stat: make add lat percentile functions inline
[fio.git] / HOWTO
CommitLineData
f80dba8d
MT
1How fio works
2-------------
3
4The first step in getting fio to simulate a desired I/O workload, is writing a
5job file describing that specific setup. A job file may contain any number of
6threads and/or files -- the typical contents of the job file is a *global*
7section defining shared parameters, and one or more job sections describing the
8jobs involved. When run, fio parses this file and sets everything up as
9described. If we break down a job from top to bottom, it contains the following
10basic parameters:
11
12`I/O type`_
13
14 Defines the I/O pattern issued to the file(s). We may only be reading
15 sequentially from this file(s), or we may be writing randomly. Or even
16 mixing reads and writes, sequentially or randomly.
17 Should we be doing buffered I/O, or direct/raw I/O?
18
19`Block size`_
20
21 In how large chunks are we issuing I/O? This may be a single value,
22 or it may describe a range of block sizes.
23
24`I/O size`_
25
26 How much data are we going to be reading/writing.
27
28`I/O engine`_
29
30 How do we issue I/O? We could be memory mapping the file, we could be
31 using regular read/write, we could be using splice, async I/O, or even
32 SG (SCSI generic sg).
33
34`I/O depth`_
35
36 If the I/O engine is async, how large a queuing depth do we want to
37 maintain?
38
39
40`Target file/device`_
41
42 How many files are we spreading the workload over.
43
44`Threads, processes and job synchronization`_
45
46 How many threads or processes should we spread this workload over.
47
48The above are the basic parameters defined for a workload, in addition there's a
49multitude of parameters that modify other aspects of how this job behaves.
50
51
52Command line options
53--------------------
54
55.. option:: --debug=type
56
f50fbdda 57 Enable verbose tracing `type` of various fio actions. May be ``all`` for all types
b034c0dd
TK
58 or individual types separated by a comma (e.g. ``--debug=file,mem`` will
59 enable file and memory debugging). Currently, additional logging is
60 available for:
f80dba8d 61
b034c0dd 62 *process*
f80dba8d 63 Dump info related to processes.
b034c0dd 64 *file*
f80dba8d 65 Dump info related to file actions.
b034c0dd 66 *io*
f80dba8d 67 Dump info related to I/O queuing.
b034c0dd 68 *mem*
f80dba8d 69 Dump info related to memory allocations.
b034c0dd 70 *blktrace*
f80dba8d 71 Dump info related to blktrace setup.
b034c0dd 72 *verify*
f80dba8d 73 Dump info related to I/O verification.
b034c0dd 74 *all*
f80dba8d 75 Enable all debug options.
b034c0dd 76 *random*
f80dba8d 77 Dump info related to random offset generation.
b034c0dd 78 *parse*
f80dba8d 79 Dump info related to option matching and parsing.
b034c0dd 80 *diskutil*
f80dba8d 81 Dump info related to disk utilization updates.
b034c0dd 82 *job:x*
f80dba8d 83 Dump info only related to job number x.
b034c0dd 84 *mutex*
f80dba8d 85 Dump info only related to mutex up/down ops.
b034c0dd 86 *profile*
f80dba8d 87 Dump info related to profile extensions.
b034c0dd 88 *time*
f80dba8d 89 Dump info related to internal time keeping.
b034c0dd 90 *net*
f80dba8d 91 Dump info related to networking connections.
b034c0dd 92 *rate*
f80dba8d 93 Dump info related to I/O rate switching.
b034c0dd 94 *compress*
f80dba8d 95 Dump info related to log compress/decompress.
a02ec45a
VF
96 *steadystate*
97 Dump info related to steadystate detection.
98 *helperthread*
99 Dump info related to the helper thread.
100 *zbd*
101 Dump info related to support for zoned block devices.
b034c0dd 102 *?* or *help*
f80dba8d
MT
103 Show available debug options.
104
105.. option:: --parse-only
106
25cd4b95 107 Parse options only, don't start any I/O.
f80dba8d 108
b9921d1a
DZ
109.. option:: --merge-blktrace-only
110
111 Merge blktraces only, don't start any I/O.
112
f80dba8d
MT
113.. option:: --output=filename
114
115 Write output to file `filename`.
116
f50fbdda 117.. option:: --output-format=format
b8f7e412 118
f50fbdda 119 Set the reporting `format` to `normal`, `terse`, `json`, or `json+`. Multiple
b8f7e412
TK
120 formats can be selected, separated by a comma. `terse` is a CSV based
121 format. `json+` is like `json`, except it adds a full dump of the latency
122 buckets.
123
f80dba8d
MT
124.. option:: --bandwidth-log
125
126 Generate aggregate bandwidth logs.
127
128.. option:: --minimal
129
130 Print statistics in a terse, semicolon-delimited format.
131
132.. option:: --append-terse
133
b034c0dd
TK
134 Print statistics in selected mode AND terse, semicolon-delimited format.
135 **Deprecated**, use :option:`--output-format` instead to select multiple
136 formats.
f80dba8d 137
f50fbdda 138.. option:: --terse-version=version
f80dba8d 139
f50fbdda 140 Set terse `version` output format (default 3, or 2 or 4 or 5).
f80dba8d
MT
141
142.. option:: --version
143
b8f7e412 144 Print version information and exit.
f80dba8d
MT
145
146.. option:: --help
147
113f0e7c 148 Print a summary of the command line options and exit.
f80dba8d
MT
149
150.. option:: --cpuclock-test
151
152 Perform test and validation of internal CPU clock.
153
113f0e7c 154.. option:: --crctest=[test]
f80dba8d 155
b034c0dd
TK
156 Test the speed of the built-in checksumming functions. If no argument is
157 given, all of them are tested. Alternatively, a comma separated list can
158 be passed, in which case the given ones are tested.
f80dba8d
MT
159
160.. option:: --cmdhelp=command
161
162 Print help information for `command`. May be ``all`` for all commands.
163
164.. option:: --enghelp=[ioengine[,command]]
165
f50fbdda
TK
166 List all commands defined by `ioengine`, or print help for `command`
167 defined by `ioengine`. If no `ioengine` is given, list all
b034c0dd 168 available ioengines.
f80dba8d
MT
169
170.. option:: --showcmd=jobfile
171
b8f7e412 172 Convert `jobfile` to a set of command-line options.
f80dba8d
MT
173
174.. option:: --readonly
175
4027b2a1
VF
176 Turn on safety read-only checks, preventing writes and trims. The
177 ``--readonly`` option is an extra safety guard to prevent users from
178 accidentally starting a write or trim workload when that is not desired.
179 Fio will only modify the device under test if
180 `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite` is given. This
181 safety net can be used as an extra precaution.
f80dba8d
MT
182
183.. option:: --eta=when
184
b8f7e412 185 Specifies when real-time ETA estimate should be printed. `when` may be
db37d890
JA
186 `always`, `never` or `auto`. `auto` is the default, it prints ETA
187 when requested if the output is a TTY. `always` disregards the output
188 type, and prints ETA when requested. `never` never prints ETA.
189
190.. option:: --eta-interval=time
191
192 By default, fio requests client ETA status roughly every second. With
193 this option, the interval is configurable. Fio imposes a minimum
194 allowed time to avoid flooding the console, less than 250 msec is
195 not supported.
f80dba8d
MT
196
197.. option:: --eta-newline=time
198
947e0fe0
SW
199 Force a new line for every `time` period passed. When the unit is omitted,
200 the value is interpreted in seconds.
f80dba8d
MT
201
202.. option:: --status-interval=time
203
aa6cb459
VF
204 Force a full status dump of cumulative (from job start) values at `time`
205 intervals. This option does *not* provide per-period measurements. So
206 values such as bandwidth are running averages. When the time unit is omitted,
c1f4de8a
JA
207 `time` is interpreted in seconds. Note that using this option with
208 ``--output-format=json`` will yield output that technically isn't valid
209 json, since the output will be collated sets of valid json. It will need
210 to be split into valid sets of json after the run.
f80dba8d
MT
211
212.. option:: --section=name
213
b034c0dd
TK
214 Only run specified section `name` in job file. Multiple sections can be specified.
215 The ``--section`` option allows one to combine related jobs into one file.
216 E.g. one job file could define light, moderate, and heavy sections. Tell
217 fio to run only the "heavy" section by giving ``--section=heavy``
218 command line option. One can also specify the "write" operations in one
219 section and "verify" operation in another section. The ``--section`` option
220 only applies to job sections. The reserved *global* section is always
221 parsed and used.
f80dba8d
MT
222
223.. option:: --alloc-size=kb
224
4a419903
VF
225 Allocate additional internal smalloc pools of size `kb` in KiB. The
226 ``--alloc-size`` option increases shared memory set aside for use by fio.
b034c0dd
TK
227 If running large jobs with randommap enabled, fio can run out of memory.
228 Smalloc is an internal allocator for shared structures from a fixed size
229 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
f80dba8d 230
b034c0dd
TK
231 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
232 in :file:`/tmp`.
f80dba8d
MT
233
234.. option:: --warnings-fatal
235
b034c0dd
TK
236 All fio parser warnings are fatal, causing fio to exit with an
237 error.
f80dba8d
MT
238
239.. option:: --max-jobs=nr
240
f50fbdda 241 Set the maximum number of threads/processes to support to `nr`.
818322cc 242 NOTE: On Linux, it may be necessary to increase the shared-memory
71aa48eb 243 limit (:file:`/proc/sys/kernel/shmmax`) if fio runs into errors while
818322cc 244 creating jobs.
f80dba8d
MT
245
246.. option:: --server=args
247
b034c0dd
TK
248 Start a backend server, with `args` specifying what to listen to.
249 See `Client/Server`_ section.
f80dba8d
MT
250
251.. option:: --daemonize=pidfile
252
b034c0dd 253 Background a fio server, writing the pid to the given `pidfile` file.
f80dba8d
MT
254
255.. option:: --client=hostname
256
f50fbdda 257 Instead of running the jobs locally, send and run them on the given `hostname`
71aa48eb 258 or set of `hostname`\s. See `Client/Server`_ section.
f80dba8d
MT
259
260.. option:: --remote-config=file
261
f50fbdda 262 Tell fio server to load this local `file`.
f80dba8d
MT
263
264.. option:: --idle-prof=option
265
b8f7e412 266 Report CPU idleness. `option` is one of the following:
113f0e7c
SW
267
268 **calibrate**
269 Run unit work calibration only and exit.
270
271 **system**
272 Show aggregate system idleness and unit work.
273
274 **percpu**
275 As **system** but also show per CPU idleness.
f80dba8d
MT
276
277.. option:: --inflate-log=log
278
f50fbdda 279 Inflate and output compressed `log`.
f80dba8d
MT
280
281.. option:: --trigger-file=file
282
f50fbdda 283 Execute trigger command when `file` exists.
f80dba8d 284
f50fbdda 285.. option:: --trigger-timeout=time
f80dba8d 286
f50fbdda 287 Execute trigger at this `time`.
f80dba8d 288
f50fbdda 289.. option:: --trigger=command
f80dba8d 290
f50fbdda 291 Set this `command` as local trigger.
f80dba8d 292
f50fbdda 293.. option:: --trigger-remote=command
f80dba8d 294
f50fbdda 295 Set this `command` as remote trigger.
f80dba8d
MT
296
297.. option:: --aux-path=path
298
f4401bf8
SW
299 Use the directory specified by `path` for generated state files instead
300 of the current working directory.
f80dba8d
MT
301
302Any parameters following the options will be assumed to be job files, unless
303they match a job file parameter. Multiple job files can be listed and each job
304file will be regarded as a separate group. Fio will :option:`stonewall`
305execution between each group.
306
307
308Job file format
309---------------
310
311As previously described, fio accepts one or more job files describing what it is
312supposed to do. The job file format is the classic ini file, where the names
c60ebc45 313enclosed in [] brackets define the job name. You are free to use any ASCII name
f80dba8d
MT
314you want, except *global* which has special meaning. Following the job name is
315a sequence of zero or more parameters, one per line, that define the behavior of
316the job. If the first character in a line is a ';' or a '#', the entire line is
317discarded as a comment.
318
319A *global* section sets defaults for the jobs described in that file. A job may
320override a *global* section parameter, and a job file may even have several
321*global* sections if so desired. A job is only affected by a *global* section
322residing above it.
323
f50fbdda
TK
324The :option:`--cmdhelp` option also lists all options. If used with a `command`
325argument, :option:`--cmdhelp` will detail the given `command`.
f80dba8d
MT
326
327See the `examples/` directory for inspiration on how to write job files. Note
328the copyright and license requirements currently apply to `examples/` files.
329
330So let's look at a really simple job file that defines two processes, each
331randomly reading from a 128MiB file:
332
333.. code-block:: ini
334
335 ; -- start job file --
336 [global]
337 rw=randread
338 size=128m
339
340 [job1]
341
342 [job2]
343
344 ; -- end job file --
345
346As you can see, the job file sections themselves are empty as all the described
347parameters are shared. As no :option:`filename` option is given, fio makes up a
348`filename` for each of the jobs as it sees fit. On the command line, this job
349would look as follows::
350
351$ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
352
353
354Let's look at an example that has a number of processes writing randomly to
355files:
356
357.. code-block:: ini
358
359 ; -- start job file --
360 [random-writers]
361 ioengine=libaio
362 iodepth=4
363 rw=randwrite
364 bs=32k
365 direct=0
366 size=64m
367 numjobs=4
368 ; -- end job file --
369
370Here we have no *global* section, as we only have one job defined anyway. We
371want to use async I/O here, with a depth of 4 for each file. We also increased
372the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
373jobs. The result is 4 processes each randomly writing to their own 64MiB
374file. Instead of using the above job file, you could have given the parameters
375on the command line. For this case, you would specify::
376
377$ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
378
379When fio is utilized as a basis of any reasonably large test suite, it might be
380desirable to share a set of standardized settings across multiple job files.
381Instead of copy/pasting such settings, any section may pull in an external
382:file:`filename.fio` file with *include filename* directive, as in the following
383example::
384
385 ; -- start job file including.fio --
386 [global]
387 filename=/tmp/test
388 filesize=1m
389 include glob-include.fio
390
391 [test]
392 rw=randread
393 bs=4k
394 time_based=1
395 runtime=10
396 include test-include.fio
397 ; -- end job file including.fio --
398
399.. code-block:: ini
400
401 ; -- start job file glob-include.fio --
402 thread=1
403 group_reporting=1
404 ; -- end job file glob-include.fio --
405
406.. code-block:: ini
407
408 ; -- start job file test-include.fio --
409 ioengine=libaio
410 iodepth=4
411 ; -- end job file test-include.fio --
412
413Settings pulled into a section apply to that section only (except *global*
414section). Include directives may be nested in that any included file may contain
415further include directive(s). Include files may not contain [] sections.
416
417
418Environment variables
419~~~~~~~~~~~~~~~~~~~~~
420
421Fio also supports environment variable expansion in job files. Any sub-string of
422the form ``${VARNAME}`` as part of an option value (in other words, on the right
423of the '='), will be expanded to the value of the environment variable called
424`VARNAME`. If no such environment variable is defined, or `VARNAME` is the
425empty string, the empty string will be substituted.
426
427As an example, let's look at a sample fio invocation and job file::
428
429$ SIZE=64m NUMJOBS=4 fio jobfile.fio
430
431.. code-block:: ini
432
433 ; -- start job file --
434 [random-writers]
435 rw=randwrite
436 size=${SIZE}
437 numjobs=${NUMJOBS}
438 ; -- end job file --
439
440This will expand to the following equivalent job file at runtime:
441
442.. code-block:: ini
443
444 ; -- start job file --
445 [random-writers]
446 rw=randwrite
447 size=64m
448 numjobs=4
449 ; -- end job file --
450
451Fio ships with a few example job files, you can also look there for inspiration.
452
453Reserved keywords
454~~~~~~~~~~~~~~~~~
455
456Additionally, fio has a set of reserved keywords that will be replaced
457internally with the appropriate value. Those keywords are:
458
459**$pagesize**
460
461 The architecture page size of the running system.
462
463**$mb_memory**
464
465 Megabytes of total memory in the system.
466
467**$ncpus**
468
469 Number of online available CPUs.
470
471These can be used on the command line or in the job file, and will be
472automatically substituted with the current system values when the job is
473run. Simple math is also supported on these keywords, so you can perform actions
474like::
475
b034c0dd 476 size=8*$mb_memory
f80dba8d
MT
477
478and get that properly expanded to 8 times the size of memory in the machine.
479
480
481Job file parameters
482-------------------
483
484This section describes in details each parameter associated with a job. Some
485parameters take an option of a given type, such as an integer or a
486string. Anywhere a numeric value is required, an arithmetic expression may be
487used, provided it is surrounded by parentheses. Supported operators are:
488
489 - addition (+)
490 - subtraction (-)
491 - multiplication (*)
492 - division (/)
493 - modulus (%)
494 - exponentiation (^)
495
496For time values in expressions, units are microseconds by default. This is
497different than for time values not in expressions (not enclosed in
498parentheses). The following types are used:
499
500
501Parameter types
502~~~~~~~~~~~~~~~
503
504**str**
b034c0dd 505 String: A sequence of alphanumeric characters.
f80dba8d
MT
506
507**time**
008d0feb
SW
508 Integer with possible time suffix. Without a unit value is interpreted as
509 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
510 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
511 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
f80dba8d
MT
512
513.. _int:
514
515**int**
516 Integer. A whole number value, which may contain an integer prefix
517 and an integer suffix:
518
b034c0dd 519 [*integer prefix*] **number** [*integer suffix*]
f80dba8d
MT
520
521 The optional *integer prefix* specifies the number's base. The default
522 is decimal. *0x* specifies hexadecimal.
523
524 The optional *integer suffix* specifies the number's units, and includes an
525 optional unit prefix and an optional unit. For quantities of data, the
947e0fe0
SW
526 default unit is bytes. For quantities of time, the default unit is seconds
527 unless otherwise specified.
f80dba8d 528
9207a0cb 529 With :option:`kb_base`\=1000, fio follows international standards for unit
f80dba8d
MT
530 prefixes. To specify power-of-10 decimal values defined in the
531 International System of Units (SI):
532
eccce61a
TK
533 * *K* -- means kilo (K) or 1000
534 * *M* -- means mega (M) or 1000**2
535 * *G* -- means giga (G) or 1000**3
536 * *T* -- means tera (T) or 1000**4
537 * *P* -- means peta (P) or 1000**5
f80dba8d
MT
538
539 To specify power-of-2 binary values defined in IEC 80000-13:
540
eccce61a
TK
541 * *Ki* -- means kibi (Ki) or 1024
542 * *Mi* -- means mebi (Mi) or 1024**2
543 * *Gi* -- means gibi (Gi) or 1024**3
544 * *Ti* -- means tebi (Ti) or 1024**4
545 * *Pi* -- means pebi (Pi) or 1024**5
f80dba8d 546
193aaf6a
G
547 For Zone Block Device Mode:
548 * *z* -- means Zone
549
9207a0cb 550 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
f80dba8d
MT
551 from those specified in the SI and IEC 80000-13 standards to provide
552 compatibility with old scripts. For example, 4k means 4096.
