bad63
TK
2764.B io
2765Continue on any I/O error, exit on all others.
de890a1e 2766.TP
523bad63
TK
2767.B verify
2768Continue on verify errors, exit on all others.
de890a1e 2769.TP
523bad63
TK
2770.B all
2771Continue on all errors.
b93b6a2e 2772.TP
523bad63
TK
2773.B 0
2774Backward\-compatible alias for 'none'.
d3a623de 2775.TP
523bad63
TK
2776.B 1
2777Backward\-compatible alias for 'all'.
2778.RE
2779.RE
1d360ffb 2780.TP
523bad63
TK
2781.BI ignore_error \fR=\fPstr
2782Sometimes you want to ignore some errors during test in that case you can
2783specify error list for each error type, instead of only being able to
2784ignore the default 'non\-fatal error' using \fBcontinue_on_error\fR.
2785`ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST' errors for
2786given error type is separated with ':'. Error may be symbol ('ENOSPC', 'ENOMEM')
2787or integer. Example:
de890a1e
SL
2788.RS
2789.RS
523bad63
TK
2790.P
2791ignore_error=EAGAIN,ENOSPC:122
2792.RE
2793.P
2794This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
2795WRITE. This option works by overriding \fBcontinue_on_error\fR with
2796the list of errors for each error type if any.
2797.RE
de890a1e 2798.TP
523bad63
TK
2799.BI error_dump \fR=\fPbool
2800If set dump every error even if it is non fatal, true by default. If
2801disabled only fatal error will be dumped.
2802.SS "Running predefined workloads"
2803Fio includes predefined profiles that mimic the I/O workloads generated by
2804other tools.
49ccb8c1 2805.TP
523bad63
TK
2806.BI profile \fR=\fPstr
2807The predefined workload to run. Current profiles are:
2808.RS
2809.RS
de890a1e 2810.TP
523bad63
TK
2811.B tiobench
2812Threaded I/O bench (tiotest/tiobench) like workload.
49ccb8c1 2813.TP
523bad63
TK
2814.B act
2815Aerospike Certification Tool (ACT) like workload.
2816.RE
de890a1e
SL
2817.RE
2818.P
523bad63
TK
2819To view a profile's additional options use \fB\-\-cmdhelp\fR after specifying
2820the profile. For example:
2821.RS
2822.TP
2823$ fio \-\-profile=act \-\-cmdhelp
de890a1e 2824.RE
523bad63 2825.SS "Act profile options"
de890a1e 2826.TP
523bad63
TK
2827.BI device\-names \fR=\fPstr
2828Devices to use.
d54fce84 2829.TP
523bad63
TK
2830.BI load \fR=\fPint
2831ACT load multiplier. Default: 1.
7aeb1e94 2832.TP
523bad63
TK
2833.BI test\-duration\fR=\fPtime
2834How long the entire test takes to run. When the unit is omitted, the value
2835is given in seconds. Default: 24h.
1008602c 2836.TP
523bad63
TK
2837.BI threads\-per\-queue\fR=\fPint
2838Number of read I/O threads per device. Default: 8.
e5f34d95 2839.TP
523bad63
TK
2840.BI read\-req\-num\-512\-blocks\fR=\fPint
2841Number of 512B blocks to read at the time. Default: 3.
d54fce84 2842.TP
523bad63
TK
2843.BI large\-block\-op\-kbytes\fR=\fPint
2844Size of large block ops in KiB (writes). Default: 131072.
d54fce84 2845.TP
523bad63
TK
2846.BI prep
2847Set to run ACT prep phase.
2848.SS "Tiobench profile options"
6d500c2e 2849.TP
523bad63
TK
2850.BI size\fR=\fPstr
2851Size in MiB.
0d978694 2852.TP
523bad63
TK
2853.BI block\fR=\fPint
2854Block size in bytes. Default: 4096.
0d978694 2855.TP
523bad63
TK
2856.BI numruns\fR=\fPint
2857Number of runs.
0d978694 2858.TP
523bad63
TK
2859.BI dir\fR=\fPstr
2860Test directory.
65fa28ca 2861.TP
523bad63
TK
2862.BI threads\fR=\fPint
2863Number of threads.
d60e92d1 2864.SH OUTPUT
40943b9a
TK
2865Fio spits out a lot of output. While running, fio will display the status of the
2866jobs created. An example of that would be:
d60e92d1 2867.P
40943b9a
TK
2868.nf
2869 Jobs: 1 (f=1): [_(1),M(1)][24.8%][r=20.5MiB/s,w=23.5MiB/s][r=82,w=94 IOPS][eta 01m:31s]
2870.fi
d1429b5c 2871.P
40943b9a
TK
2872The characters inside the first set of square brackets denote the current status of
2873each thread. The first character is the first job defined in the job file, and so
2874forth. The possible values (in typical life cycle order) are:
d60e92d1
AC
2875.RS
2876.TP
40943b9a 2877.PD 0
d60e92d1 2878.B P
40943b9a 2879Thread setup, but not started.
d60e92d1
AC
2880.TP
2881.B C
2882Thread created.