553
554 For quantities of data, an optional unit of 'B' may be included
b8f7e412 555 (e.g., 'kB' is the same as 'k').
f80dba8d
MT
556
557 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
558 not milli). 'b' and 'B' both mean byte, not bit.
559
9207a0cb 560 Examples with :option:`kb_base`\=1000:
f80dba8d
MT
561
562 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
563 * *1 MiB*: 1048576, 1mi, 1024ki
564 * *1 MB*: 1000000, 1m, 1000k
565 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
566 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
567
9207a0cb 568 Examples with :option:`kb_base`\=1024 (default):
f80dba8d
MT
569
570 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
571 * *1 MiB*: 1048576, 1m, 1024k
572 * *1 MB*: 1000000, 1mi, 1000ki
573 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
574 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
575
576 To specify times (units are not case sensitive):
577
578 * *D* -- means days
579 * *H* -- means hours
4502cb42 580 * *M* -- means minutes
f80dba8d
MT
581 * *s* -- or sec means seconds (default)
582 * *ms* -- or *msec* means milliseconds
583 * *us* -- or *usec* means microseconds
584
585 If the option accepts an upper and lower range, use a colon ':' or
586 minus '-' to separate such values. See :ref:`irange <irange>`.
4502cb42
SW
587 If the lower value specified happens to be larger than the upper value
588 the two values are swapped.
f80dba8d
MT
589
590.. _bool:
591
592**bool**
593 Boolean. Usually parsed as an integer, however only defined for
594 true and false (1 and 0).
595
596.. _irange:
597
598**irange**
599 Integer range with suffix. Allows value range to be given, such as
c60ebc45 600 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
f80dba8d
MT
601 option allows two sets of ranges, they can be specified with a ',' or '/'
602 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
603
604**float_list**
605 A list of floating point numbers, separated by a ':' character.
606
f5c3bcf2
TK
607With the above in mind, here follows the complete list of fio job parameters.
608
f80dba8d
MT
609
610Units
611~~~~~
612
613.. option:: kb_base=int
614
615 Select the interpretation of unit prefixes in input parameters.
616
617 **1000**
618 Inputs comply with IEC 80000-13 and the International
619 System of Units (SI). Use:
620
621 - power-of-2 values with IEC prefixes (e.g., KiB)
622 - power-of-10 values with SI prefixes (e.g., kB)
623
624 **1024**
625 Compatibility mode (default). To avoid breaking old scripts:
626
627 - power-of-2 values with SI prefixes
628 - power-of-10 values with IEC prefixes
629
630 See :option:`bs` for more details on input parameters.
631
632 Outputs always use correct prefixes. Most outputs include both
633 side-by-side, like::
634
635 bw=2383.3kB/s (2327.4KiB/s)
636
637 If only one value is reported, then kb_base selects the one to use:
638
639 **1000** -- SI prefixes
640
641 **1024** -- IEC prefixes
642
643.. option:: unit_base=int
644
645 Base unit for reporting. Allowed values are:
646
647 **0**
648 Use auto-detection (default).
649 **8**
650 Byte based.
651 **1**
652 Bit based.
653
654
f80dba8d
MT
655Job description
656~~~~~~~~~~~~~~~
657
658.. option:: name=str
659
660 ASCII name of the job. This may be used to override the name printed by fio
661 for this job. Otherwise the job name is used. On the command line this
662 parameter has the special purpose of also signaling the start of a new job.
663
664.. option:: description=str
665
666 Text description of the job. Doesn't do anything except dump this text
667 description when this job is run. It's not parsed.
668
669.. option:: loops=int
670
671 Run the specified number of iterations of this job. Used to repeat the same
672 workload a given number of times. Defaults to 1.
673
674.. option:: numjobs=int
675
79591fa9
TK
676 Create the specified number of clones of this job. Each clone of job
677 is spawned as an independent thread or process. May be used to setup a
f80dba8d
MT
678 larger number of threads/processes doing the same thing. Each thread is
679 reported separately; to see statistics for all clones as a whole, use
680 :option:`group_reporting` in conjunction with :option:`new_group`.
a47b697c 681 See :option:`--max-jobs`. Default: 1.
f80dba8d
MT
682
683
684Time related parameters
685~~~~~~~~~~~~~~~~~~~~~~~
686
687.. option:: runtime=time
688
f75ede1d 689 Tell fio to terminate processing after the specified period of time. It
f80dba8d 690 can be quite hard to determine for how long a specified job will run, so
f75ede1d 691 this parameter is handy to cap the total runtime to a given time. When
804c0839 692 the unit is omitted, the value is interpreted in seconds.
f80dba8d
MT
693
694.. option:: time_based
695
696 If set, fio will run for the duration of the :option:`runtime` specified
697 even if the file(s) are completely read or written. It will simply loop over
698 the same workload as many times as the :option:`runtime` allows.
699
a881438b 700.. option:: startdelay=irange(time)
f80dba8d 701
947e0fe0
SW
702 Delay the start of job for the specified amount of time. Can be a single
703 value or a range. When given as a range, each thread will choose a value
704 randomly from within the range. Value is in seconds if a unit is omitted.
f80dba8d
MT
705
706.. option:: ramp_time=time
707
708 If set, fio will run the specified workload for this amount of time before
709 logging any performance numbers. Useful for letting performance settle
710 before logging results, thus minimizing the runtime required for stable
711 results. Note that the ``ramp_time`` is considered lead in time for a job,
712 thus it will increase the total runtime if a special timeout or
f75ede1d
SW
713 :option:`runtime` is specified. When the unit is omitted, the value is
714 given in seconds.
f80dba8d
MT
715
716.. option:: clocksource=str
717
718 Use the given clocksource as the base of timing. The supported options are:
719
720 **gettimeofday**
721 :manpage:`gettimeofday(2)`
722
723 **clock_gettime**
724 :manpage:`clock_gettime(2)`
725
726 **cpu**
727 Internal CPU clock source
728
729 cpu is the preferred clocksource if it is reliable, as it is very fast (and
730 fio is heavy on time calls). Fio will automatically use this clocksource if
731 it's supported and considered reliable on the system it is running on,
732 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
733 means supporting TSC Invariant.
734
735.. option:: gtod_reduce=bool
736
737 Enable all of the :manpage:`gettimeofday(2)` reducing options
f75ede1d 738 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
f80dba8d
MT
739 reduce precision of the timeout somewhat to really shrink the
740 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
741 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
742 time keeping was enabled.
743
744.. option:: gtod_cpu=int
745
746 Sometimes it's cheaper to dedicate a single thread of execution to just
747 getting the current time. Fio (and databases, for instance) are very
748 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
749 one CPU aside for doing nothing but logging current time to a shared memory
750 location. Then the other threads/processes that run I/O workloads need only
751 copy that segment, instead of entering the kernel with a
752 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
753 calls will be excluded from other uses. Fio will manually clear it from the
754 CPU mask of other jobs.
755
756
757Target file/device
758~~~~~~~~~~~~~~~~~~
759
760.. option:: directory=str
761
762 Prefix filenames with this directory. Used to place files in a different
763 location than :file:`./`. You can specify a number of directories by
764 separating the names with a ':' character. These directories will be
02dd2689 765 assigned equally distributed to job clones created by :option:`numjobs` as
f80dba8d
MT
766 long as they are using generated filenames. If specific `filename(s)` are
767 set fio will use the first listed directory, and thereby matching the
f4401bf8
SW
768 `filename` semantic (which generates a file for each clone if not
769 specified, but lets all clones use the same file if set).
f80dba8d 770
3b803fe1
SW
771 See the :option:`filename` option for information on how to escape "``:``"
772 characters within the directory path itself.
f80dba8d 773
f4401bf8
SW
774 Note: To control the directory fio will use for internal state files
775 use :option:`--aux-path`.
776
f80dba8d
MT
777.. option:: filename=str
778
779 Fio normally makes up a `filename` based on the job name, thread number, and
02dd2689
SW
780 file number (see :option:`filename_format`). If you want to share files
781 between threads in a job or several
79591fa9
TK
782 jobs with fixed file paths, specify a `filename` for each of them to override
783 the default. If the ioengine is file based, you can specify a number of files
784 by separating the names with a ':' colon. So if you wanted a job to open
785 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
786 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
787 specified, :option:`nrfiles` is ignored. The size of regular files specified
02dd2689 788 by this option will be :option:`size` divided by number of files unless an
79591fa9
TK
789 explicit size is specified by :option:`filesize`.
790
3b803fe1 791 Each colon in the wanted path must be escaped with a ``\``
02dd2689
SW
792 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
793 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
3b803fe1 794 :file:`F:\\filename` then you would use ``filename=F\:\filename``.
02dd2689 795
f80dba8d
MT
796 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
797 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
798 Note: Windows and FreeBSD prevent write access to areas
02dd2689
SW
799 of the disk containing in-use data (e.g. filesystems).
800
801 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
802 of the two depends on the read/write direction set.
f80dba8d
MT
803
804.. option:: filename_format=str
805
806 If sharing multiple files between jobs, it is usually necessary to have fio
807 generate the exact names that you want. By default, fio will name a file
808 based on the default file format specification of
809 :file:`jobname.jobnumber.filenumber`. With this option, that can be
810 customized. Fio will recognize and replace the following keywords in this
811 string:
812
813 **$jobname**
814 The name of the worker thread or process.
8d53c5f8
TG
815 **$clientuid**
816 IP of the fio process when using client/server mode.
f80dba8d
MT
817 **$jobnum**
818 The incremental number of the worker thread or process.
819 **$filenum**
820 The incremental number of the file for that worker thread or
821 process.
822
823 To have dependent jobs share a set of files, this option can be set to have
824 fio generate filenames that are shared between the two. For instance, if
825 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
826 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
827 will be used if no other format specifier is given.
828
645943c0
JB
829 If you specify a path then the directories will be created up to the
830 main directory for the file. So for example if you specify
831 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
832 created before the file setup part of the job. If you specify
833 :option:`directory` then the path will be relative that directory,
834 otherwise it is treated as the absolute path.
835
f80dba8d
MT
836.. option:: unique_filename=bool
837
838 To avoid collisions between networked clients, fio defaults to prefixing any
839 generated filenames (with a directory specified) with the source of the
840 client connecting. To disable this behavior, set this option to 0.
841
842.. option:: opendir=str
843
844 Recursively open any files below directory `str`.
845
846.. option:: lockfile=str
847
848 Fio defaults to not locking any files before it does I/O to them. If a file
849 or file descriptor is shared, fio can serialize I/O to that file to make the
850 end result consistent. This is usual for emulating real workloads that share
851 files. The lock modes are:
852
853 **none**
854 No locking. The default.
855 **exclusive**
856 Only one thread or process may do I/O at a time, excluding all
857 others.
858 **readwrite**
859 Read-write locking on the file. Many readers may
860 access the file at the same time, but writes get exclusive access.
861
862.. option:: nrfiles=int
863
79591fa9
TK
864 Number of files to use for this job. Defaults to 1. The size of files
865 will be :option:`size` divided by this unless explicit size is specified by
866 :option:`filesize`. Files are created for each thread separately, and each
867 file will have a file number within its name by default, as explained in
868 :option:`filename` section.
869
f80dba8d
MT
870
871.. option:: openfiles=int
872
873 Number of files to keep open at the same time. Defaults to the same as
874 :option:`nrfiles`, can be set smaller to limit the number simultaneous
875 opens.
876
877.. option:: file_service_type=str
878
879 Defines how fio decides which file from a job to service next. The following
880 types are defined:
881
882 **random**
883 Choose a file at random.
884
885 **roundrobin**
886 Round robin over opened files. This is the default.
887
888 **sequential**
889 Finish one file before moving on to the next. Multiple files can
f50fbdda 890 still be open depending on :option:`openfiles`.
f80dba8d
MT
891
892 **zipf**
c60ebc45 893 Use a *Zipf* distribution to decide what file to access.
f80dba8d
MT
894
895 **pareto**
c60ebc45 896 Use a *Pareto* distribution to decide what file to access.
f80dba8d 897
dd3503d3 898 **normal**
c60ebc45 899 Use a *Gaussian* (normal) distribution to decide what file to
f80dba8d
MT
900 access.
901
dd3503d3
SW
902 **gauss**
903 Alias for normal.
904
f80dba8d
MT
905 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
906 tell fio how many I/Os to issue before switching to a new file. For example,
907 specifying ``file_service_type=random:8`` would cause fio to issue
908 8 I/Os before selecting a new file at random. For the non-uniform
909 distributions, a floating point postfix can be given to influence how the
910 distribution is skewed. See :option:`random_distribution` for a description
911 of how that would work.
912
913.. option:: ioscheduler=str
914
915 Attempt to switch the device hosting the file to the specified I/O scheduler
916 before running.
917
918.. option:: create_serialize=bool
919
920 If true, serialize the file creation for the jobs. This may be handy to
921 avoid interleaving of data files, which may greatly depend on the filesystem
a47b697c 922 used and even the number of processors in the system. Default: true.
f80dba8d
MT
923
924.. option:: create_fsync=bool
925
22413915 926 :manpage:`fsync(2)` the data file after creation. This is the default.
f80dba8d
MT
927
928.. option:: create_on_open=bool
929
730bd7d9
SW
930 If true, don't pre-create files but allow the job's open() to create a file
931 when it's time to do I/O. Default: false -- pre-create all necessary files
932 when the job starts.
f80dba8d
MT
933
934.. option:: create_only=bool
935
936 If true, fio will only run the setup phase of the job. If files need to be
4502cb42 937 laid out or updated on disk, only that will be done -- the actual job contents
a47b697c 938 are not executed. Default: false.
f80dba8d
MT
939
940.. option:: allow_file_create=bool
941
730bd7d9
SW
942 If true, fio is permitted to create files as part of its workload. If this
943 option is false, then fio will error out if
f80dba8d
MT
944 the files it needs to use don't already exist. Default: true.
945
946.. option:: allow_mounted_write=bool
947
c60ebc45 948 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
f80dba8d
MT
949 to what appears to be a mounted device or partition. This should help catch
950 creating inadvertently destructive tests, not realizing that the test will
b1db0375
TK
951 destroy data on the mounted file system. Note that some platforms don't allow
952 writing against a mounted device regardless of this option. Default: false.
f80dba8d
MT
953
954.. option:: pre_read=bool
955
956 If this is given, files will be pre-read into memory before starting the
957 given I/O operation. This will also clear the :option:`invalidate` flag,
958 since it is pointless to pre-read and then drop the cache. This will only
959 work for I/O engines that are seek-able, since they allow you to read the
a47b697c
SW
960 same data multiple times. Thus it will not work on non-seekable I/O engines
961 (e.g. network, splice). Default: false.
f80dba8d
MT
962
963.. option:: unlink=bool
964
965 Unlink the job files when done. Not the default, as repeated runs of that
a47b697c
SW
966 job would then waste time recreating the file set again and again. Default:
967 false.
f80dba8d
MT
968
969.. option:: unlink_each_loop=bool
970
a47b697c 971 Unlink job files after each iteration or loop. Default: false.
f80dba8d 972
7b865a2f
BVA
973.. option:: zonemode=str
974
975 Accepted values are:
976
977 **none**
b8dd9750
HH
978 The :option:`zonerange`, :option:`zonesize`,
979 :option `zonecapacity` and option:`zoneskip`
980 parameters are ignored.
7b865a2f
BVA
981 **strided**
982 I/O happens in a single zone until
983 :option:`zonesize` bytes have been transferred.
984 After that number of bytes has been
985 transferred processing of the next zone
b8dd9750 986 starts. :option `zonecapacity` is ignored.
7b865a2f
BVA
987 **zbd**
988 Zoned block device mode. I/O happens
989 sequentially in each zone, even if random I/O
990 has been selected. Random I/O happens across
991 all zones instead of being restricted to a
992 single zone. The :option:`zoneskip` parameter
993 is ignored. :option:`zonerange` and
994 :option:`zonesize` must be identical.
2455851d
SK
995 Trim is handled using a zone reset operation.
996 Trim only considers non-empty sequential write
997 required and sequential write preferred zones.
7b865a2f 998
5faddc64 999.. option:: zonerange=int
f80dba8d 1000
7b865a2f
BVA
1001 Size of a single zone. See also :option:`zonesize` and
1002 :option:`zoneskip`.
f80dba8d 1003
5faddc64 1004.. option:: zonesize=int
f80dba8d 1005
7b865a2f
BVA
1006 For :option:`zonemode` =strided, this is the number of bytes to
1007 transfer before skipping :option:`zoneskip` bytes. If this parameter
1008 is smaller than :option:`zonerange` then only a fraction of each zone
1009 with :option:`zonerange` bytes will be accessed. If this parameter is
1010 larger than :option:`zonerange` then each zone will be accessed
1011 multiple times before skipping to the next zone.
1012
1013 For :option:`zonemode` =zbd, this is the size of a single zone. The
1014 :option:`zonerange` parameter is ignored in this mode.
f80dba8d 1015
b8dd9750
HH
1016
1017.. option:: zonecapacity=int
1018
1019 For :option:`zonemode` =zbd, this defines the capacity of a single zone,
1020 which is the accessible area starting from the zone start address.
1021 This parameter only applies when using :option:`zonemode` =zbd in
1022 combination with regular block devices. If not specified it defaults to
1023 the zone size. If the target device is a zoned block device, the zone
1024 capacity is obtained from the device information and this option is
1025 ignored.
1026
f80dba8d
MT
1027.. option:: zoneskip=int
1028
7b865a2f
BVA
1029 For :option:`zonemode` =strided, the number of bytes to skip after
1030 :option:`zonesize` bytes of data have been transferred. This parameter
1031 must be zero for :option:`zonemode` =zbd.
f80dba8d 1032
bfbdd35b
BVA
1033.. option:: read_beyond_wp=bool
1034
1035 This parameter applies to :option:`zonemode` =zbd only.
1036
1037 Zoned block devices are block devices that consist of multiple zones.
1038 Each zone has a type, e.g. conventional or sequential. A conventional
1039 zone can be written at any offset that is a multiple of the block
1040 size. Sequential zones must be written sequentially. The position at
1041 which a write must occur is called the write pointer. A zoned block
1042 device can be either drive managed, host managed or host aware. For
1043 host managed devices the host must ensure that writes happen
1044 sequentially. Fio recognizes host managed devices and serializes
1045 writes to sequential zones for these devices.
1046
1047 If a read occurs in a sequential zone beyond the write pointer then
1048 the zoned block device will complete the read without reading any data
1049 from the storage medium. Since such reads lead to unrealistically high
1050 bandwidth and IOPS numbers fio only reads beyond the write pointer if
1051 explicitly told to do so. Default: false.
1052
59b07544
BVA
1053.. option:: max_open_zones=int
1054
1055 When running a random write test across an entire drive many more
1056 zones will be open than in a typical application workload. Hence this
1057 command line option that allows to limit the number of open zones. The
1058 number of open zones is defined as the number of zones to which write
1059 commands are issued.
1060
3b78a972
AK
1061.. option:: job_max_open_zones=int
1062
1063 Limit on the number of simultaneously opened zones per single
1064 thread/process.
1065
a7c2b6fc
BVA
1066.. option:: zone_reset_threshold=float
1067
1068 A number between zero and one that indicates the ratio of logical
1069 blocks with data to the total number of logical blocks in the test
1070 above which zones should be reset periodically.
1071
1072.. option:: zone_reset_frequency=float
1073
1074 A number between zero and one that indicates how often a zone reset
1075 should be issued if the zone reset threshold has been exceeded. A zone
1076 reset is submitted after each (1 / zone_reset_frequency) write
1077 requests. This and the previous parameter can be used to simulate
1078 garbage collection activity.
1079
f80dba8d
MT
1080
1081I/O type
1082~~~~~~~~
1083
1084.. option:: direct=bool
1085
1086 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
8e889110 1087 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
f80dba8d
MT
1088 ioengines don't support direct I/O. Default: false.
1089
1090.. option:: atomic=bool
1091
1092 If value is true, attempt to use atomic direct I/O. Atomic writes are
1093 guaranteed to be stable once acknowledged by the operating system. Only
1094 Linux supports O_ATOMIC right now.
1095
1096.. option:: buffered=bool
1097
1098 If value is true, use buffered I/O. This is the opposite of the
1099 :option:`direct` option. Defaults to true.
1100
1101.. option:: readwrite=str, rw=str
1102
1103 Type of I/O pattern. Accepted values are:
1104
1105 **read**
1106 Sequential reads.
1107 **write**
1108 Sequential writes.
1109 **trim**
3740cfc8
VF
1110 Sequential trims (Linux block devices and SCSI
1111 character devices only).
f80dba8d
MT
1112 **randread**
1113 Random reads.
2831be97
SW
1114 **randwrite**
1115 Random writes.
f80dba8d 1116 **randtrim**
3740cfc8
VF
1117 Random trims (Linux block devices and SCSI
1118 character devices only).
f80dba8d
MT
1119 **rw,readwrite**
1120 Sequential mixed reads and writes.
1121 **randrw**
1122 Random mixed reads and writes.
1123 **trimwrite**
1124 Sequential trim+write sequences. Blocks will be trimmed first,
1125 then the same blocks will be written to.
1126
1127 Fio defaults to read if the option is not specified. For the mixed I/O
1128 types, the default is to split them 50/50. For certain types of I/O the
730bd7d9
SW
1129 result may still be skewed a bit, since the speed may be different.
1130
1131 It is possible to specify the number of I/Os to do before getting a new
1132 offset by appending ``:<nr>`` to the end of the string given. For a
f80dba8d
MT
1133 random read, it would look like ``rw=randread:8`` for passing in an offset
1134 modifier with a value of 8. If the suffix is used with a sequential I/O
730bd7d9
SW
1135 pattern, then the *<nr>* value specified will be **added** to the generated
1136 offset for each I/O turning sequential I/O into sequential I/O with holes.
1137 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1138 the :option:`rw_sequencer` option.
f80dba8d
MT
1139
1140.. option:: rw_sequencer=str
1141
1142 If an offset modifier is given by appending a number to the ``rw=<str>``
1143 line, then this option controls how that number modifies the I/O offset
1144 being generated. Accepted values are:
1145
1146 **sequential**
1147 Generate sequential offset.
1148 **identical**
1149 Generate the same offset.
1150
1151 ``sequential`` is only useful for random I/O, where fio would normally
c60ebc45 1152 generate a new random offset for every I/O. If you append e.g. 8 to randread,
007c7be9
SW
1153 you would get a new random offset for every 8 I/Os. The result would be a
1154 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
f80dba8d
MT
1155 to specify that. As sequential I/O is already sequential, setting
1156 ``sequential`` for that would not result in any differences. ``identical``
1157 behaves in a similar fashion, except it sends the same offset 8 number of
1158 times before generating a new offset.
1159
5cb8a8cd 1160.. option:: unified_rw_reporting=str
f80dba8d
MT
1161
1162 Fio normally reports statistics on a per data direction basis, meaning that
5cb8a8cd
BP
1163 reads, writes, and trims are accounted and reported separately. This option
1164 determines whether fio reports the results normally, summed together, or as
1165 both options.
1166 Accepted values are:
1167
1168 **none**
1169 Normal statistics reporting.
1170
1171 **mixed**
1172 Statistics are summed per data direction and reported together.
1173
1174 **both**
1175 Statistics are reported normally, followed by the mixed statistics.
1176
1177 **0**
1178 Backward-compatible alias for **none**.
1179
1180 **1**
1181 Backward-compatible alias for **mixed**.
9326926b 1182
5cb8a8cd
BP
1183 **2**
1184 Alias for **both**.
f80dba8d
MT
1185
1186.. option:: randrepeat=bool
1187
1188 Seed the random number generator used for random I/O patterns in a
1189 predictable way so the pattern is repeatable across runs. Default: true.
1190
1191.. option:: allrandrepeat=bool
1192
1193 Seed all random number generators in a predictable way so results are
1194 repeatable across runs. Default: false.
1195
1196.. option:: randseed=int
1197
1198 Seed the random number generators based on this seed value, to be able to
1199 control what sequence of output is being generated. If not set, the random
1200 sequence depends on the :option:`randrepeat` setting.
1201
1202.. option:: fallocate=str
1203
1204 Whether pre-allocation is performed when laying down files.
1205 Accepted values are:
1206
1207 **none**
1208 Do not pre-allocate space.
1209
2c3e17be
SW
1210 **native**
1211 Use a platform's native pre-allocation call but fall back to
1212 **none** behavior if it fails/is not implemented.
1213
f80dba8d
MT
1214 **posix**
1215 Pre-allocate via :manpage:`posix_fallocate(3)`.
1216
1217 **keep**
1218 Pre-allocate via :manpage:`fallocate(2)` with
1219 FALLOC_FL_KEEP_SIZE set.
1220
38ca5f03
TV
1221 **truncate**
1222 Extend file to final size via :manpage:`ftruncate(2)`
1223 instead of allocating.
1224
f80dba8d
MT
1225 **0**
1226 Backward-compatible alias for **none**.
1227
1228 **1**
1229 Backward-compatible alias for **posix**.
1230
1231 May not be available on all supported platforms. **keep** is only available
2c3e17be
SW
1232 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1233 because ZFS doesn't support pre-allocation. Default: **native** if any
38ca5f03
TV
1234 pre-allocation methods except **truncate** are available, **none** if not.
1235
1236 Note that using **truncate** on Windows will interact surprisingly
1237 with non-sequential write patterns. When writing to a file that has
1238 been extended by setting the end-of-file information, Windows will
1239 backfill the unwritten portion of the file up to that offset with
1240 zeroes before issuing the new write. This means that a single small
1241 write to the end of an extended file will stall until the entire
1242 file has been filled with zeroes.
f80dba8d
MT
1243
1244.. option:: fadvise_hint=str
1245
c712c97a
JA
1246 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1247 advise the kernel on what I/O patterns are likely to be issued.
1248 Accepted values are:
f80dba8d
MT
1249
1250 **0**
1251 Backwards-compatible hint for "no hint".
1252
1253 **1**
1254 Backwards compatible hint for "advise with fio workload type". This
1255 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1256 for a sequential workload.
1257
1258 **sequential**
1259 Advise using **FADV_SEQUENTIAL**.
1260
1261 **random**
1262 Advise using **FADV_RANDOM**.
1263
8f4b9f24 1264.. option:: write_hint=str
f80dba8d 1265
8f4b9f24
JA
1266 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1267 from a write. Only supported on Linux, as of version 4.13. Accepted
1268 values are:
1269
1270 **none**
1271 No particular life time associated with this file.
1272
1273 **short**
1274 Data written to this file has a short life time.
1275
1276 **medium**
1277 Data written to this file has a medium life time.
1278
1279 **long**
1280 Data written to this file has a long life time.
1281
1282 **extreme**
1283 Data written to this file has a very long life time.
1284
1285 The values are all relative to each other, and no absolute meaning
1286 should be associated with them.
f80dba8d
MT
1287
1288.. option:: offset=int
1289
82dbb8cb 1290 Start I/O at the provided offset in the file, given as either a fixed size in
193aaf6a 1291 bytes, zones or a percentage. If a percentage is given, the generated offset will be
83c8b093
JF
1292 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1293 provided. Data before the given offset will not be touched. This
89978a6b
BW
1294 effectively caps the file size at `real_size - offset`. Can be combined with
1295 :option:`size` to constrain the start and end range of the I/O workload.
44bb1142 1296 A percentage can be specified by a number between 1 and 100 followed by '%',
193aaf6a
G
1297 for example, ``offset=20%`` to specify 20%. In ZBD mode, value can be set as
1298 number of zones using 'z'.
f80dba8d 1299
83c8b093
JF
1300.. option:: offset_align=int
1301
1302 If set to non-zero value, the byte offset generated by a percentage ``offset``
1303 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1304 offset is aligned to the minimum block size.
1305
f80dba8d
MT
1306.. option:: offset_increment=int
1307
1308 If this is provided, then the real offset becomes `offset + offset_increment
1309 * thread_number`, where the thread number is a counter that starts at 0 and
1310 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1311 specified). This option is useful if there are several jobs which are
1312 intended to operate on a file in parallel disjoint segments, with even
0b288ba1
VF
1313 spacing between the starting points. Percentages can be used for this option.
1314 If a percentage is given, the generated offset will be aligned to the minimum
193aaf6a
G
1315 ``blocksize`` or to the value of ``offset_align`` if provided. In ZBD mode, value can
1316 also be set as number of zones using 'z'.
f80dba8d
MT
1317
1318.. option:: number_ios=int
1319
c60ebc45 1320 Fio will normally perform I/Os until it has exhausted the size of the region
f80dba8d
MT
1321 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1322 condition). With this setting, the range/size can be set independently of
c60ebc45 1323 the number of I/Os to perform. When fio reaches this number, it will exit
f80dba8d
MT
1324 normally and report status. Note that this does not extend the amount of I/O
1325 that will be done, it will only stop fio if this condition is met before
1326 other end-of-job criteria.
1327
1328.. option:: fsync=int
1329
730bd7d9
SW
1330 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1331 the dirty data for every number of blocks given. For example, if you give 32
1332 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1333 using non-buffered I/O, we may not sync the file. The exception is the sg
1334 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1335 means fio does not periodically issue and wait for a sync to complete. Also
1336 see :option:`end_fsync` and :option:`fsync_on_close`.
f80dba8d
MT
1337
1338.. option:: fdatasync=int
1339
1340 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
44f668d7 1341 not metadata blocks. In Windows, FreeBSD, DragonFlyBSD or OSX there is no
730bd7d9
SW
1342 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1343 Defaults to 0, which means fio does not periodically issue and wait for a
1344 data-only sync to complete.
f80dba8d
MT
1345
1346.. option:: write_barrier=int
1347
2831be97 1348 Make every `N-th` write a barrier write.
f80dba8d 1349
f50fbdda 1350.. option:: sync_file_range=str:int
f80dba8d 1351
f50fbdda 1352 Use :manpage:`sync_file_range(2)` for every `int` number of write
f80dba8d
MT
1353 operations. Fio will track range of writes that have happened since the last
1354 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1355
1356 **wait_before**
1357 SYNC_FILE_RANGE_WAIT_BEFORE
1358 **write**
1359 SYNC_FILE_RANGE_WRITE
1360 **wait_after**
1361 SYNC_FILE_RANGE_WAIT_AFTER
1362
1363 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1364 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1365 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1366 Linux specific.
1367
1368.. option:: overwrite=bool
1369
1370 If true, writes to a file will always overwrite existing data. If the file
1371 doesn't already exist, it will be created before the write phase begins. If
1372 the file exists and is large enough for the specified write phase, nothing
a47b697c 1373 will be done. Default: false.
f80dba8d
MT
1374
1375.. option:: end_fsync=bool
1376
a47b697c
SW
1377 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1378 Default: false.
f80dba8d
MT
1379
1380.. option:: fsync_on_close=bool
1381
1382 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
a47b697c
SW
1383 from :option:`end_fsync` in that it will happen on every file close, not
1384 just at the end of the job. Default: false.
f80dba8d
MT
1385
1386.. option:: rwmixread=int
1387
1388 Percentage of a mixed workload that should be reads. Default: 50.
1389
1390.. option:: rwmixwrite=int
1391
1392 Percentage of a mixed workload that should be writes. If both
1393 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1394 add up to 100%, the latter of the two will be used to override the
1395 first. This may interfere with a given rate setting, if fio is asked to
1396 limit reads or writes to a certain rate. If that is the case, then the
1397 distribution may be skewed. Default: 50.
1398
a87c90fd 1399.. option:: random_distribution=str:float[:float][,str:float][,str:float]
f80dba8d
MT
1400
1401 By default, fio will use a completely uniform random distribution when asked
1402 to perform random I/O. Sometimes it is useful to skew the distribution in
1403 specific ways, ensuring that some parts of the data is more hot than others.
1404 fio includes the following distribution models:
1405
1406 **random**
1407 Uniform random distribution
1408
1409 **zipf**
1410 Zipf distribution
1411
1412 **pareto**
1413 Pareto distribution
1414
b2f4b559 1415 **normal**
c60ebc45 1416 Normal (Gaussian) distribution
f80dba8d
MT
1417
1418 **zoned**
1419 Zoned random distribution
1420
59466396
JA
1421 **zoned_abs**
1422 Zone absolute random distribution
1423
f80dba8d 1424 When using a **zipf** or **pareto** distribution, an input value is also
f50fbdda 1425 needed to define the access pattern. For **zipf**, this is the `Zipf
c60ebc45 1426 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
f50fbdda 1427 program, :command:`fio-genzipf`, that can be used visualize what the given input
f80dba8d
MT
1428 values will yield in terms of hit rates. If you wanted to use **zipf** with
1429 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1430 option. If a non-uniform model is used, fio will disable use of the random
b2f4b559
SW
1431 map. For the **normal** distribution, a normal (Gaussian) deviation is
1432 supplied as a value between 0 and 100.
f80dba8d 1433
a87c90fd
AK
1434 The second, optional float is allowed for **pareto**, **zipf** and **normal** distributions.
1435 It allows to set base of distribution in non-default place, giving more control
1436 over most probable outcome. This value is in range [0-1] which maps linearly to
1437 range of possible random values.
1438 Defaults are: random for **pareto** and **zipf**, and 0.5 for **normal**.
1439 If you wanted to use **zipf** with a `theta` of 1.2 centered on 1/4 of allowed value range,
1440 you would use ``random_distibution=zipf:1.2:0.25``.
1441
f80dba8d
MT
1442 For a **zoned** distribution, fio supports specifying percentages of I/O
1443 access that should fall within what range of the file or device. For
1444 example, given a criteria of:
1445
f50fbdda
TK
1446 * 60% of accesses should be to the first 10%
1447 * 30% of accesses should be to the next 20%
1448 * 8% of accesses should be to the next 30%
1449 * 2% of accesses should be to the next 40%
f80dba8d
MT
1450
1451 we can define that through zoning of the random accesses. For the above
1452 example, the user would do::
1453
1454 random_distribution=zoned:60/10:30/20:8/30:2/40
1455
59466396
JA
1456 A **zoned_abs** distribution works exactly like the **zoned**, except
1457 that it takes absolute sizes. For example, let's say you wanted to
1458 define access according to the following criteria:
1459
1460 * 60% of accesses should be to the first 20G
1461 * 30% of accesses should be to the next 100G
1462 * 10% of accesses should be to the next 500G
1463
1464 we can define an absolute zoning distribution with:
1465
1466 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1467
6a16ece8
JA
1468 For both **zoned** and **zoned_abs**, fio supports defining up to
1469 256 separate zones.
1470
59466396
JA
1471 Similarly to how :option:`bssplit` works for setting ranges and
1472 percentages of block sizes. Like :option:`bssplit`, it's possible to
1473 specify separate zones for reads, writes, and trims. If just one set
1474 is given, it'll apply to all of them. This goes for both **zoned**
1475 **zoned_abs** distributions.
f80dba8d
MT
1476
1477.. option:: percentage_random=int[,int][,int]
1478
1479 For a random workload, set how big a percentage should be random. This
1480 defaults to 100%, in which case the workload is fully random. It can be set
1481 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1482 sequential. Any setting in between will result in a random mix of sequential
1483 and random I/O, at the given percentages. Comma-separated values may be
1484 specified for reads, writes, and trims as described in :option:`blocksize`.
1485
1486.. option:: norandommap
1487
1488 Normally fio will cover every block of the file when doing random I/O. If
1489 this option is given, fio will just get a new random offset without looking
1490 at past I/O history. This means that some blocks may not be read or written,
1491 and that some blocks may be read/written more than once. If this option is
1492 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1493 only intact blocks are verified, i.e., partially-overwritten blocks are
47e6a6e5
SW
1494 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1495 the same block to be overwritten, which can cause verification errors. Either
1496 do not use norandommap in this case, or also use the lfsr random generator.
f80dba8d
MT
1497
1498.. option:: softrandommap=bool
1499
1500 See :option:`norandommap`. If fio runs with the random block map enabled and
1501 it fails to allocate the map, if this option is set it will continue without
1502 a random block map. As coverage will not be as complete as with random maps,
1503 this option is disabled by default.
1504
1505.. option:: random_generator=str
1506
f50fbdda 1507 Fio supports the following engines for generating I/O offsets for random I/O:
f80dba8d
MT
1508
1509 **tausworthe**
f50fbdda 1510 Strong 2^88 cycle random number generator.
f80dba8d 1511 **lfsr**
f50fbdda 1512 Linear feedback shift register generator.
f80dba8d 1513 **tausworthe64**
f50fbdda 1514 Strong 64-bit 2^258 cycle random number generator.
f80dba8d
MT
1515
1516 **tausworthe** is a strong random number generator, but it requires tracking
1517 on the side if we want to ensure that blocks are only read or written
f50fbdda 1518 once. **lfsr** guarantees that we never generate the same offset twice, and
f80dba8d 1519 it's also less computationally expensive. It's not a true random generator,
f50fbdda 1520 however, though for I/O purposes it's typically good enough. **lfsr** only
f80dba8d
MT
1521 works with single block sizes, not with workloads that use multiple block
1522 sizes. If used with such a workload, fio may read or write some blocks
1523 multiple times. The default value is **tausworthe**, unless the required
1524 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1525 selected automatically.
1526
1527
1528Block size
1529~~~~~~~~~~
1530
1531.. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1532
1533 The block size in bytes used for I/O units. Default: 4096. A single value
1534 applies to reads, writes, and trims. Comma-separated values may be
1535 specified for reads, writes, and trims. A value not terminated in a comma
1536 applies to subsequent types.
1537
1538 Examples:
1539
1540 **bs=256k**
1541 means 256k for reads, writes and trims.
1542
1543 **bs=8k,32k**
1544 means 8k for reads, 32k for writes and trims.
1545
1546 **bs=8k,32k,**
1547 means 8k for reads, 32k for writes, and default for trims.
1548
1549 **bs=,8k**
1550 means default for reads, 8k for writes and trims.