2883.TP
2884.B I
40943b9a
TK
2885Thread initialized, waiting or generating necessary data.
2886.TP
522c29f6 2887.B p
40943b9a
TK
2888Thread running pre\-reading file(s).
2889.TP
2890.B /
2891Thread is in ramp period.
d60e92d1
AC
2892.TP
2893.B R
2894Running, doing sequential reads.
2895.TP
2896.B r
2897Running, doing random reads.
2898.TP
2899.B W
2900Running, doing sequential writes.
2901.TP
2902.B w
2903Running, doing random writes.
2904.TP
2905.B M
2906Running, doing mixed sequential reads/writes.
2907.TP
2908.B m
2909Running, doing mixed random reads/writes.
2910.TP
40943b9a
TK
2911.B D
2912Running, doing sequential trims.
2913.TP
2914.B d
2915Running, doing random trims.
2916.TP
d60e92d1
AC
2917.B F
2918Running, currently waiting for \fBfsync\fR\|(2).
2919.TP
2920.B V
40943b9a
TK
2921Running, doing verification of written data.
2922.TP
2923.B f
2924Thread finishing.
d60e92d1
AC
2925.TP
2926.B E
40943b9a 2927Thread exited, not reaped by main thread yet.
d60e92d1
AC
2928.TP
2929.B \-
40943b9a
TK
2930Thread reaped.
2931.TP
2932.B X
2933Thread reaped, exited with an error.
2934.TP
2935.B K
2936Thread reaped, exited due to signal.
d1429b5c 2937.PD
40943b9a
TK
2938.RE
2939.P
2940Fio will condense the thread string as not to take up more space on the command
2941line than needed. For instance, if you have 10 readers and 10 writers running,
2942the output would look like this:
2943.P
2944.nf
2945 Jobs: 20 (f=20): [R(10),W(10)][4.0%][r=20.5MiB/s,w=23.5MiB/s][r=82,w=94 IOPS][eta 57m:36s]
2946.fi
d60e92d1 2947.P
40943b9a
TK
2948Note that the status string is displayed in order, so it's possible to tell which of
2949the jobs are currently doing what. In the example above this means that jobs 1\-\-10
2950are readers and 11\-\-20 are writers.
d60e92d1 2951.P
40943b9a
TK
2952The other values are fairly self explanatory \-\- number of threads currently
2953running and doing I/O, the number of currently open files (f=), the estimated
2954completion percentage, the rate of I/O since last check (read speed listed first,
2955then write speed and optionally trim speed) in terms of bandwidth and IOPS,
2956and time to completion for the current running group. It's impossible to estimate
2957runtime of the following groups (if any).
d60e92d1 2958.P
40943b9a
TK
2959When fio is done (or interrupted by Ctrl\-C), it will show the data for
2960each thread, group of threads, and disks in that order. For each overall thread (or
2961group) the output looks like:
2962.P
2963.nf
2964 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
2965 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
2966 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
2967 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
2968 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
2969 clat percentiles (usec):
2970 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
2971 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
2972 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
2973 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
2974 | 99.99th=[78119]
2975 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
2976 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
d3b9694d
VF
2977 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
2978 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
2979 lat (msec) : 100=0.65%
40943b9a
TK
2980 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
2981 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
2982 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
2983 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
2984 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
2985 latency : target=0, window=0, percentile=100.00%, depth=8
2986.fi
2987.P
2988The job name (or first job's name when using \fBgroup_reporting\fR) is printed,
2989along with the group id, count of jobs being aggregated, last error id seen (which
2990is 0 when there are no errors), pid/tid of that thread and the time the job/group
2991completed. Below are the I/O statistics for each data direction performed (showing
2992writes in the example above). In the order listed, they denote:
d60e92d1 2993.RS
d60e92d1 2994.TP
40943b9a
TK
2995.B read/write/trim
2996The string before the colon shows the I/O direction the statistics
2997are for. \fIIOPS\fR is the average I/Os performed per second. \fIBW\fR
2998is the average bandwidth rate shown as: value in power of 2 format
2999(value in power of 10 format). The last two values show: (total
3000I/O performed in power of 2 format / \fIruntime\fR of that thread).
d60e92d1
AC
3001.TP
3002.B slat
40943b9a
TK
3003Submission latency (\fImin\fR being the minimum, \fImax\fR being the
3004maximum, \fIavg\fR being the average, \fIstdev\fR being the standard
3005deviation). This is the time it took to submit the I/O. For
3006sync I/O this row is not displayed as the slat is really the
3007completion latency (since queue/complete is one operation there).