1551
1552 **bs=,8k,**
b443ae44 1553 means default for reads, 8k for writes, and default for trims.
f80dba8d
MT
1554
1555.. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1556
1557 A range of block sizes in bytes for I/O units. The issued I/O unit will
1558 always be a multiple of the minimum size, unless
1559 :option:`blocksize_unaligned` is set.
1560
1561 Comma-separated ranges may be specified for reads, writes, and trims as
1562 described in :option:`blocksize`.
1563
1564 Example: ``bsrange=1k-4k,2k-8k``.
1565
1566.. option:: bssplit=str[,str][,str]
1567
6a16ece8
JA
1568 Sometimes you want even finer grained control of the block sizes
1569 issued, not just an even split between them. This option allows you to
1570 weight various block sizes, so that you are able to define a specific
1571 amount of block sizes issued. The format for this option is::
f80dba8d
MT
1572
1573 bssplit=blocksize/percentage:blocksize/percentage
1574
6a16ece8
JA
1575 for as many block sizes as needed. So if you want to define a workload
1576 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1577 write::
f80dba8d
MT
1578
1579 bssplit=4k/10:64k/50:32k/40
1580
6a16ece8
JA
1581 Ordering does not matter. If the percentage is left blank, fio will
1582 fill in the remaining values evenly. So a bssplit option like this one::
f80dba8d
MT
1583
1584 bssplit=4k/50:1k/:32k/
1585
6a16ece8
JA
1586 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1587 add up to 100, if bssplit is given a range that adds up to more, it
1588 will error out.
f80dba8d
MT
1589
1590 Comma-separated values may be specified for reads, writes, and trims as
1591 described in :option:`blocksize`.
1592
6a16ece8
JA
1593 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1594 having 90% 4k writes and 10% 8k writes, you would specify::
f80dba8d 1595
cf04b906 1596 bssplit=2k/50:4k/50,4k/90:8k/10
f80dba8d 1597
6a16ece8
JA
1598 Fio supports defining up to 64 different weights for each data
1599 direction.
1600
f80dba8d
MT
1601.. option:: blocksize_unaligned, bs_unaligned
1602
1603 If set, fio will issue I/O units with any size within
1604 :option:`blocksize_range`, not just multiples of the minimum size. This
1605 typically won't work with direct I/O, as that normally requires sector
1606 alignment.
1607
589e88b7 1608.. option:: bs_is_seq_rand=bool
f80dba8d
MT
1609
1610 If this option is set, fio will use the normal read,write blocksize settings
1611 as sequential,random blocksize settings instead. Any random read or write
1612 will use the WRITE blocksize settings, and any sequential read or write will
1613 use the READ blocksize settings.
1614
1615.. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1616
1617 Boundary to which fio will align random I/O units. Default:
1618 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1619 I/O, though it usually depends on the hardware block size. This option is
1620 mutually exclusive with using a random map for files, so it will turn off
1621 that option. Comma-separated values may be specified for reads, writes, and
1622 trims as described in :option:`blocksize`.
1623
1624
1625Buffers and memory
1626~~~~~~~~~~~~~~~~~~
1627
1628.. option:: zero_buffers
1629
1630 Initialize buffers with all zeros. Default: fill buffers with random data.
1631
1632.. option:: refill_buffers
1633
1634 If this option is given, fio will refill the I/O buffers on every
72592780
SW
1635 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1636 naturally. Defaults to being unset i.e., the buffer is only filled at
1637 init time and the data in it is reused when possible but if any of
1638 :option:`verify`, :option:`buffer_compress_percentage` or
1639 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1640 automatically enabled.
f80dba8d
MT
1641
1642.. option:: scramble_buffers=bool
1643
1644 If :option:`refill_buffers` is too costly and the target is using data
1645 deduplication, then setting this option will slightly modify the I/O buffer
1646 contents to defeat normal de-dupe attempts. This is not enough to defeat
1647 more clever block compression attempts, but it will stop naive dedupe of
1648 blocks. Default: true.
1649
1650.. option:: buffer_compress_percentage=int
1651
72592780
SW
1652 If this is set, then fio will attempt to provide I/O buffer content
1653 (on WRITEs) that compresses to the specified level. Fio does this by
1654 providing a mix of random data followed by fixed pattern data. The
1655 fixed pattern is either zeros, or the pattern specified by
1656 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1657 might skew the compression ratio slightly. Setting
1658 `buffer_compress_percentage` to a value other than 100 will also
1659 enable :option:`refill_buffers` in order to reduce the likelihood that
1660 adjacent blocks are so similar that they over compress when seen
1661 together. See :option:`buffer_compress_chunk` for how to set a finer or
1662 coarser granularity for the random/fixed data region. Defaults to unset
1663 i.e., buffer data will not adhere to any compression level.
f80dba8d
MT
1664
1665.. option:: buffer_compress_chunk=int
1666
72592780
SW
1667 This setting allows fio to manage how big the random/fixed data region
1668 is when using :option:`buffer_compress_percentage`. When
1669 `buffer_compress_chunk` is set to some non-zero value smaller than the
1670 block size, fio can repeat the random/fixed region throughout the I/O
1671 buffer at the specified interval (which particularly useful when
1672 bigger block sizes are used for a job). When set to 0, fio will use a
1673 chunk size that matches the block size resulting in a single
1674 random/fixed region within the I/O buffer. Defaults to 512. When the
1675 unit is omitted, the value is interpreted in bytes.
f80dba8d
MT
1676
1677.. option:: buffer_pattern=str
1678
a1554f65
SB
1679 If set, fio will fill the I/O buffers with this pattern or with the contents
1680 of a file. If not set, the contents of I/O buffers are defined by the other
1681 options related to buffer contents. The setting can be any pattern of bytes,
1682 and can be prefixed with 0x for hex values. It may also be a string, where
1683 the string must then be wrapped with ``""``. Or it may also be a filename,
1684 where the filename must be wrapped with ``''`` in which case the file is
1685 opened and read. Note that not all the file contents will be read if that
1686 would cause the buffers to overflow. So, for example::
1687
1688 buffer_pattern='filename'
1689
1690 or::
f80dba8d
MT
1691
1692 buffer_pattern="abcd"
1693
1694 or::
1695
1696 buffer_pattern=-12
1697
1698 or::
1699
1700 buffer_pattern=0xdeadface
1701
1702 Also you can combine everything together in any order::
1703
a1554f65 1704 buffer_pattern=0xdeadface"abcd"-12'filename'
f80dba8d
MT
1705
1706.. option:: dedupe_percentage=int
1707
1708 If set, fio will generate this percentage of identical buffers when
1709 writing. These buffers will be naturally dedupable. The contents of the
1710 buffers depend on what other buffer compression settings have been set. It's
1711 possible to have the individual buffers either fully compressible, or not at
72592780
SW
1712 all -- this option only controls the distribution of unique buffers. Setting
1713 this option will also enable :option:`refill_buffers` to prevent every buffer
1714 being identical.
f80dba8d 1715
0d71aa98
BD
1716.. option:: dedupe_mode=str
1717
1718 If ``dedupe_percentage=<int>`` is given, then this option controls how fio
1719 generates the dedupe buffers.
1720
1721 **repeat**
1722 Generate dedupe buffers by repeating previous writes
1723 **working_set**
1724 Generate dedupe buffers from working set
1725
1726 ``repeat`` is the default option for fio. Dedupe buffers are generated
1727 by repeating previous unique write.
1728
1729 ``working_set`` is a more realistic workload.
1730 With ``working_set``, ``dedupe_working_set_percentage=<int>`` should be provided.
1731 Given that, fio will use the initial unique write buffers as its working set.
1732 Upon deciding to dedupe, fio will randomly choose a buffer from the working set.
1733 Note that by using ``working_set`` the dedupe percentage will converge
1734 to the desired over time while ``repeat`` maintains the desired percentage
1735 throughout the job.
1736
1737.. option:: dedupe_working_set_percentage=int
1738
1739 If ``dedupe_mode=<str>`` is set to ``working_set``, then this controls
1740 the percentage of size of the file or device used as the buffers
1741 fio will choose to generate the dedupe buffers from
1742
1743 Note that size needs to be explicitly provided and only 1 file per
1744 job is supported
1745
f80dba8d
MT
1746.. option:: invalidate=bool
1747
730bd7d9
SW
1748 Invalidate the buffer/page cache parts of the files to be used prior to
1749 starting I/O if the platform and file type support it. Defaults to true.
21c1b29e
TK
1750 This will be ignored if :option:`pre_read` is also specified for the
1751 same job.
f80dba8d 1752
eb9f8d7f
AF
1753.. option:: sync=str
1754
1755 Whether, and what type, of synchronous I/O to use for writes. The allowed
1756 values are:
1757
1758 **none**
1759 Do not use synchronous IO, the default.
1760
1761 **0**
1762 Same as **none**.
1763
1764 **sync**
1765 Use synchronous file IO. For the majority of I/O engines,
1766 this means using O_SYNC.
1767
1768 **1**
1769 Same as **sync**.
1770
1771 **dsync**
1772 Use synchronous data IO. For the majority of I/O engines,
1773 this means using O_DSYNC.
f80dba8d 1774
f80dba8d
MT
1775
1776.. option:: iomem=str, mem=str
1777
1778 Fio can use various types of memory as the I/O unit buffer. The allowed
1779 values are:
1780
1781 **malloc**
1782 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1783 type.
1784
1785 **shm**
1786 Use shared memory as the buffers. Allocated through
1787 :manpage:`shmget(2)`.
1788
1789 **shmhuge**
1790 Same as shm, but use huge pages as backing.
1791
1792 **mmap**
22413915 1793 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
f80dba8d
MT
1794 be file backed if a filename is given after the option. The format
1795 is `mem=mmap:/path/to/file`.
1796
1797 **mmaphuge**
1798 Use a memory mapped huge file as the buffer backing. Append filename
1799 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1800
1801 **mmapshared**
1802 Same as mmap, but use a MMAP_SHARED mapping.
1803
03553853
YR
1804 **cudamalloc**
1805 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
f50fbdda 1806 The :option:`ioengine` must be `rdma`.
03553853 1807
f80dba8d
MT
1808 The area allocated is a function of the maximum allowed bs size for the job,
1809 multiplied by the I/O depth given. Note that for **shmhuge** and
1810 **mmaphuge** to work, the system must have free huge pages allocated. This
1811 can normally be checked and set by reading/writing
1812 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1813 is 4MiB in size. So to calculate the number of huge pages you need for a
1814 given job file, add up the I/O depth of all jobs (normally one unless
1815 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1816 that number by the huge page size. You can see the size of the huge pages in
1817 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1818 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1819 see :option:`hugepage-size`.
1820
1821 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1822 should point there. So if it's mounted in :file:`/huge`, you would use
1823 `mem=mmaphuge:/huge/somefile`.
1824
f50fbdda 1825.. option:: iomem_align=int, mem_align=int
f80dba8d
MT
1826
1827 This indicates the memory alignment of the I/O memory buffers. Note that
1828 the given alignment is applied to the first I/O unit buffer, if using
1829 :option:`iodepth` the alignment of the following buffers are given by the
1830 :option:`bs` used. In other words, if using a :option:`bs` that is a
1831 multiple of the page sized in the system, all buffers will be aligned to
1832 this value. If using a :option:`bs` that is not page aligned, the alignment
1833 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1834 :option:`bs` used.
1835
1836.. option:: hugepage-size=int
1837
1838 Defines the size of a huge page. Must at least be equal to the system
1839 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1840 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1841 preferred way to set this to avoid setting a non-pow-2 bad value.
1842
1843.. option:: lockmem=int
1844
1845 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1846 simulate a smaller amount of memory. The amount specified is per worker.
1847
1848
1849I/O size
1850~~~~~~~~
1851
1852.. option:: size=int
1853
79591fa9
TK
1854 The total size of file I/O for each thread of this job. Fio will run until
1855 this many bytes has been transferred, unless runtime is limited by other options
1856 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1857 Fio will divide this size between the available files determined by options
1858 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1859 specified by the job. If the result of division happens to be 0, the size is
c4aa2d08 1860 set to the physical size of the given files or devices if they exist.
79591fa9 1861 If this option is not specified, fio will use the full size of the given
f80dba8d
MT
1862 files or devices. If the files do not exist, size must be given. It is also
1863 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
193aaf6a
G
1864 given, fio will use 20% of the full size of the given files or devices.
1865 In ZBD mode, value can also be set as number of zones using 'z'.
9d25d068
SW
1866 Can be combined with :option:`offset` to constrain the start and end range
1867 that I/O will be done within.
f80dba8d
MT
1868
1869.. option:: io_size=int, io_limit=int
1870
1871 Normally fio operates within the region set by :option:`size`, which means
1872 that the :option:`size` option sets both the region and size of I/O to be
1873 performed. Sometimes that is not what you want. With this option, it is
1874 possible to define just the amount of I/O that fio should do. For instance,
1875 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1876 will perform I/O within the first 20GiB but exit when 5GiB have been
1877 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1878 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1879 the 0..20GiB region.
1880
7fdd97ca 1881.. option:: filesize=irange(int)
f80dba8d
MT
1882
1883 Individual file sizes. May be a range, in which case fio will select sizes
1884 for files at random within the given range and limited to :option:`size` in
1885 total (if that is given). If not given, each created file is the same size.
79591fa9
TK
1886 This option overrides :option:`size` in terms of file size, which means
1887 this value is used as a fixed size or possible range of each file.
f80dba8d
MT
1888
1889.. option:: file_append=bool
1890
1891 Perform I/O after the end of the file. Normally fio will operate within the
1892 size of a file. If this option is set, then fio will append to the file
1893 instead. This has identical behavior to setting :option:`offset` to the size
1894 of a file. This option is ignored on non-regular files.
1895
1896.. option:: fill_device=bool, fill_fs=bool
1897
1898 Sets size to something really large and waits for ENOSPC (no space left on
418f5399
MB
1899 device) or EDQUOT (disk quota exceeded)
1900 as the terminating condition. Only makes sense with sequential
f80dba8d
MT
1901 write. For a read workload, the mount point will be filled first then I/O
1902 started on the result. This option doesn't make sense if operating on a raw
1903 device node, since the size of that is already known by the file system.
1904 Additionally, writing beyond end-of-device will not return ENOSPC there.
1905
1906
1907I/O engine
1908~~~~~~~~~~
1909
1910.. option:: ioengine=str
1911
1912 Defines how the job issues I/O to the file. The following types are defined:
1913
1914 **sync**
1915 Basic :manpage:`read(2)` or :manpage:`write(2)`
1916 I/O. :manpage:`lseek(2)` is used to position the I/O location.
54227e6b 1917 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
f80dba8d
MT
1918
1919 **psync**
1920 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1921 all supported operating systems except for Windows.
1922
1923 **vsync**
1924 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
c60ebc45 1925 queuing by coalescing adjacent I/Os into a single submission.
f80dba8d
MT
1926
1927 **pvsync**
1928 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1929
1930 **pvsync2**
1931 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1932
029b42ac
JA
1933 **io_uring**
1934 Fast Linux native asynchronous I/O. Supports async IO
1935 for both direct and buffered IO.
1936 This engine defines engine specific options.
1937
f80dba8d
MT
1938 **libaio**
1939 Linux native asynchronous I/O. Note that Linux may only support
22413915 1940 queued behavior with non-buffered I/O (set ``direct=1`` or
f80dba8d
MT
1941 ``buffered=0``).
1942 This engine defines engine specific options.
1943
1944 **posixaio**
1945 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1946 :manpage:`aio_write(3)`.
1947
1948 **solarisaio**
1949 Solaris native asynchronous I/O.
1950
1951 **windowsaio**
1952 Windows native asynchronous I/O. Default on Windows.
1953
1954 **mmap**
1955 File is memory mapped with :manpage:`mmap(2)` and data copied
1956 to/from using :manpage:`memcpy(3)`.
1957
1958 **splice**
1959 :manpage:`splice(2)` is used to transfer the data and
1960 :manpage:`vmsplice(2)` to transfer data from user space to the
1961 kernel.
1962
1963 **sg**
1964 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1965 ioctl, or if the target is an sg character device we use
1966 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
f50fbdda 1967 I/O. Requires :option:`filename` option to specify either block or
3740cfc8 1968 character devices. This engine supports trim operations.
52b81b7c 1969 The sg engine includes engine specific options.
f80dba8d 1970
2455851d
SK
1971 **libzbc**
1972 Read, write, trim and ZBC/ZAC operations to a zoned
1973 block device using libzbc library. The target can be
1974 either an SG character device or a block device file.
1975
f80dba8d
MT
1976 **null**
1977 Doesn't transfer any data, just pretends to. This is mainly used to
1978 exercise fio itself and for debugging/testing purposes.
1979
1980 **net**
1981 Transfer over the network to given ``host:port``. Depending on the
1982 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1983 :option:`listen` and :option:`filename` options are used to specify
1984 what sort of connection to make, while the :option:`protocol` option
1985 determines which protocol will be used. This engine defines engine
1986 specific options.
1987
1988 **netsplice**
1989 Like **net**, but uses :manpage:`splice(2)` and
1990 :manpage:`vmsplice(2)` to map data and send/receive.
1991 This engine defines engine specific options.
1992
1993 **cpuio**
1994 Doesn't transfer any data, but burns CPU cycles according to the
9de473a8
EV
1995 :option:`cpuload`, :option:`cpuchunks` and :option:`cpumode` options.
1996 Setting :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
71aa48eb 1997 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
f50fbdda 1998 to get desired CPU usage, as the cpuload only loads a
f80dba8d
MT
1999 single CPU at the desired rate. A job never finishes unless there is
2000 at least one non-cpuio job.
9de473a8
EV
2001 Setting :option:`cpumode`\=qsort replace the default noop instructions loop
2002 by a qsort algorithm to consume more energy.
f80dba8d 2003
f80dba8d
MT
2004 **rdma**
2005 The RDMA I/O engine supports both RDMA memory semantics
2006 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
609ac152
SB
2007 InfiniBand, RoCE and iWARP protocols. This engine defines engine
2008 specific options.
f80dba8d
MT
2009
2010 **falloc**
2011 I/O engine that does regular fallocate to simulate data transfer as
2012 fio ioengine.
2013
2014 DDIR_READ
2015 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
2016
2017 DDIR_WRITE
2018 does fallocate(,mode = 0).
2019
2020 DDIR_TRIM
2021 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
2022
761cd093
SW
2023 **ftruncate**
2024 I/O engine that sends :manpage:`ftruncate(2)` operations in response
2025 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
f50fbdda 2026 size to the current block offset. :option:`blocksize` is ignored.
761cd093 2027
f80dba8d
MT
2028 **e4defrag**
2029 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
2030 defragment activity in request to DDIR_WRITE event.
2031
f3f96717
IF
2032 **rados**
2033 I/O engine supporting direct access to Ceph Reliable Autonomic
2034 Distributed Object Store (RADOS) via librados. This ioengine
2035 defines engine specific options.