3008This value can be in nanoseconds, microseconds or milliseconds \-\-\-
3009fio will choose the most appropriate base and print that (in the
3010example above nanoseconds was the best scale). Note: in \fB\-\-minimal\fR mode
3011latencies are always expressed in microseconds.
d60e92d1
AC
3012.TP
3013.B clat
40943b9a
TK
3014Completion latency. Same names as slat, this denotes the time from
3015submission to completion of the I/O pieces. For sync I/O, clat will
3016usually be equal (or very close) to 0, as the time from submit to
3017complete is basically just CPU time (I/O has already been done, see slat
3018explanation).
d60e92d1 3019.TP
d3b9694d
VF
3020.B lat
3021Total latency. Same names as slat and clat, this denotes the time from
3022when fio created the I/O unit to completion of the I/O operation.
3023.TP
d60e92d1 3024.B bw
40943b9a
TK
3025Bandwidth statistics based on samples. Same names as the xlat stats,
3026but also includes the number of samples taken (\fIsamples\fR) and an
3027approximate percentage of total aggregate bandwidth this thread
3028received in its group (\fIper\fR). This last value is only really
3029useful if the threads in this group are on the same disk, since they
3030are then competing for disk access.
3031.TP
3032.B iops
3033IOPS statistics based on samples. Same names as \fBbw\fR.
d60e92d1 3034.TP
d3b9694d
VF
3035.B lat (nsec/usec/msec)
3036The distribution of I/O completion latencies. This is the time from when
3037I/O leaves fio and when it gets completed. Unlike the separate
3038read/write/trim sections above, the data here and in the remaining
3039sections apply to all I/Os for the reporting group. 250=0.04% means that
30400.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
3041of the I/Os required 250 to 499us for completion.
3042.TP
d60e92d1 3043.B cpu
40943b9a
TK
3044CPU usage. User and system time, along with the number of context
3045switches this thread went through, usage of system and user time, and
3046finally the number of major and minor page faults. The CPU utilization
3047numbers are averages for the jobs in that reporting group, while the
3048context and fault counters are summed.
d60e92d1
AC
3049.TP
3050.B IO depths
40943b9a
TK
3051The distribution of I/O depths over the job lifetime. The numbers are
3052divided into powers of 2 and each entry covers depths from that value
3053up to those that are lower than the next entry \-\- e.g., 16= covers
3054depths from 16 to 31. Note that the range covered by a depth
3055distribution entry can be different to the range covered by the
3056equivalent \fBsubmit\fR/\fBcomplete\fR distribution entry.
3057.TP
3058.B IO submit
3059How many pieces of I/O were submitting in a single submit call. Each
3060entry denotes that amount and below, until the previous entry \-\- e.g.,
306116=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3062call. Note that the range covered by a \fBsubmit\fR distribution entry can
3063be different to the range covered by the equivalent depth distribution
3064entry.
3065.TP
3066.B IO complete
3067Like the above \fBsubmit\fR number, but for completions instead.
3068.TP
3069.B IO issued rwt
3070The number of \fBread/write/trim\fR requests issued, and how many of them were
3071short or dropped.
d60e92d1 3072.TP
d3b9694d 3073.B IO latency
ee21ebee 3074These values are for \fBlatency_target\fR and related options. When
d3b9694d
VF
3075these options are engaged, this section describes the I/O depth required
3076to meet the specified latency target.
d60e92d1 3077.RE
d60e92d1 3078.P
40943b9a
TK
3079After each client has been listed, the group statistics are printed. They
3080will look like this:
3081.P
3082.nf
3083 Run status group 0 (all jobs):
3084 READ: bw=20.9MiB/s (21.9MB/s), 10.4MiB/s\-10.8MiB/s (10.9MB/s\-11.3MB/s), io=64.0MiB (67.1MB), run=2973\-3069msec
3085 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s\-621KiB/s (630kB/s\-636kB/s), io=64.0MiB (67.1MB), run=52747\-53223msec
3086.fi
3087.P
3088For each data direction it prints:
d60e92d1
AC
3089.RS
3090.TP
40943b9a
TK
3091.B bw
3092Aggregate bandwidth of threads in this group followed by the
3093minimum and maximum bandwidth of all the threads in this group.
3094Values outside of brackets are power\-of\-2 format and those
3095within are the equivalent value in a power\-of\-10 format.
d60e92d1 3096.TP
40943b9a
TK
3097.B io
3098Aggregate I/O performed of all threads in this group. The
3099format is the same as \fBbw\fR.
d60e92d1 3100.TP
40943b9a
TK
3101.B run
3102The smallest and longest runtimes of the threads in this group.
d60e92d1 3103.RE
d60e92d1 3104.P
40943b9a
TK
3105And finally, the disk statistics are printed. This is Linux specific.
3106They will look like this:
3107.P
3108.nf
3109 Disk stats (read/write):
3110 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3111.fi
3112.P
3113Each value is printed for both reads and writes, with reads first. The
3114numbers denote:
d60e92d1
AC
3115.RS
3116.TP
3117.B ios
3118Number of I/Os performed by all groups.
3119.TP
3120.B merge
007c7be9 3121Number of merges performed by the I/O scheduler.
d60e92d1
AC
3122.TP
3123.B ticks
3124Number of ticks we kept the disk busy.
3125.TP
40943b9a 3126.B in_queue
d60e92d1
AC
3127Total time spent in the disk queue.
3128.TP
3129.B util
40943b9a
TK
3130The disk utilization. A value of 100% means we kept the disk
3131busy constantly, 50% would be a disk idling half of the time.
d60e92d1 3132.RE
8423bd11 3133.P
40943b9a
TK
3134It is also possible to get fio to dump the current output while it is running,
3135without terminating the job. To do that, send fio the USR1 signal. You can
3136also get regularly timed dumps by using the \fB\-\-status\-interval\fR
3137parameter, or by creating a file in `/tmp' named
3138`fio\-dump\-status'. If fio sees this file, it will unlink it and dump the
3139current output status.
d60e92d1 3140.SH TERSE OUTPUT
40943b9a
TK
3141For scripted usage where you typically want to generate tables or graphs of the
3142results, fio can output the results in a semicolon separated format. The format
3143is one long line of values, such as:
d60e92d1 3144.P
40943b9a
TK
3145.nf
3146 2;card0;0;0;7139336;121836;60004;1;10109;27.932460;116.933948;220;126861;3495.446807;1085.368601;226;126864;3523.635629;1089.012448;24063;99944;50.275485%;59818.274627;5540.657370;7155060;122104;60004;1;8338;29.086342;117.839068;388;128077;5032.488518;1234.785715;391;128085;5061.839412;1236.909129;23436;100928;50.287926%;59964.832030;5644.844189;14.595833%;19.394167%;123706;0;7313;0.1%;0.1%;0.1%;0.1%;0.1%;0.1%;100.0%;0.00%;0.00%;0.00%;0.00%;0.00%;0.00%;0.01%;0.02%;0.05%;0.16%;6.04%;40.40%;52.68%;0.64%;0.01%;0.00%;0.01%;0.00%;0.00%;0.00%;0.00%;0.00%
3147 A description of this job goes here.
3148.fi
d60e92d1 3149.P
40943b9a 3150The job description (if provided) follows on a second line.
d60e92d1 3151.P
40943b9a
TK
3152To enable terse output, use the \fB\-\-minimal\fR or
3153`\-\-output\-format=terse' command line options. The
3154first value is the version of the terse output format. If the output has to be
3155changed for some reason, this number will be incremented by 1 to signify that
3156change.
d60e92d1 3157.P
40943b9a
TK
3158Split up, the format is as follows (comments in brackets denote when a
3159field was introduced or whether it's specific to some terse version):
d60e92d1 3160.P
40943b9a
TK
3161.nf
3162 terse version, fio version [v3], jobname, groupid, error
3163.fi
525c2bfa 3164.RS
40943b9a
TK
3165.P
3166.B
3167READ status:
525c2bfa 3168.RE
40943b9a
TK
3169.P
3170.nf
3171 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3172 Submission latency: min, max, mean, stdev (usec)
3173 Completion latency: min, max, mean, stdev (usec)
3174 Completion latency percentiles: 20 fields (see below)
3175 Total latency: min, max, mean, stdev (usec)
3176 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3177 IOPS [v5]: min, max, mean, stdev, number of samples
3178.fi
d60e92d1 3179.RS
40943b9a
TK
3180.P
3181.B
3182WRITE status:
a2c95580 3183.RE
40943b9a
TK
3184.P
3185.nf
3186 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3187 Submission latency: min, max, mean, stdev (usec)
3188 Completion latency: min, max, mean, stdev (usec)
3189 Completion latency percentiles: 20 fields (see below)
3190 Total latency: min, max, mean, stdev (usec)
3191 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3192 IOPS [v5]: min, max, mean, stdev, number of samples
3193.fi
a2c95580 3194.RS
40943b9a
TK
3195.P
3196.B
3197TRIM status [all but version 3]:
d60e92d1
AC
3198.RE
3199.P
40943b9a
TK
3200.nf
3201 Fields are similar to \fBREAD/WRITE\fR status.
3202.fi
a2c95580 3203.RS
a2c95580 3204.P
40943b9a 3205.B
d1429b5c 3206CPU usage:
d60e92d1
AC
3207.RE
3208.P
40943b9a
TK
3209.nf
3210 user, system, context switches, major faults, minor faults
3211.fi
d60e92d1 3212.RS
40943b9a
TK
3213.P
3214.B
3215I/O depths:
d60e92d1
AC
3216.RE
3217.P
40943b9a
TK
3218.nf
3219 <=1, 2, 4, 8, 16, 32, >=64
3220.fi
562c2d2f 3221.RS
40943b9a
TK
3222.P
3223.B
3224I/O latencies microseconds:
562c2d2f 3225.RE
40943b9a
TK
3226.P
3227.nf
3228 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3229.fi
562c2d2f 3230.RS
40943b9a
TK
3231.P
3232.B
3233I/O latencies milliseconds:
562c2d2f
DN
3234.RE
3235.P
40943b9a
TK
3236.nf
3237 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3238.fi
f2f788dd 3239.RS
40943b9a
TK
3240.P
3241.B
3242Disk utilization [v3]:
f2f788dd
JA
3243.RE
3244.P
40943b9a
TK
3245.nf
3246 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks, time spent in queue, disk utilization percentage
3247.fi
562c2d2f 3248.RS
d60e92d1 3249.P
40943b9a
TK
3250.B
3251Additional Info (dependent on continue_on_error, default off):
d60e92d1 3252.RE
2fc26c3d 3253.P
40943b9a
TK
3254.nf
3255 total # errors, first error code
3256.fi
2fc26c3d
IC
3257.RS
3258.P
40943b9a
TK
3259.B
3260Additional Info (dependent on description being set):
3261.RE
3262.P
2fc26c3d 3263.nf
40943b9a
TK
3264 Text description
3265.fi
3266.P
3267Completion latency percentiles can be a grouping of up to 20 sets, so for the
3268terse output fio writes all of them. Each field will look like this:
3269.P
3270.nf
3271 1.00%=6112
3272.fi
3273.P
3274which is the Xth percentile, and the `usec' latency associated with it.
3275.P
3276For \fBDisk utilization\fR, all disks used by fio are shown. So for each disk there
3277will be a disk utilization section.
3278.P
3279Below is a single line containing short names for each of the fields in the
3280minimal output v3, separated by semicolons:
3281.P
3282.nf
3283 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth;read_iops;read_runtime_ms;read_slat_min;read_slat_max;read_slat_mean;read_slat_dev;read_clat_min;read_clat_max;read_clat_mean;read_clat_dev;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min;read_lat_max;read_lat_mean;read_lat_dev;read_bw_min;read_bw_max;read_bw_agg_pct;read_bw_mean;read_bw_dev;write_kb;write_bandwidth;write_iops;write_runtime_ms;write_slat_min;write_slat_max;write_slat_mean;write_slat_dev;write_clat_min;write_clat_max;write_clat_mean;write_clat_dev;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min;write_lat_max;write_lat_mean;write_lat_dev;write_bw_min;write_bw_max;write_bw_agg_pct;write_bw_mean;write_bw_dev;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
2fc26c3d 3284.fi
44c82dba
VF
3285.SH JSON OUTPUT
3286The \fBjson\fR output format is intended to be both human readable and convenient
3287for automated parsing. For the most part its sections mirror those of the
3288\fBnormal\fR output. The \fBruntime\fR value is reported in msec and the \fBbw\fR value is
3289reported in 1024 bytes per second units.
3290.fi
d9e557ab
VF
3291.SH JSON+ OUTPUT
3292The \fBjson+\fR output format is identical to the \fBjson\fR output format except that it
3293adds a full dump of the completion latency bins. Each \fBbins\fR object contains a
3294set of (key, value) pairs where keys are latency durations and values count how
3295many I/Os had completion latencies of the corresponding duration. For example,
3296consider:
d9e557ab 3297.RS
40943b9a 3298.P
d9e557ab
VF
3299"bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
3300.RE
40943b9a 3301.P
d9e557ab
VF
3302This data indicates that one I/O required 87,552ns to complete, two I/Os required
3303100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
40943b9a 3304.P
d9e557ab 3305Also included with fio is a Python script \fBfio_jsonplus_clat2csv\fR that takes
40943b9a
TK
3306json+ output and generates CSV\-formatted latency data suitable for plotting.
3307.P
d9e557ab 3308The latency durations actually represent the midpoints of latency intervals.
40943b9a 3309For details refer to `stat.h' in the fio source.
29dbd1e5 3310.SH TRACE FILE FORMAT
40943b9a
TK
3311There are two trace file format that you can encounter. The older (v1) format is
3312unsupported since version 1.20\-rc3 (March 2008). It will still be described
29dbd1e5 3313below in case that you get an old trace and want to understand it.
29dbd1e5 3314.P
40943b9a
TK
3315In any case the trace is a simple text file with a single action per line.
3316.TP
29dbd1e5 3317.B Trace file format v1
40943b9a 3318Each line represents a single I/O action in the following format:
29dbd1e5 3319.RS
40943b9a
TK
3320.RS
3321.P
29dbd1e5 3322rw, offset, length
29dbd1e5
JA
3323.RE
3324.P
40943b9a
TK
3325where `rw=0/1' for read/write, and the `offset' and `length' entries being in bytes.
3326.P
3327This format is not supported in fio versions >= 1.20\-rc3.
3328.RE
3329.TP
29dbd1e5 3330.B Trace file format v2
40943b9a
TK
3331The second version of the trace file format was added in fio version 1.17. It
3332allows to access more then one file per trace and has a bigger set of possible
3333file actions.
29dbd1e5 3334.RS
40943b9a 3335.P
29dbd1e5 3336The first line of the trace file has to be:
40943b9a
TK
3337.RS
3338.P
3339"fio version 2 iolog"
3340.RE
3341.P
29dbd1e5 3342Following this can be lines in two different formats, which are described below.