2036
f80dba8d
MT
2037 **rbd**
2038 I/O engine supporting direct access to Ceph Rados Block Devices
2039 (RBD) via librbd without the need to use the kernel rbd driver. This
2040 ioengine defines engine specific options.
2041
c2f6a13d
LMB
2042 **http**
2043 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
2044 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
2045
2046 This engine only supports direct IO of iodepth=1; you need to scale this
2047 via numjobs. blocksize defines the size of the objects to be created.
2048
2049 TRIM is translated to object deletion.
2050
f80dba8d 2051 **gfapi**
ac8ca2af
SW
2052 Using GlusterFS libgfapi sync interface to direct access to
2053 GlusterFS volumes without having to go through FUSE. This ioengine
f80dba8d
MT
2054 defines engine specific options.
2055
2056 **gfapi_async**
ac8ca2af
SW
2057 Using GlusterFS libgfapi async interface to direct access to
2058 GlusterFS volumes without having to go through FUSE. This ioengine
f80dba8d
MT
2059 defines engine specific options.
2060
2061 **libhdfs**
f50fbdda 2062 Read and write through Hadoop (HDFS). The :option:`filename` option
f80dba8d
MT
2063 is used to specify host,port of the hdfs name-node to connect. This
2064 engine interprets offsets a little differently. In HDFS, files once
e25c0c91
SW
2065 created cannot be modified so random writes are not possible. To
2066 imitate this the libhdfs engine expects a bunch of small files to be
2067 created over HDFS and will randomly pick a file from them
2068 based on the offset generated by fio backend (see the example
f80dba8d 2069 job file to create such files, use ``rw=write`` option). Please
e25c0c91
SW
2070 note, it may be necessary to set environment variables to work
2071 with HDFS/libhdfs properly. Each job uses its own connection to
f80dba8d
MT
2072 HDFS.
2073
2074 **mtd**
2075 Read, write and erase an MTD character device (e.g.,
2076 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
2077 underlying device type, the I/O may have to go in a certain pattern,
2078 e.g., on NAND, writing sequentially to erase blocks and discarding
c298ee71 2079 before overwriting. The `trimwrite` mode works well for this
f80dba8d
MT
2080 constraint.
2081
2082 **pmemblk**
2083 Read and write using filesystem DAX to a file on a filesystem
363a5f65 2084 mounted with DAX on a persistent memory device through the PMDK
f80dba8d
MT
2085 libpmemblk library.
2086
2087 **dev-dax**
2088 Read and write using device DAX to a persistent memory device (e.g.,
363a5f65 2089 /dev/dax0.0) through the PMDK libpmem library.
f80dba8d
MT
2090
2091 **external**
2092 Prefix to specify loading an external I/O engine object file. Append
c60ebc45 2093 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
d243fd6d
TK
2094 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
2095 absolute or relative. See :file:`engines/skeleton_external.c` for
2096 details of writing an external I/O engine.
f80dba8d 2097
1216cc5a 2098 **filecreate**
b71968b1 2099 Simply create the files and do no I/O to them. You still need to
1216cc5a 2100 set `filesize` so that all the accounting still occurs, but no
b71968b1 2101 actual I/O will be done other than creating the file.
f80dba8d 2102
73ccd14e
SF
2103 **filestat**
2104 Simply do stat() and do no I/O to the file. You need to set 'filesize'
2105 and 'nrfiles', so that files will be created.
2106 This engine is to measure file lookup and meta data access.
2107
5561e9dd
FS
2108 **filedelete**
2109 Simply delete the files by unlink() and do no I/O to them. You need to set 'filesize'
2110 and 'nrfiles', so that the files will be created.
2111 This engine is to measure file delete.
2112
ae0db592
TI
2113 **libpmem**
2114 Read and write using mmap I/O to a file on a filesystem
363a5f65 2115 mounted with DAX on a persistent memory device through the PMDK
ae0db592
TI
2116 libpmem library.
2117
a40e7a59
GB
2118 **ime_psync**
2119 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2120 This engine is very basic and issues calls to IME whenever an IO is
2121 queued.
2122
2123 **ime_psyncv**
2124 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2125 This engine uses iovecs and will try to stack as much IOs as possible
2126 (if the IOs are "contiguous" and the IO depth is not exceeded)
2127 before issuing a call to IME.
2128
2129 **ime_aio**
2130 Asynchronous read and write using DDN's Infinite Memory Engine (IME).
2131 This engine will try to stack as much IOs as possible by creating
2132 requests for IME. FIO will then decide when to commit these requests.
247ef2aa
KZ
2133 **libiscsi**
2134 Read and write iscsi lun with libiscsi.
d643a1e2 2135 **nbd**
f2d6de5d 2136 Read and write a Network Block Device (NBD).
a40e7a59 2137
10756b2c
BS
2138 **libcufile**
2139 I/O engine supporting libcufile synchronous access to nvidia-fs and a
2140 GPUDirect Storage-supported filesystem. This engine performs
2141 I/O without transferring buffers between user-space and the kernel,
2142 unless :option:`verify` is set or :option:`cuda_io` is `posix`.
2143 :option:`iomem` must not be `cudamalloc`. This ioengine defines
2144 engine specific options.
c363fdd7
JL
2145 **dfs**
2146 I/O engine supporting asynchronous read and write operations to the
2147 DAOS File System (DFS) via libdfs.
10756b2c 2148
9326926b
TG
2149 **nfs**
2150 I/O engine supporting asynchronous read and write operations to
2151 NFS filesystems from userspace via libnfs. This is useful for
2152 achieving higher concurrency and thus throughput than is possible
2153 via kernel NFS.
2154
b50590bc
EV
2155 **exec**
2156 Execute 3rd party tools. Could be used to perform monitoring during jobs runtime.
2157
f80dba8d
MT
2158I/O engine specific parameters
2159~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2160
2161In addition, there are some parameters which are only valid when a specific
f50fbdda
TK
2162:option:`ioengine` is in use. These are used identically to normal parameters,
2163with the caveat that when used on the command line, they must come after the
f80dba8d
MT
2164:option:`ioengine` that defines them is selected.
2165
e9f6567a
DLM
2166.. option:: cmdprio_percentage=int[,int] : [io_uring] [libaio]
2167
2168 Set the percentage of I/O that will be issued with the highest priority.
2169 Default: 0. A single value applies to reads and writes. Comma-separated
acf2e2d9
NC
2170 values may be specified for reads and writes. For this option to be
2171 effective, NCQ priority must be supported and enabled, and `direct=1'
bebf1407
NC
2172 option must be used. fio must also be run as the root user. Unlike
2173 slat/clat/lat stats, which can be tracked and reported independently, per
2174 priority stats only track and report a single type of latency. By default,
2175 completion latency (clat) will be reported, if :option:`lat_percentiles` is
2176 set, total latency (lat) will be reported.
029b42ac 2177
12f9d54a
DLM
2178.. option:: cmdprio_class=int[,int] : [io_uring] [libaio]
2179
2180 Set the I/O priority class to use for I/Os that must be issued with
a48f0cc7
DLM
2181 a priority when :option:`cmdprio_percentage` or
2182 :option:`cmdprio_bssplit` is set. If not specified when
2183 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2184 this defaults to the highest priority class. A single value applies
2185 to reads and writes. Comma-separated values may be specified for
2186 reads and writes. See :manpage:`ionice(1)`. See also the
2187 :option:`prioclass` option.
12f9d54a
DLM
2188
2189.. option:: cmdprio=int[,int] : [io_uring] [libaio]
2190
2191 Set the I/O priority value to use for I/Os that must be issued with
a48f0cc7
DLM
2192 a priority when :option:`cmdprio_percentage` or
2193 :option:`cmdprio_bssplit` is set. If not specified when
2194 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2195 this defaults to 0.
12f9d54a
DLM
2196 Linux limits us to a positive value between 0 and 7, with 0 being the
2197 highest. A single value applies to reads and writes. Comma-separated
2198 values may be specified for reads and writes. See :manpage:`ionice(1)`.
2199 Refer to an appropriate manpage for other operating systems since
2200 meaning of priority may differ. See also the :option:`prio` option.
2201
a48f0cc7
DLM
2202.. option:: cmdprio_bssplit=str[,str] : [io_uring] [libaio]
2203 To get a finer control over I/O priority, this option allows
2204 specifying the percentage of IOs that must have a priority set
2205 depending on the block size of the IO. This option is useful only
2206 when used together with the :option:`bssplit` option, that is,
2207 multiple different block sizes are used for reads and writes.
2208 The format for this option is the same as the format of the
2209 :option:`bssplit` option, with the exception that values for
2210 trim IOs are ignored. This option is mutually exclusive with the
2211 :option:`cmdprio_percentage` option.
2212
029b42ac
JA
2213.. option:: fixedbufs : [io_uring]
2214
b2a432bf
PC
2215 If fio is asked to do direct IO, then Linux will map pages for each
2216 IO call, and release them when IO is done. If this option is set, the
2217 pages are pre-mapped before IO is started. This eliminates the need to
2218 map and release for each IO. This is more efficient, and reduces the
2219 IO latency as well.
2220
2221.. option:: hipri : [io_uring]
2222
2223 If this option is set, fio will attempt to use polled IO completions.
2224 Normal IO completions generate interrupts to signal the completion of
2225 IO, polled completions do not. Hence they are require active reaping
2226 by the application. The benefits are more efficient IO for high IOPS
2227 scenarios, and lower latencies for low queue depth IO.
029b42ac 2228
5ffd5626 2229.. option:: registerfiles : [io_uring]
2c870598 2230
5ffd5626
JA
2231 With this option, fio registers the set of files being used with the
2232 kernel. This avoids the overhead of managing file counts in the kernel,
2233 making the submission and completion part more lightweight. Required
2234 for the below :option:`sqthread_poll` option.
2235
029b42ac
JA
2236.. option:: sqthread_poll : [io_uring]
2237
2238 Normally fio will submit IO by issuing a system call to notify the
2239 kernel of available items in the SQ ring. If this option is set, the
2240 act of submitting IO will be done by a polling thread in the kernel.
2241 This frees up cycles for fio, at the cost of using more CPU in the
2242 system.
2243
2244.. option:: sqthread_poll_cpu : [io_uring]
2245
2246 When :option:`sqthread_poll` is set, this option provides a way to
2247 define which CPU should be used for the polling thread.
2248
f80dba8d
MT
2249.. option:: userspace_reap : [libaio]
2250
2251 Normally, with the libaio engine in use, fio will use the
2252 :manpage:`io_getevents(2)` system call to reap newly returned events. With
2253 this flag turned on, the AIO ring will be read directly from user-space to
2254 reap events. The reaping mode is only enabled when polling for a minimum of
c60ebc45 2255 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
f80dba8d 2256
9d25d068 2257.. option:: hipri : [pvsync2]
f80dba8d
MT
2258
2259 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2260 than normal.
2261
a0679ce5
SB
2262.. option:: hipri_percentage : [pvsync2]
2263
f50fbdda 2264 When hipri is set this determines the probability of a pvsync2 I/O being high
a0679ce5
SB
2265 priority. The default is 100%.
2266
7d42e66e
KK
2267.. option:: nowait : [pvsync2] [libaio] [io_uring]
2268
2269 By default if a request cannot be executed immediately (e.g. resource starvation,
2270 waiting on locks) it is queued and the initiating process will be blocked until
2271 the required resource becomes free.
2272
2273 This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and
2274 the call will return instantly with EAGAIN or a partial result rather than waiting.
2275
2276 It is useful to also use ignore_error=EAGAIN when using this option.
2277
2278 Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2.
2279 They return EOPNOTSUP instead of EAGAIN.
2280
2281 For cached I/O, using this option usually means a request operates only with
2282 cached data. Currently the RWF_NOWAIT flag does not supported for cached write.
2283
2284 For direct I/O, requests will only succeed if cache invalidation isn't required,
2285 file blocks are fully allocated and the disk request could be issued immediately.
2286
f80dba8d
MT
2287.. option:: cpuload=int : [cpuio]
2288
da19cdb4
TK
2289 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2290 option when using cpuio I/O engine.
f80dba8d
MT
2291
2292.. option:: cpuchunks=int : [cpuio]
2293
2294 Split the load into cycles of the given time. In microseconds.
2295
2296.. option:: exit_on_io_done=bool : [cpuio]
2297
2298 Detect when I/O threads are done, then exit.
2299
f80dba8d
MT
2300.. option:: namenode=str : [libhdfs]
2301
22413915 2302 The hostname or IP address of a HDFS cluster namenode to contact.
f80dba8d
MT
2303
2304.. option:: port=int
2305
f50fbdda
TK
2306 [libhdfs]
2307
2308 The listening port of the HFDS cluster namenode.
2309
f80dba8d
MT
2310 [netsplice], [net]
2311
2312 The TCP or UDP port to bind to or connect to. If this is used with
2313 :option:`numjobs` to spawn multiple instances of the same job type, then
2314 this will be the starting port number since fio will use a range of
2315 ports.
2316
e4c4625f 2317 [rdma], [librpma_*]
609ac152
SB
2318
2319 The port to use for RDMA-CM communication. This should be the same value
2320 on the client and the server side.
2321
2322.. option:: hostname=str : [netsplice] [net] [rdma]
f80dba8d 2323
609ac152
SB
2324 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
2325 is a TCP listener or UDP reader, the hostname is not used and must be omitted
f50fbdda 2326 unless it is a valid UDP multicast address.
f80dba8d 2327
e4c4625f
JM
2328.. option:: serverip=str : [librpma_*]
2329
2330 The IP address to be used for RDMA-CM based I/O.
2331
2332.. option:: direct_write_to_pmem=bool : [librpma_*]
2333
2334 Set to 1 only when Direct Write to PMem from the remote host is possible.
2335 Otherwise, set to 0.
2336
6a229978
OS
2337.. option:: busy_wait_polling=bool : [librpma_*_server]
2338
2339 Set to 0 to wait for completion instead of busy-wait polling completion.
2340 Default: 1.
2341
f80dba8d
MT
2342.. option:: interface=str : [netsplice] [net]
2343
2344 The IP address of the network interface used to send or receive UDP
2345 multicast.
2346
2347.. option:: ttl=int : [netsplice] [net]
2348
2349 Time-to-live value for outgoing UDP multicast packets. Default: 1.
2350
2351.. option:: nodelay=bool : [netsplice] [net]
2352
2353 Set TCP_NODELAY on TCP connections.
2354
f50fbdda 2355.. option:: protocol=str, proto=str : [netsplice] [net]
f80dba8d
MT
2356
2357 The network protocol to use. Accepted values are:
2358
2359 **tcp**
2360 Transmission control protocol.
2361 **tcpv6**
2362 Transmission control protocol V6.
2363 **udp**
2364 User datagram protocol.
2365 **udpv6**
2366 User datagram protocol V6.
2367 **unix**
2368 UNIX domain socket.
2369
2370 When the protocol is TCP or UDP, the port must also be given, as well as the
2371 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
f50fbdda 2372 normal :option:`filename` option should be used and the port is invalid.
f80dba8d 2373
e9184ec1 2374.. option:: listen : [netsplice] [net]
f80dba8d
MT
2375
2376 For TCP network connections, tell fio to listen for incoming connections
2377 rather than initiating an outgoing connection. The :option:`hostname` must
2378 be omitted if this option is used.
2379
e9184ec1 2380.. option:: pingpong : [netsplice] [net]
f80dba8d
MT
2381
2382 Normally a network writer will just continue writing data, and a network
2383 reader will just consume packages. If ``pingpong=1`` is set, a writer will
2384 send its normal payload to the reader, then wait for the reader to send the
2385 same payload back. This allows fio to measure network latencies. The
2386 submission and completion latencies then measure local time spent sending or
2387 receiving, and the completion latency measures how long it took for the
2388 other end to receive and send back. For UDP multicast traffic
2389 ``pingpong=1`` should only be set for a single reader when multiple readers
2390 are listening to the same address.
2391
e9184ec1 2392.. option:: window_size : [netsplice] [net]
f80dba8d
MT
2393
2394 Set the desired socket buffer size for the connection.
2395
e9184ec1 2396.. option:: mss : [netsplice] [net]
f80dba8d
MT
2397
2398 Set the TCP maximum segment size (TCP_MAXSEG).
2399
2400.. option:: donorname=str : [e4defrag]
2401
730bd7d9 2402 File will be used as a block donor (swap extents between files).
f80dba8d
MT
2403
2404.. option:: inplace=int : [e4defrag]
2405
2406 Configure donor file blocks allocation strategy:
2407
2408 **0**
2409 Default. Preallocate donor's file on init.
2410 **1**
2b455dbf 2411 Allocate space immediately inside defragment event, and free right
f80dba8d
MT
2412 after event.
2413
f3f96717 2414.. option:: clustername=str : [rbd,rados]
f80dba8d
MT
2415
2416 Specifies the name of the Ceph cluster.
2417
2418.. option:: rbdname=str : [rbd]
2419
2420 Specifies the name of the RBD.
2421
f3f96717 2422.. option:: pool=str : [rbd,rados]
f80dba8d 2423
f3f96717 2424 Specifies the name of the Ceph pool containing RBD or RADOS data.
f80dba8d 2425
f3f96717 2426.. option:: clientname=str : [rbd,rados]
f80dba8d
MT
2427
2428 Specifies the username (without the 'client.' prefix) used to access the
2429 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2430 the full *type.id* string. If no type. prefix is given, fio will add
2431 'client.' by default.
2432
f3f96717
IF
2433.. option:: busy_poll=bool : [rbd,rados]
2434
2435 Poll store instead of waiting for completion. Usually this provides better
2436 throughput at cost of higher(up to 100%) CPU utilization.
2437
2b728756
AK
2438.. option:: touch_objects=bool : [rados]
2439
2440 During initialization, touch (create if do not exist) all objects (files).
2441 Touching all objects affects ceph caches and likely impacts test results.
2442 Enabled by default.
2443
f80dba8d
MT
2444.. option:: skip_bad=bool : [mtd]
2445
2446 Skip operations against known bad blocks.
2447
2448.. option:: hdfsdirectory : [libhdfs]
2449
2450 libhdfs will create chunk in this HDFS directory.
2451
2452.. option:: chunk_size : [libhdfs]
2453
2b455dbf 2454 The size of the chunk to use for each file.
f80dba8d 2455
609ac152
SB
2456.. option:: verb=str : [rdma]
2457
2458 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2459 values are write, read, send and recv. These correspond to the equivalent
2460 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2461 specified on the client side of the connection. See the examples folder.
2462
2463.. option:: bindname=str : [rdma]
2464
2465 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2466 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2467 will be passed into the rdma_bind_addr() function and on the client site it
2468 will be used in the rdma_resolve_add() function. This can be useful when
2469 multiple paths exist between the client and the server or in certain loopback
2470 configurations.
f80dba8d 2471
93a13ba5 2472.. option:: stat_type=str : [filestat]
c446eff0 2473
93a13ba5
TK
2474 Specify stat system call type to measure lookup/getattr performance.
2475 Default is **stat** for :manpage:`stat(2)`.
c446eff0 2476
52b81b7c
KD
2477.. option:: readfua=bool : [sg]
2478
2479 With readfua option set to 1, read operations include
2480 the force unit access (fua) flag. Default is 0.