40943b9a
TK
3343.P
3344.B
29dbd1e5 3345The file management format:
40943b9a
TK
3346.RS
3347filename action
29dbd1e5 3348.P
40943b9a 3349The `filename' is given as an absolute path. The `action' can be one of these:
29dbd1e5
JA
3350.RS
3351.TP
3352.B add
40943b9a 3353Add the given `filename' to the trace.
29dbd1e5
JA
3354.TP
3355.B open
40943b9a
TK
3356Open the file with the given `filename'. The `filename' has to have
3357been added with the \fBadd\fR action before.
29dbd1e5
JA
3358.TP
3359.B close
40943b9a
TK
3360Close the file with the given `filename'. The file has to have been
3361\fBopen\fRed before.
3362.RE
29dbd1e5 3363.RE
29dbd1e5 3364.P
40943b9a
TK
3365.B
3366The file I/O action format:
3367.RS
3368filename action offset length
29dbd1e5 3369.P
40943b9a
TK
3370The `filename' is given as an absolute path, and has to have been \fBadd\fRed and
3371\fBopen\fRed before it can be used with this format. The `offset' and `length' are
3372given in bytes. The `action' can be one of these:
29dbd1e5
JA
3373.RS
3374.TP
3375.B wait
40943b9a
TK
3376Wait for `offset' microseconds. Everything below 100 is discarded.
3377The time is relative to the previous `wait' statement.
29dbd1e5
JA
3378.TP
3379.B read
40943b9a 3380Read `length' bytes beginning from `offset'.
29dbd1e5
JA
3381.TP
3382.B write
40943b9a 3383Write `length' bytes beginning from `offset'.
29dbd1e5
JA
3384.TP
3385.B sync
40943b9a 3386\fBfsync\fR\|(2) the file.
29dbd1e5
JA
3387.TP
3388.B datasync
40943b9a 3389\fBfdatasync\fR\|(2) the file.
29dbd1e5
JA
3390.TP
3391.B trim
40943b9a
TK
3392Trim the given file from the given `offset' for `length' bytes.
3393.RE
29dbd1e5 3394.RE
29dbd1e5 3395.SH CPU IDLENESS PROFILING
40943b9a
TK
3396In some cases, we want to understand CPU overhead in a test. For example, we
3397test patches for the specific goodness of whether they reduce CPU usage.
3398Fio implements a balloon approach to create a thread per CPU that runs at idle
3399priority, meaning that it only runs when nobody else needs the cpu.
3400By measuring the amount of work completed by the thread, idleness of each CPU
3401can be derived accordingly.
3402.P
3403An unit work is defined as touching a full page of unsigned characters. Mean and
3404standard deviation of time to complete an unit work is reported in "unit work"
3405section. Options can be chosen to report detailed percpu idleness or overall
3406system idleness by aggregating percpu stats.
29dbd1e5 3407.SH VERIFICATION AND TRIGGERS
40943b9a
TK
3408Fio is usually run in one of two ways, when data verification is done. The first
3409is a normal write job of some sort with verify enabled. When the write phase has
3410completed, fio switches to reads and verifies everything it wrote. The second
3411model is running just the write phase, and then later on running the same job
3412(but with reads instead of writes) to repeat the same I/O patterns and verify
3413the contents. Both of these methods depend on the write phase being completed,
3414as fio otherwise has no idea how much data was written.
3415.P
3416With verification triggers, fio supports dumping the current write state to
3417local files. Then a subsequent read verify workload can load this state and know
3418exactly where to stop. This is useful for testing cases where power is cut to a
3419server in a managed fashion, for instance.
3420.P
29dbd1e5 3421A verification trigger consists of two things:
29dbd1e5 3422.RS
40943b9a
TK
3423.P
34241) Storing the write state of each job.
3425.P
34262) Executing a trigger command.
29dbd1e5 3427.RE
40943b9a
TK
3428.P
3429The write state is relatively small, on the order of hundreds of bytes to single
3430kilobytes. It contains information on the number of completions done, the last X
3431completions, etc.
3432.P
3433A trigger is invoked either through creation ('touch') of a specified file in
3434the system, or through a timeout setting. If fio is run with
3435`\-\-trigger\-file=/tmp/trigger\-file', then it will continually
3436check for the existence of `/tmp/trigger\-file'. When it sees this file, it
3437will fire off the trigger (thus saving state, and executing the trigger
29dbd1e5 3438command).
40943b9a
TK
3439.P
3440For client/server runs, there's both a local and remote trigger. If fio is
3441running as a server backend, it will send the job states back to the client for
3442safe storage, then execute the remote trigger, if specified. If a local trigger
3443is specified, the server will still send back the write state, but the client
3444will then execute the trigger.
29dbd1e5
JA
3445.RE
3446.P
3447.B Verification trigger example
3448.RS
40943b9a
TK
3449Let's say we want to run a powercut test on the remote Linux machine 'server'.
3450Our write workload is in `write\-test.fio'. We want to cut power to 'server' at
3451some point during the run, and we'll run this test from the safety or our local
3452machine, 'localbox'. On the server, we'll start the fio backend normally:
3453.RS
3454.P
3455server# fio \-\-server
3456.RE
3457.P
29dbd1e5 3458and on the client, we'll fire off the workload:
40943b9a
TK
3459.RS
3460.P
3461localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger\-remote="bash \-c "echo b > /proc/sysrq\-triger""
3462.RE
3463.P
3464We set `/tmp/my\-trigger' as the trigger file, and we tell fio to execute:
3465.RS
3466.P
3467echo b > /proc/sysrq\-trigger
3468.RE
3469.P
3470on the server once it has received the trigger and sent us the write state. This
3471will work, but it's not really cutting power to the server, it's merely
3472abruptly rebooting it. If we have a remote way of cutting power to the server
3473through IPMI or similar, we could do that through a local trigger command
3474instead. Let's assume we have a script that does IPMI reboot of a given hostname,
3475ipmi\-reboot. On localbox, we could then have run fio with a local trigger
3476instead:
3477.RS
3478.P
3479localbox$ fio \-\-client=server \-\-trigger\-file=/tmp/my\-trigger \-\-trigger="ipmi\-reboot server"
3480.RE
3481.P
3482For this case, fio would wait for the server to send us the write state, then
3483execute `ipmi\-reboot server' when that happened.
29dbd1e5
JA
3484.RE
3485.P
3486.B Loading verify state
3487.RS
40943b9a
TK
3488To load stored write state, a read verification job file must contain the
3489\fBverify_state_load\fR option. If that is set, fio will load the previously
29dbd1e5 3490stored state. For a local fio run this is done by loading the files directly,
40943b9a
TK
3491and on a client/server run, the server backend will ask the client to send the
3492files over and load them from there.
29dbd1e5 3493.RE
a3ae5b05 3494.SH LOG FILE FORMATS
a3ae5b05
JA
3495Fio supports a variety of log file formats, for logging latencies, bandwidth,
3496and IOPS. The logs share a common format, which looks like this:
40943b9a 3497.RS
a3ae5b05 3498.P
40943b9a
TK
3499time (msec), value, data direction, block size (bytes), offset (bytes)
3500.RE
3501.P
3502`Time' for the log entry is always in milliseconds. The `value' logged depends
3503on the type of log, it will be one of the following:
3504.RS
a3ae5b05
JA
3505.TP
3506.B Latency log
168bb587 3507Value is latency in nsecs
a3ae5b05
JA
3508.TP
3509.B Bandwidth log
6d500c2e 3510Value is in KiB/sec
a3ae5b05
JA
3511.TP
3512.B IOPS log
40943b9a
TK
3513Value is IOPS
3514.RE
a3ae5b05 3515.P
40943b9a
TK
3516`Data direction' is one of the following:
3517.RS
a3ae5b05
JA
3518.TP
3519.B 0
40943b9a 3520I/O is a READ
a3ae5b05
JA
3521.TP
3522.B 1
40943b9a 3523I/O is a WRITE
a3ae5b05
JA
3524.TP
3525.B 2
40943b9a 3526I/O is a TRIM
a3ae5b05 3527.RE
40943b9a
TK
3528.P
3529The entry's `block size' is always in bytes. The `offset' is the offset, in bytes,
3530from the start of the file, for that particular I/O. The logging of the offset can be
3531toggled with \fBlog_offset\fR.
3532.P
3533Fio defaults to logging every individual I/O. When IOPS are logged for individual
3534I/Os the `value' entry will always be 1. If windowed logging is enabled through
3535\fBlog_avg_msec\fR, fio logs the average values over the specified period of time.
3536If windowed logging is enabled and \fBlog_max_value\fR is set, then fio logs
3537maximum values in that window instead of averages. Since `data direction', `block size'
3538and `offset' are per\-I/O values, if windowed logging is enabled they
3539aren't applicable and will be 0.
49da1240 3540.SH CLIENT / SERVER
40943b9a
TK
3541Normally fio is invoked as a stand\-alone application on the machine where the
3542I/O workload should be generated. However, the backend and frontend of fio can
3543be run separately i.e., the fio server can generate an I/O workload on the "Device
3544Under Test" while being controlled by a client on another machine.
3545.P
3546Start the server on the machine which has access to the storage DUT:
3547.RS
3548.P
3549$ fio \-\-server=args
3550.RE
3551.P
3552where `args' defines what fio listens to. The arguments are of the form
3553`type,hostname' or `IP,port'. `type' is either `ip' (or ip4) for TCP/IP
3554v4, `ip6' for TCP/IP v6, or `sock' for a local unix domain socket.
3555`hostname' is either a hostname or IP address, and `port' is the port to listen
3556to (only valid for TCP/IP, not a local socket). Some examples:
3557.RS
3558.TP
e0ee7a8b 35591) \fBfio \-\-server\fR
40943b9a
TK
3560Start a fio server, listening on all interfaces on the default port (8765).