2481
2482.. option:: writefua=bool : [sg]
2483
2484 With writefua option set to 1, write operations include
2485 the force unit access (fua) flag. Default is 0.
2486
2c3a9150 2487.. option:: sg_write_mode=str : [sg]
3740cfc8 2488
2c3a9150
VF
2489 Specify the type of write commands to issue. This option can take three values:
2490
2491 **write**
2492 This is the default where write opcodes are issued as usual.
2493 **verify**
2494 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2495 directs the device to carry out a medium verification with no data
2496 comparison. The writefua option is ignored with this selection.
2497 **same**
2498 Issue WRITE SAME commands. This transfers a single block to the device
2499 and writes this same block of data to a contiguous sequence of LBAs
2500 beginning at the specified offset. fio's block size parameter specifies
2501 the amount of data written with each command. However, the amount of data
2502 actually transferred to the device is equal to the device's block
2503 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2504 write 16 sectors with each command. fio will still generate 8k of data
2505 for each command but only the first 512 bytes will be used and
2506 transferred to the device. The writefua option is ignored with this
2507 selection.
52b81b7c 2508
e493ceae
JA
2509.. option:: hipri : [sg]
2510
2511 If this option is set, fio will attempt to use polled IO completions.
2512 This will have a similar effect as (io_uring)hipri. Only SCSI READ and
2513 WRITE commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor
2514 VERIFY). Older versions of the Linux sg driver that do not support
2515 hipri will simply ignore this flag and do normal IO. The Linux SCSI
2516 Low Level Driver (LLD) that "owns" the device also needs to support
2517 hipri (also known as iopoll and mq_poll). The MegaRAID driver is an
2518 example of a SCSI LLD. Default: clear (0) which does normal
2519 (interrupted based) IO.
2520
c2f6a13d
LMB
2521.. option:: http_host=str : [http]
2522
2523 Hostname to connect to. For S3, this could be the bucket hostname.
2524 Default is **localhost**
2525
2526.. option:: http_user=str : [http]
2527
2528 Username for HTTP authentication.
2529
2530.. option:: http_pass=str : [http]
2531
2532 Password for HTTP authentication.
2533
09fd2966 2534.. option:: https=str : [http]
c2f6a13d 2535
09fd2966
LMB
2536 Enable HTTPS instead of http. *on* enables HTTPS; *insecure*
2537 will enable HTTPS, but disable SSL peer verification (use with
2538 caution!). Default is **off**
c2f6a13d 2539
09fd2966 2540.. option:: http_mode=str : [http]
c2f6a13d 2541
09fd2966
LMB
2542 Which HTTP access mode to use: *webdav*, *swift*, or *s3*.
2543 Default is **webdav**
c2f6a13d
LMB
2544
2545.. option:: http_s3_region=str : [http]
2546
2547 The S3 region/zone string.
2548 Default is **us-east-1**
2549
2550.. option:: http_s3_key=str : [http]
2551
2552 The S3 secret key.
2553
2554.. option:: http_s3_keyid=str : [http]
2555
2556 The S3 key/access id.
2557
09fd2966
LMB
2558.. option:: http_swift_auth_token=str : [http]
2559
2560 The Swift auth token. See the example configuration file on how
2561 to retrieve this.
2562
c2f6a13d
LMB
2563.. option:: http_verbose=int : [http]
2564
2565 Enable verbose requests from libcurl. Useful for debugging. 1
2566 turns on verbose logging from libcurl, 2 additionally enables
2567 HTTP IO tracing. Default is **0**
2568
f2d6de5d
RJ
2569.. option:: uri=str : [nbd]
2570
2571 Specify the NBD URI of the server to test. The string
2572 is a standard NBD URI
2573 (see https://github.com/NetworkBlockDevice/nbd/tree/master/doc).
2574 Example URIs: nbd://localhost:10809
2575 nbd+unix:///?socket=/tmp/socket
2576 nbds://tlshost/exportname
2577
10756b2c
BS
2578.. option:: gpu_dev_ids=str : [libcufile]
2579
2580 Specify the GPU IDs to use with CUDA. This is a colon-separated list of
2581 int. GPUs are assigned to workers roundrobin. Default is 0.
2582
2583.. option:: cuda_io=str : [libcufile]
2584
2585 Specify the type of I/O to use with CUDA. Default is **cufile**.
2586
2587 **cufile**
2588 Use libcufile and nvidia-fs. This option performs I/O directly
2589 between a GPUDirect Storage filesystem and GPU buffers,
2590 avoiding use of a bounce buffer. If :option:`verify` is set,
2591 cudaMemcpy is used to copy verificaton data between RAM and GPU.
2592 Verification data is copied from RAM to GPU before a write
2593 and from GPU to RAM after a read. :option:`direct` must be 1.
2594 **posix**
2595 Use POSIX to perform I/O with a RAM buffer, and use cudaMemcpy
2596 to transfer data between RAM and the GPUs. Data is copied from
2597 GPU to RAM before a write and copied from RAM to GPU after a
2598 read. :option:`verify` does not affect use of cudaMemcpy.
2599
c363fdd7
JL
2600.. option:: pool=str : [dfs]
2601
2819492b 2602 Specify the label or UUID of the DAOS pool to connect to.
c363fdd7
JL
2603
2604.. option:: cont=str : [dfs]
2605
2819492b 2606 Specify the label or UUID of the DAOS container to open.
c363fdd7
JL
2607
2608.. option:: chunk_size=int : [dfs]
2609
2610 Specificy a different chunk size (in bytes) for the dfs file.
2611 Use DAOS container's chunk size by default.
2612
2613.. option:: object_class=str : [dfs]
2614
2615 Specificy a different object class for the dfs file.
2616 Use DAOS container's object class by default.
2617
9326926b
TG
2618.. option:: nfs_url=str : [nfs]
2619
2620 URL in libnfs format, eg nfs://<server|ipv4|ipv6>/path[?arg=val[&arg=val]*]
2621 Refer to the libnfs README for more details.
2622
b50590bc
EV
2623.. option:: program=str : [exec]
2624
2625 Specify the program to execute.
2626
2627.. option:: arguments=str : [exec]
2628
2629 Specify arguments to pass to program.
2630 Some special variables can be expanded to pass fio's job details to the program.
2631
2632 **%r**
2633 Replaced by the duration of the job in seconds.
2634 **%n**
2635 Replaced by the name of the job.
2636
2637.. option:: grace_time=int : [exec]
2638
2639 Specify the time between the SIGTERM and SIGKILL signals. Default is 1 second.
2640
81c7079c 2641.. option:: std_redirect=bool : [exec]
b50590bc
EV
2642
2643 If set, stdout and stderr streams are redirected to files named from the job name. Default is true.
2644
f80dba8d
MT
2645I/O depth
2646~~~~~~~~~
2647
2648.. option:: iodepth=int
2649
2650 Number of I/O units to keep in flight against the file. Note that
2651 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
c60ebc45 2652 for small degrees when :option:`verify_async` is in use). Even async
f80dba8d
MT
2653 engines may impose OS restrictions causing the desired depth not to be
2654 achieved. This may happen on Linux when using libaio and not setting
9207a0cb 2655 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
f80dba8d
MT
2656 eye on the I/O depth distribution in the fio output to verify that the
2657 achieved depth is as expected. Default: 1.
2658
2659.. option:: iodepth_batch_submit=int, iodepth_batch=int
2660
2661 This defines how many pieces of I/O to submit at once. It defaults to 1
2662 which means that we submit each I/O as soon as it is available, but can be
2663 raised to submit bigger batches of I/O at the time. If it is set to 0 the
2664 :option:`iodepth` value will be used.
2665
2666.. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
2667
2668 This defines how many pieces of I/O to retrieve at once. It defaults to 1
2669 which means that we'll ask for a minimum of 1 I/O in the retrieval process
2670 from the kernel. The I/O retrieval will go on until we hit the limit set by
2671 :option:`iodepth_low`. If this variable is set to 0, then fio will always
2672 check for completed events before queuing more I/O. This helps reduce I/O
2673 latency, at the cost of more retrieval system calls.
2674
2675.. option:: iodepth_batch_complete_max=int
2676
2677 This defines maximum pieces of I/O to retrieve at once. This variable should
9207a0cb 2678 be used along with :option:`iodepth_batch_complete_min`\=int variable,
f80dba8d 2679 specifying the range of min and max amount of I/O which should be
730bd7d9 2680 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
f80dba8d
MT
2681 value.
2682
2683 Example #1::
2684
2685 iodepth_batch_complete_min=1
2686 iodepth_batch_complete_max=<iodepth>
2687
2688 which means that we will retrieve at least 1 I/O and up to the whole
2689 submitted queue depth. If none of I/O has been completed yet, we will wait.
2690
2691 Example #2::
2692
2693 iodepth_batch_complete_min=0
2694 iodepth_batch_complete_max=<iodepth>
2695
2696 which means that we can retrieve up to the whole submitted queue depth, but
2697 if none of I/O has been completed yet, we will NOT wait and immediately exit
2698 the system call. In this example we simply do polling.
2699
2700.. option:: iodepth_low=int
2701
2702 The low water mark indicating when to start filling the queue
2703 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2704 attempt to keep the queue full at all times. If :option:`iodepth` is set to
c60ebc45 2705 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
f80dba8d
MT
2706 16 requests, it will let the depth drain down to 4 before starting to fill
2707 it again.
2708
997b5680
SW
2709.. option:: serialize_overlap=bool
2710
2711 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2712 When two or more I/Os are submitted simultaneously, there is no guarantee that
2713 the I/Os will be processed or completed in the submitted order. Further, if
2714 two or more of those I/Os are writes, any overlapping region between them can
2715 become indeterminate/undefined on certain storage. These issues can cause
2716 verification to fail erratically when at least one of the racing I/Os is
2717 changing data and the overlapping region has a non-zero size. Setting
2718 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2719 serializing in-flight I/Os that have a non-zero overlap. Note that setting
ee21ebee 2720 this option can reduce both performance and the :option:`iodepth` achieved.
3d6a6f04
VF
2721
2722 This option only applies to I/Os issued for a single job except when it is
a02ec45a 2723 enabled along with :option:`io_submit_mode`\=offload. In offload mode, fio
3d6a6f04 2724 will check for overlap among all I/Os submitted by offload jobs with :option:`serialize_overlap`
307f2246 2725 enabled.
3d6a6f04
VF
2726
2727 Default: false.
997b5680 2728
f80dba8d
MT
2729.. option:: io_submit_mode=str
2730
2731 This option controls how fio submits the I/O to the I/O engine. The default
2732 is `inline`, which means that the fio job threads submit and reap I/O
2733 directly. If set to `offload`, the job threads will offload I/O submission
2734 to a dedicated pool of I/O threads. This requires some coordination and thus
2735 has a bit of extra overhead, especially for lower queue depth I/O where it
2736 can increase latencies. The benefit is that fio can manage submission rates
2737 independently of the device completion rates. This avoids skewed latency
730bd7d9 2738 reporting if I/O gets backed up on the device side (the coordinated omission
abfd235a
JA
2739 problem). Note that this option cannot reliably be used with async IO
2740 engines.
f80dba8d
MT
2741
2742
2743I/O rate
2744~~~~~~~~
2745
a881438b 2746.. option:: thinktime=time
f80dba8d 2747
f75ede1d
SW
2748 Stall the job for the specified period of time after an I/O has completed before issuing the
2749 next. May be used to simulate processing being done by an application.
947e0fe0 2750 When the unit is omitted, the value is interpreted in microseconds. See
f7942acd 2751 :option:`thinktime_blocks`, :option:`thinktime_iotime` and :option:`thinktime_spin`.
f80dba8d 2752
a881438b 2753.. option:: thinktime_spin=time
f80dba8d
MT
2754
2755 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2756 something with the data received, before falling back to sleeping for the
f75ede1d 2757 rest of the period specified by :option:`thinktime`. When the unit is
947e0fe0 2758 omitted, the value is interpreted in microseconds.
f80dba8d
MT
2759
2760.. option:: thinktime_blocks=int
2761
2762 Only valid if :option:`thinktime` is set - control how many blocks to issue,
f50fbdda
TK
2763 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
2764 fio wait :option:`thinktime` usecs after every block. This effectively makes any
f80dba8d 2765 queue depth setting redundant, since no more than 1 I/O will be queued
f50fbdda 2766 before we have to complete it and do our :option:`thinktime`. In other words, this
f80dba8d 2767 setting effectively caps the queue depth if the latter is larger.
71bfa161 2768
33f42c20
HQ
2769.. option:: thinktime_blocks_type=str
2770
2771 Only valid if :option:`thinktime` is set - control how :option:`thinktime_blocks`
2772 triggers. The default is `complete`, which triggers thinktime when fio completes
2773 :option:`thinktime_blocks` blocks. If this is set to `issue`, then the trigger happens
2774 at the issue side.
2775
f7942acd
SK
2776.. option:: thinktime_iotime=time
2777
2778 Only valid if :option:`thinktime` is set - control :option:`thinktime`
2779 interval by time. The :option:`thinktime` stall is repeated after IOs
2780 are executed for :option:`thinktime_iotime`. For example,
2781 ``--thinktime_iotime=9s --thinktime=1s`` repeat 10-second cycle with IOs
2782 for 9 seconds and stall for 1 second. When the unit is omitted,
2783 :option:`thinktime_iotime` is interpreted as a number of seconds. If
2784 this option is used together with :option:`thinktime_blocks`, the
2785 :option:`thinktime` stall is repeated after :option:`thinktime_iotime`
2786 or after :option:`thinktime_blocks` IOs, whichever happens first.
2787
f80dba8d 2788.. option:: rate=int[,int][,int]
71bfa161 2789
f80dba8d
MT
2790 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2791 suffix rules apply. Comma-separated values may be specified for reads,
2792 writes, and trims as described in :option:`blocksize`.
71bfa161 2793
b25b3464
SW
2794 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2795 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2796 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2797 latter will only limit reads.
2798
f80dba8d 2799.. option:: rate_min=int[,int][,int]
71bfa161 2800
f80dba8d
MT
2801 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2802 to meet this requirement will cause the job to exit. Comma-separated values
2803 may be specified for reads, writes, and trims as described in
2804 :option:`blocksize`.
71bfa161 2805
f80dba8d 2806.. option:: rate_iops=int[,int][,int]
71bfa161 2807
f80dba8d
MT
2808 Cap the bandwidth to this number of IOPS. Basically the same as
2809 :option:`rate`, just specified independently of bandwidth. If the job is
2810 given a block size range instead of a fixed value, the smallest block size
2811 is used as the metric. Comma-separated values may be specified for reads,
2812 writes, and trims as described in :option:`blocksize`.
71bfa161 2813
f80dba8d 2814.. option:: rate_iops_min=int[,int][,int]
71bfa161 2815
f80dba8d
MT
2816 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2817 Comma-separated values may be specified for reads, writes, and trims as
2818 described in :option:`blocksize`.
71bfa161 2819
f80dba8d 2820.. option:: rate_process=str
66c098b8 2821
f80dba8d
MT
2822 This option controls how fio manages rated I/O submissions. The default is
2823 `linear`, which submits I/O in a linear fashion with fixed delays between
c60ebc45 2824 I/Os that gets adjusted based on I/O completion rates. If this is set to
f80dba8d
MT
2825 `poisson`, fio will submit I/O based on a more real world random request
2826 flow, known as the Poisson process
2827 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2828 10^6 / IOPS for the given workload.
71bfa161 2829
1a9bf814
JA
2830.. option:: rate_ignore_thinktime=bool
2831
2832 By default, fio will attempt to catch up to the specified rate setting,
2833 if any kind of thinktime setting was used. If this option is set, then
2834 fio will ignore the thinktime and continue doing IO at the specified
2835 rate, instead of entering a catch-up mode after thinktime is done.
2836
71bfa161 2837
f80dba8d
MT
2838I/O latency
2839~~~~~~~~~~~
71bfa161 2840
a881438b 2841.. option:: latency_target=time
71bfa161 2842
f80dba8d 2843 If set, fio will attempt to find the max performance point that the given
f75ede1d 2844 workload will run at while maintaining a latency below this target. When
947e0fe0 2845 the unit is omitted, the value is interpreted in microseconds. See
f75ede1d 2846 :option:`latency_window` and :option:`latency_percentile`.
71bfa161 2847
a881438b 2848.. option:: latency_window=time
71bfa161 2849
f80dba8d 2850 Used with :option:`latency_target` to specify the sample window that the job
f75ede1d 2851 is run at varying queue depths to test the performance. When the unit is
947e0fe0 2852 omitted, the value is interpreted in microseconds.
b4692828 2853
f80dba8d 2854.. option:: latency_percentile=float
71bfa161 2855
c60ebc45 2856 The percentage of I/Os that must fall within the criteria specified by
f80dba8d 2857 :option:`latency_target` and :option:`latency_window`. If not set, this
c60ebc45 2858 defaults to 100.0, meaning that all I/Os must be equal or below to the value
f80dba8d 2859 set by :option:`latency_target`.
71bfa161 2860
e1bcd541
SL
2861.. option:: latency_run=bool
2862
2863 Used with :option:`latency_target`. If false (default), fio will find
2864 the highest queue depth that meets :option:`latency_target` and exit. If
2865 true, fio will continue running and try to meet :option:`latency_target`
2866 by adjusting queue depth.
2867
f7cf63bf 2868.. option:: max_latency=time[,time][,time]
71bfa161 2869
f75ede1d 2870 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
947e0fe0 2871 maximum latency. When the unit is omitted, the value is interpreted in
f7cf63bf
VR
2872 microseconds. Comma-separated values may be specified for reads, writes,
2873 and trims as described in :option:`blocksize`.
71bfa161 2874
f80dba8d 2875.. option:: rate_cycle=int
71bfa161 2876
f80dba8d 2877 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
a47b697c 2878 of milliseconds. Defaults to 1000.
71bfa161 2879
71bfa161 2880
f80dba8d
MT
2881I/O replay
2882~~~~~~~~~~
71bfa161 2883
f80dba8d 2884.. option:: write_iolog=str
c2b1e753 2885
f80dba8d
MT
2886 Write the issued I/O patterns to the specified file. See
2887 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2888 iologs will be interspersed and the file may be corrupt.
c2b1e753 2889
f80dba8d 2890.. option:: read_iolog=str
71bfa161 2891
22413915 2892 Open an iolog with the specified filename and replay the I/O patterns it
f80dba8d
MT
2893 contains. This can be used to store a workload and replay it sometime
2894 later. The iolog given may also be a blktrace binary file, which allows fio
2895 to replay a workload captured by :command:`blktrace`. See
2896 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2897 replay, the file needs to be turned into a blkparse binary data file first
2898 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
78439a18
JA
2899 You can specify a number of files by separating the names with a ':'
2900 character. See the :option:`filename` option for information on how to
3b803fe1 2901 escape ':' characters within the file names. These files will
78439a18 2902 be sequentially assigned to job clones created by :option:`numjobs`.
d19c04d1 2903 '-' is a reserved name, meaning read from stdin, notably if
2904 :option:`filename` is set to '-' which means stdin as well, then
2905 this flag can't be set to '-'.