3561.TP
e0ee7a8b 35622) \fBfio \-\-server=ip:hostname,4444\fR
40943b9a
TK
3563Start a fio server, listening on IP belonging to hostname and on port 4444.
3564.TP
e0ee7a8b 35653) \fBfio \-\-server=ip6:::1,4444\fR
40943b9a
TK
3566Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3567.TP
e0ee7a8b 35684) \fBfio \-\-server=,4444\fR
40943b9a
TK
3569Start a fio server, listening on all interfaces on port 4444.
3570.TP
e0ee7a8b 35715) \fBfio \-\-server=1.2.3.4\fR
40943b9a
TK
3572Start a fio server, listening on IP 1.2.3.4 on the default port.
3573.TP
e0ee7a8b 35746) \fBfio \-\-server=sock:/tmp/fio.sock\fR
40943b9a
TK
3575Start a fio server, listening on the local socket `/tmp/fio.sock'.
3576.RE
3577.P
3578Once a server is running, a "client" can connect to the fio server with:
3579.RS
3580.P
3581$ fio <local\-args> \-\-client=<server> <remote\-args> <job file(s)>
3582.RE
3583.P
3584where `local\-args' are arguments for the client where it is running, `server'
3585is the connect string, and `remote\-args' and `job file(s)' are sent to the
3586server. The `server' string follows the same format as it does on the server
3587side, to allow IP/hostname/socket and port strings.
3588.P
3589Fio can connect to multiple servers this way:
3590.RS
3591.P
3592$ fio \-\-client=<server1> <job file(s)> \-\-client=<server2> <job file(s)>
3593.RE
3594.P
3595If the job file is located on the fio server, then you can tell the server to
3596load a local file as well. This is done by using \fB\-\-remote\-config\fR:
3597.RS
3598.P
3599$ fio \-\-client=server \-\-remote\-config /path/to/file.fio
3600.RE
3601.P
3602Then fio will open this local (to the server) job file instead of being passed
3603one from the client.
3604.P
ff6bb260 3605If you have many servers (example: 100 VMs/containers), you can input a pathname
40943b9a
TK
3606of a file containing host IPs/names as the parameter value for the
3607\fB\-\-client\fR option. For example, here is an example `host.list'
3608file containing 2 hostnames:
3609.RS
3610.P
3611.PD 0
39b5f61e 3612host1.your.dns.domain
40943b9a 3613.P
39b5f61e 3614host2.your.dns.domain
40943b9a
TK
3615.PD
3616.RE
3617.P
39b5f61e 3618The fio command would then be:
40943b9a
TK
3619.RS
3620.P
3621$ fio \-\-client=host.list <job file(s)>
3622.RE
3623.P
3624In this mode, you cannot input server\-specific parameters or job files \-\- all
39b5f61e 3625servers receive the same job file.
40943b9a
TK
3626.P
3627In order to let `fio \-\-client' runs use a shared filesystem from multiple
3628hosts, `fio \-\-client' now prepends the IP address of the server to the
3629filename. For example, if fio is using the directory `/mnt/nfs/fio' and is
3630writing filename `fileio.tmp', with a \fB\-\-client\fR `hostfile'
3631containing two hostnames `h1' and `h2' with IP addresses 192.168.10.120 and
3632192.168.10.121, then fio will create two files:
3633.RS
3634.P
3635.PD 0
39b5f61e 3636/mnt/nfs/fio/192.168.10.120.fileio.tmp
40943b9a 3637.P
39b5f61e 3638/mnt/nfs/fio/192.168.10.121.fileio.tmp
40943b9a
TK
3639.PD
3640.RE
d60e92d1
AC
3641.SH AUTHORS
3642.B fio
d292596c 3643was written by Jens Axboe <axboe@kernel.dk>.
d1429b5c
AC
3644.br
3645This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
d60e92d1 3646on documentation by Jens Axboe.
40943b9a
TK
3647.br
3648This man page was rewritten by Tomohiro Kusumi <tkusumi@tuxera.com> based
3649on documentation by Jens Axboe.
d60e92d1 3650.SH "REPORTING BUGS"
482900c9 3651Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
6468020d 3652.br
40943b9a
TK
3653See \fBREPORTING\-BUGS\fR.
3654.P
3655\fBREPORTING\-BUGS\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/REPORTING\-BUGS\fR
d60e92d1 3656.SH "SEE ALSO"
d1429b5c
AC
3657For further documentation see \fBHOWTO\fR and \fBREADME\fR.
3658.br
40943b9a 3659Sample jobfiles are available in the `examples/' directory.
9040e236 3660.br
40943b9a
TK
3661These are typically located under `/usr/share/doc/fio'.
3662.P
3663\fBHOWTO\fR: \fIhttp://git.kernel.dk/cgit/fio/plain/HOWTO\fR
9040e236 3664.br
40943b9a 3665