71bfa161 2906
77be374d
AK
2907.. option:: read_iolog_chunked=bool
2908
2909 Determines how iolog is read. If false(default) entire :option:`read_iolog`
2910 will be read at once. If selected true, input from iolog will be read
2911 gradually. Useful when iolog is very large, or it is generated.
2912
b9921d1a
DZ
2913.. option:: merge_blktrace_file=str
2914
2915 When specified, rather than replaying the logs passed to :option:`read_iolog`,
2916 the logs go through a merge phase which aggregates them into a single
2917 blktrace. The resulting file is then passed on as the :option:`read_iolog`
2918 parameter. The intention here is to make the order of events consistent.
2919 This limits the influence of the scheduler compared to replaying multiple
2920 blktraces via concurrent jobs.
2921
87a48ada
DZ
2922.. option:: merge_blktrace_scalars=float_list
2923
2924 This is a percentage based option that is index paired with the list of
2925 files passed to :option:`read_iolog`. When merging is performed, scale
2926 the time of each event by the corresponding amount. For example,
2927 ``--merge_blktrace_scalars="50:100"`` runs the first trace in halftime
2928 and the second trace in realtime. This knob is separately tunable from
2929 :option:`replay_time_scale` which scales the trace during runtime and
2930 does not change the output of the merge unlike this option.
2931
55bfd8c8
DZ
2932.. option:: merge_blktrace_iters=float_list
2933
2934 This is a whole number option that is index paired with the list of files
2935 passed to :option:`read_iolog`. When merging is performed, run each trace
2936 for the specified number of iterations. For example,
2937 ``--merge_blktrace_iters="2:1"`` runs the first trace for two iterations
2938 and the second trace for one iteration.
2939
589e88b7 2940.. option:: replay_no_stall=bool
71bfa161 2941
f80dba8d 2942 When replaying I/O with :option:`read_iolog` the default behavior is to
22413915 2943 attempt to respect the timestamps within the log and replay them with the
f80dba8d
MT
2944 appropriate delay between IOPS. By setting this variable fio will not
2945 respect the timestamps and attempt to replay them as fast as possible while
2946 still respecting ordering. The result is the same I/O pattern to a given
2947 device, but different timings.
71bfa161 2948
6dd7fa77
JA
2949.. option:: replay_time_scale=int
2950
2951 When replaying I/O with :option:`read_iolog`, fio will honor the
2952 original timing in the trace. With this option, it's possible to scale
2953 the time. It's a percentage option, if set to 50 it means run at 50%
2954 the original IO rate in the trace. If set to 200, run at twice the
2955 original IO rate. Defaults to 100.
2956
f80dba8d 2957.. option:: replay_redirect=str
b4692828 2958
f80dba8d
MT
2959 While replaying I/O patterns using :option:`read_iolog` the default behavior
2960 is to replay the IOPS onto the major/minor device that each IOP was recorded
2961 from. This is sometimes undesirable because on a different machine those
2962 major/minor numbers can map to a different device. Changing hardware on the
2963 same system can also result in a different major/minor mapping.
730bd7d9 2964 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
f80dba8d 2965 device regardless of the device it was recorded
9207a0cb 2966 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
f80dba8d
MT
2967 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2968 multiple devices will be replayed onto a single device, if the trace
2969 contains multiple devices. If you want multiple devices to be replayed
2970 concurrently to multiple redirected devices you must blkparse your trace
2971 into separate traces and replay them with independent fio invocations.
2972 Unfortunately this also breaks the strict time ordering between multiple
2973 device accesses.
71bfa161 2974
f80dba8d 2975.. option:: replay_align=int
74929ac2 2976
350a535d
DZ
2977 Force alignment of the byte offsets in a trace to this value. The value
2978 must be a power of 2.
3c54bc46 2979
f80dba8d 2980.. option:: replay_scale=int
3c54bc46 2981
350a535d
DZ
2982 Scale byte offsets down by this factor when replaying traces. Should most
2983 likely use :option:`replay_align` as well.
3c54bc46 2984
38f68906
JA
2985.. option:: replay_skip=str
2986
2987 Sometimes it's useful to skip certain IO types in a replay trace.
2988 This could be, for instance, eliminating the writes in the trace.
2989 Or not replaying the trims/discards, if you are redirecting to
2990 a device that doesn't support them. This option takes a comma
2991 separated list of read, write, trim, sync.
2992
3c54bc46 2993
f80dba8d
MT
2994Threads, processes and job synchronization
2995~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3c54bc46 2996
f80dba8d 2997.. option:: thread
3c54bc46 2998
730bd7d9
SW
2999 Fio defaults to creating jobs by using fork, however if this option is
3000 given, fio will create jobs by using POSIX Threads' function
3001 :manpage:`pthread_create(3)` to create threads instead.
71bfa161 3002
f80dba8d 3003.. option:: wait_for=str
74929ac2 3004
730bd7d9
SW
3005 If set, the current job won't be started until all workers of the specified
3006 waitee job are done.
74929ac2 3007
f80dba8d
MT
3008 ``wait_for`` operates on the job name basis, so there are a few
3009 limitations. First, the waitee must be defined prior to the waiter job
3010 (meaning no forward references). Second, if a job is being referenced as a
3011 waitee, it must have a unique name (no duplicate waitees).
74929ac2 3012
f80dba8d 3013.. option:: nice=int
892a6ffc 3014
f80dba8d 3015 Run the job with the given nice value. See man :manpage:`nice(2)`.
892a6ffc 3016
f80dba8d
MT
3017 On Windows, values less than -15 set the process class to "High"; -1 through
3018 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
3019 priority class.
74929ac2 3020
f80dba8d 3021.. option:: prio=int
71bfa161 3022
f80dba8d
MT
3023 Set the I/O priority value of this job. Linux limits us to a positive value
3024 between 0 and 7, with 0 being the highest. See man
3025 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
b2a432bf 3026 systems since meaning of priority may differ. For per-command priority
12f9d54a
DLM
3027 setting, see I/O engine specific :option:`cmdprio_percentage` and
3028 :option:`cmdprio` options.
71bfa161 3029
f80dba8d 3030.. option:: prioclass=int
d59aa780 3031
b2a432bf 3032 Set the I/O priority class. See man :manpage:`ionice(1)`. For per-command
12f9d54a
DLM
3033 priority setting, see I/O engine specific :option:`cmdprio_percentage`
3034 and :option:`cmdprio_class` options.
d59aa780 3035
f80dba8d 3036.. option:: cpus_allowed=str
6d500c2e 3037
730bd7d9 3038 Controls the same options as :option:`cpumask`, but accepts a textual
b570e037
SW
3039 specification of the permitted CPUs instead and CPUs are indexed from 0. So
3040 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
3041 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
3042 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
3043
3044 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
3045 processor group will be used and affinity settings are inherited from the
3046 system. An fio build configured to target Windows 7 makes options that set
3047 CPUs processor group aware and values will set both the processor group
3048 and a CPU from within that group. For example, on a system where processor
3049 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
3050 values between 0 and 39 will bind CPUs from processor group 0 and
3051 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
3052 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
3053 single ``cpus_allowed`` option must be from the same processor group. For
3054 Windows fio builds not built for Windows 7, CPUs will only be selected from
3055 (and be relative to) whatever processor group fio happens to be running in
3056 and CPUs from other processor groups cannot be used.
6d500c2e 3057
f80dba8d 3058.. option:: cpus_allowed_policy=str
6d500c2e 3059
f80dba8d 3060 Set the policy of how fio distributes the CPUs specified by
730bd7d9 3061 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
6d500c2e 3062
f80dba8d
MT
3063 **shared**
3064 All jobs will share the CPU set specified.
3065 **split**
3066 Each job will get a unique CPU from the CPU set.
6d500c2e 3067
22413915 3068 **shared** is the default behavior, if the option isn't specified. If
b21fc93f 3069 **split** is specified, then fio will assign one cpu per job. If not
f80dba8d
MT
3070 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
3071 in the set.
6d500c2e 3072
b570e037
SW
3073.. option:: cpumask=int
3074
3075 Set the CPU affinity of this job. The parameter given is a bit mask of
3076 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
3077 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
3078 :manpage:`sched_setaffinity(2)`. This may not work on all supported
3079 operating systems or kernel versions. This option doesn't work well for a
3080 higher CPU count than what you can store in an integer mask, so it can only
3081 control cpus 1-32. For boxes with larger CPU counts, use
3082 :option:`cpus_allowed`.
3083
f80dba8d 3084.. option:: numa_cpu_nodes=str
6d500c2e 3085
f80dba8d
MT
3086 Set this job running on specified NUMA nodes' CPUs. The arguments allow
3087 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
ac8ca2af 3088 NUMA options support, fio must be built on a system with libnuma-dev(el)
f80dba8d 3089 installed.
61b9861d 3090
f80dba8d 3091.. option:: numa_mem_policy=str
61b9861d 3092
f80dba8d
MT
3093 Set this job's memory policy and corresponding NUMA nodes. Format of the
3094 arguments::
5c94b008 3095
f80dba8d 3096 <mode>[:<nodelist>]
ce35b1ec 3097
804c0839 3098 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
730bd7d9
SW
3099 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
3100 policies, no node needs to be specified. For ``prefer``, only one node is
3101 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
3102 follows: a comma delimited list of numbers, A-B ranges, or `all`.
71bfa161 3103
f80dba8d 3104.. option:: cgroup=str
390b1537 3105
f80dba8d
MT
3106 Add job to this control group. If it doesn't exist, it will be created. The
3107 system must have a mounted cgroup blkio mount point for this to work. If
3108 your system doesn't have it mounted, you can do so with::
5af1c6f3 3109
f80dba8d 3110 # mount -t cgroup -o blkio none /cgroup
5af1c6f3 3111
f80dba8d 3112.. option:: cgroup_weight=int
5af1c6f3 3113
f80dba8d
MT
3114 Set the weight of the cgroup to this value. See the documentation that comes
3115 with the kernel, allowed values are in the range of 100..1000.
a086c257 3116
f80dba8d 3117.. option:: cgroup_nodelete=bool
8c07860d 3118
f80dba8d
MT
3119 Normally fio will delete the cgroups it has created after the job
3120 completion. To override this behavior and to leave cgroups around after the
3121 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
3122 to inspect various cgroup files after job completion. Default: false.
8c07860d 3123
f80dba8d 3124.. option:: flow_id=int
8c07860d 3125
f80dba8d
MT
3126 The ID of the flow. If not specified, it defaults to being a global
3127 flow. See :option:`flow`.
1907dbc6 3128
f80dba8d 3129.. option:: flow=int
71bfa161 3130
f80dba8d
MT
3131 Weight in token-based flow control. If this value is used, then there is a
3132 'flow counter' which is used to regulate the proportion of activity between
3133 two or more jobs. Fio attempts to keep this flow counter near zero. The
3134 ``flow`` parameter stands for how much should be added or subtracted to the
3135 flow counter on each iteration of the main I/O loop. That is, if one job has
3136 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
3137 ratio in how much one runs vs the other.
71bfa161 3138
f80dba8d 3139.. option:: flow_sleep=int
82407585 3140
d4e74fda
DB
3141 The period of time, in microseconds, to wait after the flow counter
3142 has exceeded its proportion before retrying operations.
82407585 3143
f80dba8d 3144.. option:: stonewall, wait_for_previous
82407585 3145
f80dba8d
MT
3146 Wait for preceding jobs in the job file to exit, before starting this
3147 one. Can be used to insert serialization points in the job file. A stone
3148 wall also implies starting a new reporting group, see
3149 :option:`group_reporting`.
3150
3151.. option:: exitall
3152
64402a8a
HW
3153 By default, fio will continue running all other jobs when one job finishes.
3154 Sometimes this is not the desired action. Setting ``exitall`` will instead
3155 make fio terminate all jobs in the same group, as soon as one job of that
3156 group finishes.
3157
3158.. option:: exit_what
3159
3160 By default, fio will continue running all other jobs when one job finishes.
3161 Sometimes this is not the desired action. Setting ``exit_all`` will
3162 instead make fio terminate all jobs in the same group. The option
3163 ``exit_what`` allows to control which jobs get terminated when ``exitall`` is
3164 enabled. The default is ``group`` and does not change the behaviour of
3165 ``exitall``. The setting ``all`` terminates all jobs. The setting ``stonewall``
3166 terminates all currently running jobs across all groups and continues execution
3167 with the next stonewalled group.
f80dba8d
MT
3168
3169.. option:: exec_prerun=str
3170
3171 Before running this job, issue the command specified through
3172 :manpage:`system(3)`. Output is redirected in a file called
3173 :file:`jobname.prerun.txt`.
3174
3175.. option:: exec_postrun=str
3176
3177 After the job completes, issue the command specified though
3178 :manpage:`system(3)`. Output is redirected in a file called
3179 :file:`jobname.postrun.txt`.
3180
3181.. option:: uid=int
3182
3183 Instead of running as the invoking user, set the user ID to this value
3184 before the thread/process does any work.
3185
3186.. option:: gid=int
3187
3188 Set group ID, see :option:`uid`.
3189
3190
3191Verification
3192~~~~~~~~~~~~
3193
3194.. option:: verify_only
3195
3196 Do not perform specified workload, only verify data still matches previous
3197 invocation of this workload. This option allows one to check data multiple
3198 times at a later date without overwriting it. This option makes sense only
3199 for workloads that write data, and does not support workloads with the
3200 :option:`time_based` option set.
3201
3202.. option:: do_verify=bool
3203
3204 Run the verify phase after a write phase. Only valid if :option:`verify` is
3205 set. Default: true.
3206
3207.. option:: verify=str
3208
3209 If writing to a file, fio can verify the file contents after each iteration
3210 of the job. Each verification method also implies verification of special
3211 header, which is written to the beginning of each block. This header also
3212 includes meta information, like offset of the block, block number, timestamp
3213 when block was written, etc. :option:`verify` can be combined with
3214 :option:`verify_pattern` option. The allowed values are:
3215
3216 **md5**
3217 Use an md5 sum of the data area and store it in the header of
3218 each block.
3219
3220 **crc64**
3221 Use an experimental crc64 sum of the data area and store it in the
3222 header of each block.
3223
3224 **crc32c**
a5896300
SW
3225 Use a crc32c sum of the data area and store it in the header of
3226 each block. This will automatically use hardware acceleration
3227 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
3228 fall back to software crc32c if none is found. Generally the
804c0839 3229 fastest checksum fio supports when hardware accelerated.
f80dba8d
MT
3230
3231 **crc32c-intel**
a5896300 3232 Synonym for crc32c.
f80dba8d
MT
3233
3234 **crc32**
3235 Use a crc32 sum of the data area and store it in the header of each
3236 block.
3237
3238 **crc16**
3239 Use a crc16 sum of the data area and store it in the header of each
3240 block.
3241
3242 **crc7**
3243 Use a crc7 sum of the data area and store it in the header of each
3244 block.
3245
3246 **xxhash**
3247 Use xxhash as the checksum function. Generally the fastest software
3248 checksum that fio supports.
3249
3250 **sha512**
3251 Use sha512 as the checksum function.
3252
3253 **sha256**
3254 Use sha256 as the checksum function.
3255
3256 **sha1**
3257 Use optimized sha1 as the checksum function.
82407585 3258
ae3a5acc
JA
3259 **sha3-224**
3260 Use optimized sha3-224 as the checksum function.
3261
3262 **sha3-256**
3263 Use optimized sha3-256 as the checksum function.
3264
3265 **sha3-384**
3266 Use optimized sha3-384 as the checksum function.
3267
3268 **sha3-512**
3269 Use optimized sha3-512 as the checksum function.
3270
f80dba8d
MT
3271 **meta**
3272 This option is deprecated, since now meta information is included in
3273 generic verification header and meta verification happens by
3274 default. For detailed information see the description of the
3275 :option:`verify` setting. This option is kept because of
3276 compatibility's sake with old configurations. Do not use it.
3277
3278 **pattern**
3279 Verify a strict pattern. Normally fio includes a header with some
3280 basic information and checksumming, but if this option is set, only
3281 the specific pattern set with :option:`verify_pattern` is verified.
3282
3283 **null**
3284 Only pretend to verify. Useful for testing internals with
9207a0cb 3285 :option:`ioengine`\=null, not for much else.
f80dba8d
MT
3286
3287 This option can be used for repeated burn-in tests of a system to make sure
3288 that the written data is also correctly read back. If the data direction
3289 given is a read or random read, fio will assume that it should verify a
3290 previously written file. If the data direction includes any form of write,
3291 the verify will be of the newly written data.
3292
47e6a6e5
SW
3293 To avoid false verification errors, do not use the norandommap option when
3294 verifying data with async I/O engines and I/O depths > 1. Or use the
3295 norandommap and the lfsr random generator together to avoid writing to the
3296 same offset with muliple outstanding I/Os.
3297
f80dba8d
MT
3298.. option:: verify_offset=int
3299
3300 Swap the verification header with data somewhere else in the block before
3301 writing. It is swapped back before verifying.
3302
3303.. option:: verify_interval=int
3304
3305 Write the verification header at a finer granularity than the
3306 :option:`blocksize`. It will be written for chunks the size of
3307 ``verify_interval``. :option:`blocksize` should divide this evenly.
3308
3309.. option:: verify_pattern=str
3310
3311 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
3312 filling with totally random bytes, but sometimes it's interesting to fill
3313 with a known pattern for I/O verification purposes. Depending on the width
730bd7d9 3314 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
f80dba8d
MT
3315 be either a decimal or a hex number). The ``verify_pattern`` if larger than
3316 a 32-bit quantity has to be a hex number that starts with either "0x" or
3317 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
3318 format, which means that for each block offset will be written and then
3319 verified back, e.g.::
61b9861d
RP
3320
3321 verify_pattern=%o
3322
f80dba8d
MT
3323 Or use combination of everything::
3324
61b9861d 3325 verify_pattern=0xff%o"abcd"-12
e28218f3 3326
f80dba8d
MT
3327.. option:: verify_fatal=bool
3328
3329 Normally fio will keep checking the entire contents before quitting on a
3330 block verification failure. If this option is set, fio will exit the job on
3331 the first observed failure. Default: false.
3332
3333.. option:: verify_dump=bool
3334
3335 If set, dump the contents of both the original data block and the data block
3336 we read off disk to files. This allows later analysis to inspect just what
3337 kind of data corruption occurred. Off by default.
3338
3339.. option:: verify_async=int
3340
3341 Fio will normally verify I/O inline from the submitting thread. This option
3342 takes an integer describing how many async offload threads to create for I/O
3343 verification instead, causing fio to offload the duty of verifying I/O
3344 contents to one or more separate threads. If using this offload option, even
3345 sync I/O engines can benefit from using an :option:`iodepth` setting higher
3346 than 1, as it allows them to have I/O in flight while verifies are running.
d7e6ea1c 3347 Defaults to 0 async threads, i.e. verification is not asynchronous.
f80dba8d
MT
3348
3349.. option:: verify_async_cpus=str
3350
3351 Tell fio to set the given CPU affinity on the async I/O verification
3352 threads. See :option:`cpus_allowed` for the format used.
3353
3354.. option:: verify_backlog=int
3355
3356 Fio will normally verify the written contents of a job that utilizes verify
3357 once that job has completed. In other words, everything is written then
3358 everything is read back and verified. You may want to verify continually
3359 instead for a variety of reasons. Fio stores the meta data associated with
3360 an I/O block in memory, so for large verify workloads, quite a bit of memory
3361 would be used up holding this meta data. If this option is enabled, fio will
3362 write only N blocks before verifying these blocks.
3363
3364.. option:: verify_backlog_batch=int
3365
3366 Control how many blocks fio will verify if :option:`verify_backlog` is
3367 set. If not set, will default to the value of :option:`verify_backlog`
3368 (meaning the entire queue is read back and verified). If
3369 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
3370 blocks will be verified, if ``verify_backlog_batch`` is larger than
3371 :option:`verify_backlog`, some blocks will be verified more than once.
3372
3373.. option:: verify_state_save=bool
3374
3375 When a job exits during the write phase of a verify workload, save its
3376 current state. This allows fio to replay up until that point, if the verify
3377 state is loaded for the verify read phase. The format of the filename is,
3378 roughly::
3379
f50fbdda 3380 <type>-<jobname>-<jobindex>-verify.state.
f80dba8d
MT
3381
3382 <type> is "local" for a local run, "sock" for a client/server socket
3383 connection, and "ip" (192.168.0.1, for instance) for a networked
d7e6ea1c 3384 client/server connection. Defaults to true.
f80dba8d
MT
3385
3386.. option:: verify_state_load=bool
3387
3388 If a verify termination trigger was used, fio stores the current write state
3389 of each thread. This can be used at verification time so that fio knows how
3390 far it should verify. Without this information, fio will run a full
a47b697c
SW
3391 verification pass, according to the settings in the job file used. Default
3392 false.
f80dba8d
MT
3393
3394.. option:: trim_percentage=int
3395
3396 Number of verify blocks to discard/trim.
3397
3398.. option:: trim_verify_zero=bool
3399
22413915 3400 Verify that trim/discarded blocks are returned as zeros.
f80dba8d
MT
3401
3402.. option:: trim_backlog=int
3403
5cfd1e9a 3404 Trim after this number of blocks are written.
f80dba8d
MT
3405
3406.. option:: trim_backlog_batch=int
3407
3408 Trim this number of I/O blocks.
3409
3410.. option:: experimental_verify=bool
3411
3412 Enable experimental verification.
3413
f80dba8d
MT
3414Steady state
3415~~~~~~~~~~~~
3416
3417.. option:: steadystate=str:float, ss=str:float
3418
3419 Define the criterion and limit for assessing steady state performance. The
3420 first parameter designates the criterion whereas the second parameter sets
3421 the threshold. When the criterion falls below the threshold for the
3422 specified duration, the job will stop. For example, `iops_slope:0.1%` will
3423 direct fio to terminate the job when the least squares regression slope
3424 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
3425 this will apply to all jobs in the group. Below is the list of available
3426 steady state assessment criteria. All assessments are carried out using only
3427 data from the rolling collection window. Threshold limits can be expressed
3428 as a fixed value or as a percentage of the mean in the collection window.
3429
1cb049d9
VF
3430 When using this feature, most jobs should include the :option:`time_based`
3431 and :option:`runtime` options or the :option:`loops` option so that fio does not
3432 stop running after it has covered the full size of the specified file(s) or device(s).
3433
f80dba8d
MT
3434 **iops**
3435 Collect IOPS data. Stop the job if all individual IOPS measurements
3436 are within the specified limit of the mean IOPS (e.g., ``iops:2``
3437 means that all individual IOPS values must be within 2 of the mean,
3438 whereas ``iops:0.2%`` means that all individual IOPS values must be
3439 within 0.2% of the mean IOPS to terminate the job).
3440
3441 **iops_slope**
3442 Collect IOPS data and calculate the least squares regression
3443 slope. Stop the job if the slope falls below the specified limit.
3444
3445 **bw**
3446 Collect bandwidth data. Stop the job if all individual bandwidth
3447 measurements are within the specified limit of the mean bandwidth.
3448
3449 **bw_slope**
3450 Collect bandwidth data and calculate the least squares regression
3451 slope. Stop the job if the slope falls below the specified limit.
3452
3453.. option:: steadystate_duration=time, ss_dur=time
3454
3455 A rolling window of this duration will be used to judge whether steady state
3456 has been reached. Data will be collected once per second. The default is 0
f75ede1d 3457 which disables steady state detection. When the unit is omitted, the
947e0fe0 3458 value is interpreted in seconds.
f80dba8d
MT
3459
3460.. option:: steadystate_ramp_time=time, ss_ramp=time
3461
3462 Allow the job to run for the specified duration before beginning data
3463 collection for checking the steady state job termination criterion. The
947e0fe0 3464 default is 0. When the unit is omitted, the value is interpreted in seconds.
f80dba8d
MT
3465
3466
3467Measurements and reporting
3468~~~~~~~~~~~~~~~~~~~~~~~~~~
3469
3470.. option:: per_job_logs=bool
3471
3472 If set, this generates bw/clat/iops log with per file private filenames. If
3473 not set, jobs with identical names will share the log filename. Default:
3474 true.
3475
3476.. option:: group_reporting
3477
3478 It may sometimes be interesting to display statistics for groups of jobs as
3479 a whole instead of for each individual job. This is especially true if
3480 :option:`numjobs` is used; looking at individual thread/process output
3481 quickly becomes unwieldy. To see the final report per-group instead of
3482 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
3483 same reporting group, unless if separated by a :option:`stonewall`, or by
3484 using :option:`new_group`.
3485
3486.. option:: new_group
3487
3488 Start a new reporting group. See: :option:`group_reporting`. If not given,
3489 all jobs in a file will be part of the same reporting group, unless
3490 separated by a :option:`stonewall`.
3491
589e88b7 3492.. option:: stats=bool
8243be59
JA
3493
3494 By default, fio collects and shows final output results for all jobs
3495 that run. If this option is set to 0, then fio will ignore it in
3496 the final stat output.
3497
f80dba8d
MT
3498.. option:: write_bw_log=str
3499
3500 If given, write a bandwidth log for this job. Can be used to store data of
074f0817 3501 the bandwidth of the jobs in their lifetime.
f80dba8d 3502
074f0817
SW
3503 If no str argument is given, the default filename of
3504 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
3505 will still append the type of log. So if one specifies::
3506
3507 write_bw_log=foo
f80dba8d 3508
074f0817
SW
3509 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
3510 of the job (`1..N`, where `N` is the number of jobs). If
3511 :option:`per_job_logs` is false, then the filename will not include the
3512 `.x` job index.
e3cedca7 3513
074f0817
SW
3514 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
3515 text files into nice graphs. See `Log File Formats`_ for how data is
3516 structured within the file.
3517
3518.. option:: write_lat_log=str
e3cedca7 3519
074f0817 3520 Same as :option:`write_bw_log`, except this option creates I/O
77b7e675
SW
3521 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
3522 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
074f0817
SW
3523 latency files instead. See :option:`write_bw_log` for details about
3524 the filename format and `Log File Formats`_ for how data is structured
3525 within the files.
be4ecfdf 3526
f80dba8d 3527.. option:: write_hist_log=str
06842027 3528
074f0817 3529 Same as :option:`write_bw_log` but writes an I/O completion latency
77b7e675 3530 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
074f0817
SW
3531 file will be empty unless :option:`log_hist_msec` has also been set.
3532 See :option:`write_bw_log` for details about the filename format and
3533 `Log File Formats`_ for how data is structured within the file.
06842027 3534
f80dba8d 3535.. option:: write_iops_log=str
06842027 3536
074f0817 3537 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
15417073
SW
3538 :file:`name_iops.x.log`) instead. Because fio defaults to individual
3539 I/O logging, the value entry in the IOPS log will be 1 unless windowed
3540 logging (see :option:`log_avg_msec`) has been enabled. See
3541 :option:`write_bw_log` for details about the filename format and `Log
3542 File Formats`_ for how data is structured within the file.
06842027 3543
0a852a50
DLM
3544.. option:: log_entries=int
3545
3546 By default, fio will log an entry in the iops, latency, or bw log for
3547 every I/O that completes. The initial number of I/O log entries is 1024.
3548 When the log entries are all used, new log entries are dynamically
3549 allocated. This dynamic log entry allocation may negatively impact
3550 time-related statistics such as I/O tail latencies (e.g. 99.9th percentile
3551 completion latency). This option allows specifying a larger initial
3552 number of log entries to avoid run-time allocations of new log entries,
3553 resulting in more precise time-related I/O statistics.
3554 Also see :option:`log_avg_msec`. Defaults to 1024.
3555
f80dba8d 3556.. option:: log_avg_msec=int
06842027 3557
f80dba8d
MT
3558 By default, fio will log an entry in the iops, latency, or bw log for every
3559 I/O that completes. When writing to the disk log, that can quickly grow to a
3560 very large size. Setting this option makes fio average the each log entry
3561 over the specified period of time, reducing the resolution of the log. See
3562 :option:`log_max_value` as well. Defaults to 0, logging all entries.
6fc82095 3563 Also see `Log File Formats`_.
06842027 3564
f80dba8d 3565.. option:: log_hist_msec=int
06842027 3566
f80dba8d
MT
3567 Same as :option:`log_avg_msec`, but logs entries for completion latency
3568 histograms. Computing latency percentiles from averages of intervals using
c60ebc45 3569 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
f80dba8d
MT
3570 histogram entries over the specified period of time, reducing log sizes for
3571 high IOPS devices while retaining percentile accuracy. See
074f0817
SW
3572 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
3573 Defaults to 0, meaning histogram logging is disabled.
06842027 3574
f80dba8d 3575.. option:: log_hist_coarseness=int
06842027 3576
f80dba8d
MT
3577 Integer ranging from 0 to 6, defining the coarseness of the resolution of
3578 the histogram logs enabled with :option:`log_hist_msec`. For each increment
3579 in coarseness, fio outputs half as many bins. Defaults to 0, for which
074f0817
SW
3580 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
3581 and `Log File Formats`_.
8b28bd41 3582
f80dba8d 3583.. option:: log_max_value=bool
66c098b8 3584
f80dba8d
MT
3585 If :option:`log_avg_msec` is set, fio logs the average over that window. If
3586 you instead want to log the maximum value, set this option to 1. Defaults to
3587 0, meaning that averaged values are logged.
a696fa2a 3588
589e88b7 3589.. option:: log_offset=bool
a696fa2a 3590
f80dba8d 3591 If this is set, the iolog options will include the byte offset for the I/O
5a83478f
SW
3592 entry as well as the other data values. Defaults to 0 meaning that
3593 offsets are not present in logs. Also see `Log File Formats`_.
71bfa161 3594
f80dba8d 3595.. option:: log_compression=int
7de87099 3596
f80dba8d
MT
3597 If this is set, fio will compress the I/O logs as it goes, to keep the
3598 memory footprint lower. When a log reaches the specified size, that chunk is
3599 removed and compressed in the background. Given that I/O logs are fairly
3600 highly compressible, this yields a nice memory savings for longer runs. The
3601 downside is that the compression will consume some background CPU cycles, so
3602 it may impact the run. This, however, is also true if the logging ends up
3603 consuming most of the system memory. So pick your poison. The I/O logs are
3604 saved normally at the end of a run, by decompressing the chunks and storing
3605 them in the specified log file. This feature depends on the availability of
3606 zlib.
e0b0d892 3607
f80dba8d 3608.. option:: log_compression_cpus=str
e0b0d892 3609
f80dba8d
MT
3610 Define the set of CPUs that are allowed to handle online log compression for
3611 the I/O jobs. This can provide better isolation between performance
0cf90a62
SW
3612 sensitive jobs, and background compression work. See
3613 :option:`cpus_allowed` for the format used.
9e684a49 3614
f80dba8d 3615.. option:: log_store_compressed=bool
9e684a49 3616
f80dba8d
MT
3617 If set, fio will store the log files in a compressed format. They can be
3618 decompressed with fio, using the :option:`--inflate-log` command line
3619 parameter. The files will be stored with a :file:`.fz` suffix.
9e684a49 3620
f80dba8d 3621.. option:: log_unix_epoch=bool
9e684a49 3622
f80dba8d
MT
3623 If set, fio will log Unix timestamps to the log files produced by enabling
3624 write_type_log for each log type, instead of the default zero-based
3625 timestamps.
3626
3627.. option:: block_error_percentiles=bool
3628
3629 If set, record errors in trim block-sized units from writes and trims and
3630 output a histogram of how many trims it took to get to errors, and what kind
3631 of error was encountered.
3632
3633.. option:: bwavgtime=int
3634
3635 Average the calculated bandwidth over the given time. Value is specified in
3636 milliseconds. If the job also does bandwidth logging through
3637 :option:`write_bw_log`, then the minimum of this option and
3638 :option:`log_avg_msec` will be used. Default: 500ms.
3639
3640.. option:: iopsavgtime=int
3641
3642 Average the calculated IOPS over the given time. Value is specified in
3643 milliseconds. If the job also does IOPS logging through
3644 :option:`write_iops_log`, then the minimum of this option and
3645 :option:`log_avg_msec` will be used. Default: 500ms.
3646
3647.. option:: disk_util=bool
3648
3649 Generate disk utilization statistics, if the platform supports it.
3650 Default: true.
3651
3652.. option:: disable_lat=bool
3653
3654 Disable measurements of total latency numbers. Useful only for cutting back
3655 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
3656 performance at really high IOPS rates. Note that to really get rid of a
3657 large amount of these calls, this option must be used with
f75ede1d 3658 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
f80dba8d
MT
3659
3660.. option:: disable_clat=bool
3661
3662 Disable measurements of completion latency numbers. See
3663 :option:`disable_lat`.
3664
3665.. option:: disable_slat=bool
3666
3667 Disable measurements of submission latency numbers. See
f50fbdda 3668 :option:`disable_lat`.
f80dba8d 3669
f75ede1d 3670.. option:: disable_bw_measurement=bool, disable_bw=bool
f80dba8d
MT
3671
3672 Disable measurements of throughput/bandwidth numbers. See
3673 :option:`disable_lat`.
3674
dd39b9ce
VF
3675.. option:: slat_percentiles=bool
3676
3677 Report submission latency percentiles. Submission latency is not recorded
3678 for synchronous ioengines.
3679
f80dba8d
MT
3680.. option:: clat_percentiles=bool
3681
dd39b9ce 3682 Report completion latency percentiles.
b599759b
JA
3683
3684.. option:: lat_percentiles=bool
3685
dd39b9ce
VF
3686 Report total latency percentiles. Total latency is the sum of submission
3687 latency and completion latency.
f80dba8d
MT
3688
3689.. option:: percentile_list=float_list
3690
dd39b9ce
VF
3691 Overwrite the default list of percentiles for latencies and the block error
3692 histogram. Each number is a floating point number in the range (0,100], and
3693 the maximum length of the list is 20. Use ``:`` to separate the numbers. For
c32ba107 3694 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
dd39b9ce
VF
3695 latency durations below which 99.5% and 99.9% of the observed latencies fell,
3696 respectively.
f80dba8d 3697
e883cb35
JF
3698.. option:: significant_figures=int
3699
c32ba107
JA
3700 If using :option:`--output-format` of `normal`, set the significant
3701 figures to this value. Higher values will yield more precise IOPS and
3702 throughput units, while lower values will round. Requires a minimum
3703 value of 1 and a maximum value of 10. Defaults to 4.
e883cb35 3704
f80dba8d
MT
3705
3706Error handling
3707~~~~~~~~~~~~~~
3708
3709.. option:: exitall_on_error
3710
3711 When one job finishes in error, terminate the rest. The default is to wait
3712 for each job to finish.
3713
3714.. option:: continue_on_error=str
3715
3716 Normally fio will exit the job on the first observed failure. If this option
3717 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
3718 EILSEQ) until the runtime is exceeded or the I/O size specified is
3719 completed. If this option is used, there are two more stats that are
3720 appended, the total error count and the first error. The error field given
3721 in the stats is the first error that was hit during the run.
3722
3723 The allowed values are:
3724
3725 **none**
3726 Exit on any I/O or verify errors.
3727
3728 **read**
3729 Continue on read errors, exit on all others.
3730
3731 **write**
3732 Continue on write errors, exit on all others.
3733
3734 **io**
3735 Continue on any I/O error, exit on all others.
3736
3737 **verify**
3738 Continue on verify errors, exit on all others.
3739
3740 **all**
3741 Continue on all errors.
3742
3743 **0**
3744 Backward-compatible alias for 'none'.
3745
3746 **1**
3747 Backward-compatible alias for 'all'.
3748
3749.. option:: ignore_error=str
3750
3751 Sometimes you want to ignore some errors during test in that case you can
a35ef7cb
TK
3752 specify error list for each error type, instead of only being able to
3753 ignore the default 'non-fatal error' using :option:`continue_on_error`.
f80dba8d
MT
3754 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
3755 given error type is separated with ':'. Error may be symbol ('ENOSPC',
3756 'ENOMEM') or integer. Example::
3757
3758 ignore_error=EAGAIN,ENOSPC:122
3759
3760 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
a35ef7cb
TK
3761 WRITE. This option works by overriding :option:`continue_on_error` with
3762 the list of errors for each error type if any.
f80dba8d
MT
3763
3764.. option:: error_dump=bool
3765
3766 If set dump every error even if it is non fatal, true by default. If
3767 disabled only fatal error will be dumped.
3768
f75ede1d
SW
3769Running predefined workloads
3770----------------------------
3771
3772Fio includes predefined profiles that mimic the I/O workloads generated by
3773other tools.
3774
3775.. option:: profile=str
3776
3777 The predefined workload to run. Current profiles are:
3778
3779 **tiobench**
3780 Threaded I/O bench (tiotest/tiobench) like workload.
3781
3782 **act**
3783 Aerospike Certification Tool (ACT) like workload.
3784
3785To view a profile's additional options use :option:`--cmdhelp` after specifying
3786the profile. For example::
3787
f50fbdda 3788 $ fio --profile=act --cmdhelp
f75ede1d
SW
3789
3790Act profile options
3791~~~~~~~~~~~~~~~~~~~
3792
3793.. option:: device-names=str
3794 :noindex:
3795
3796 Devices to use.
3797
3798.. option:: load=int
3799 :noindex:
3800
3801 ACT load multiplier. Default: 1.
3802
3803.. option:: test-duration=time
3804 :noindex:
3805
947e0fe0
SW
3806 How long the entire test takes to run. When the unit is omitted, the value
3807 is given in seconds. Default: 24h.
f75ede1d
SW
3808
3809.. option:: threads-per-queue=int
3810 :noindex:
3811
f50fbdda 3812 Number of read I/O threads per device. Default: 8.
f75ede1d
SW
3813
3814.. option:: read-req-num-512-blocks=int
3815 :noindex:
3816
3817 Number of 512B blocks to read at the time. Default: 3.
381