Add support for resetting zones periodically
[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.
b034c0dd 96 *?* or *help*
f80dba8d
MT
97 Show available debug options.
98
99.. option:: --parse-only
100
25cd4b95 101 Parse options only, don't start any I/O.
f80dba8d
MT
102
103.. option:: --output=filename
104
105 Write output to file `filename`.
106
f50fbdda 107.. option:: --output-format=format
b8f7e412 108
f50fbdda 109 Set the reporting `format` to `normal`, `terse`, `json`, or `json+`. Multiple
b8f7e412
TK
110 formats can be selected, separated by a comma. `terse` is a CSV based
111 format. `json+` is like `json`, except it adds a full dump of the latency
112 buckets.
113
f80dba8d
MT
114.. option:: --bandwidth-log
115
116 Generate aggregate bandwidth logs.
117
118.. option:: --minimal
119
120 Print statistics in a terse, semicolon-delimited format.
121
122.. option:: --append-terse
123
b034c0dd
TK
124 Print statistics in selected mode AND terse, semicolon-delimited format.
125 **Deprecated**, use :option:`--output-format` instead to select multiple
126 formats.
f80dba8d 127
f50fbdda 128.. option:: --terse-version=version
f80dba8d 129
f50fbdda 130 Set terse `version` output format (default 3, or 2 or 4 or 5).
f80dba8d
MT
131
132.. option:: --version
133
b8f7e412 134 Print version information and exit.
f80dba8d
MT
135
136.. option:: --help
137
113f0e7c 138 Print a summary of the command line options and exit.
f80dba8d
MT
139
140.. option:: --cpuclock-test
141
142 Perform test and validation of internal CPU clock.
143
113f0e7c 144.. option:: --crctest=[test]
f80dba8d 145
b034c0dd
TK
146 Test the speed of the built-in checksumming functions. If no argument is
147 given, all of them are tested. Alternatively, a comma separated list can
148 be passed, in which case the given ones are tested.
f80dba8d
MT
149
150.. option:: --cmdhelp=command
151
152 Print help information for `command`. May be ``all`` for all commands.
153
154.. option:: --enghelp=[ioengine[,command]]
155
f50fbdda
TK
156 List all commands defined by `ioengine`, or print help for `command`
157 defined by `ioengine`. If no `ioengine` is given, list all
b034c0dd 158 available ioengines.
f80dba8d
MT
159
160.. option:: --showcmd=jobfile
161
b8f7e412 162 Convert `jobfile` to a set of command-line options.
f80dba8d
MT
163
164.. option:: --readonly
165
4027b2a1
VF
166 Turn on safety read-only checks, preventing writes and trims. The
167 ``--readonly`` option is an extra safety guard to prevent users from
168 accidentally starting a write or trim workload when that is not desired.
169 Fio will only modify the device under test if
170 `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite` is given. This
171 safety net can be used as an extra precaution.
f80dba8d
MT
172
173.. option:: --eta=when
174
b8f7e412 175 Specifies when real-time ETA estimate should be printed. `when` may be
db37d890
JA
176 `always`, `never` or `auto`. `auto` is the default, it prints ETA
177 when requested if the output is a TTY. `always` disregards the output
178 type, and prints ETA when requested. `never` never prints ETA.
179
180.. option:: --eta-interval=time
181
182 By default, fio requests client ETA status roughly every second. With
183 this option, the interval is configurable. Fio imposes a minimum
184 allowed time to avoid flooding the console, less than 250 msec is
185 not supported.
f80dba8d
MT
186
187.. option:: --eta-newline=time
188
947e0fe0
SW
189 Force a new line for every `time` period passed. When the unit is omitted,
190 the value is interpreted in seconds.
f80dba8d
MT
191
192.. option:: --status-interval=time
193
aa6cb459
VF
194 Force a full status dump of cumulative (from job start) values at `time`
195 intervals. This option does *not* provide per-period measurements. So
196 values such as bandwidth are running averages. When the time unit is omitted,
197 `time` is interpreted in seconds.
f80dba8d
MT
198
199.. option:: --section=name
200
b034c0dd
TK
201 Only run specified section `name` in job file. Multiple sections can be specified.
202 The ``--section`` option allows one to combine related jobs into one file.
203 E.g. one job file could define light, moderate, and heavy sections. Tell
204 fio to run only the "heavy" section by giving ``--section=heavy``
205 command line option. One can also specify the "write" operations in one
206 section and "verify" operation in another section. The ``--section`` option
207 only applies to job sections. The reserved *global* section is always
208 parsed and used.
f80dba8d
MT
209
210.. option:: --alloc-size=kb
211
b034c0dd
TK
212 Set the internal smalloc pool size to `kb` in KiB. The
213 ``--alloc-size`` switch allows one to use a larger pool size for smalloc.
214 If running large jobs with randommap enabled, fio can run out of memory.
215 Smalloc is an internal allocator for shared structures from a fixed size
216 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
f80dba8d 217
b034c0dd
TK
218 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
219 in :file:`/tmp`.
f80dba8d
MT
220
221.. option:: --warnings-fatal
222
b034c0dd
TK
223 All fio parser warnings are fatal, causing fio to exit with an
224 error.
f80dba8d
MT
225
226.. option:: --max-jobs=nr
227
f50fbdda 228 Set the maximum number of threads/processes to support to `nr`.
818322cc 229 NOTE: On Linux, it may be necessary to increase the shared-memory
71aa48eb 230 limit (:file:`/proc/sys/kernel/shmmax`) if fio runs into errors while
818322cc 231 creating jobs.
f80dba8d
MT
232
233.. option:: --server=args
234
b034c0dd
TK
235 Start a backend server, with `args` specifying what to listen to.
236 See `Client/Server`_ section.
f80dba8d
MT
237
238.. option:: --daemonize=pidfile
239
b034c0dd 240 Background a fio server, writing the pid to the given `pidfile` file.
f80dba8d
MT
241
242.. option:: --client=hostname
243
f50fbdda 244 Instead of running the jobs locally, send and run them on the given `hostname`
71aa48eb 245 or set of `hostname`\s. See `Client/Server`_ section.
f80dba8d
MT
246
247.. option:: --remote-config=file
248
f50fbdda 249 Tell fio server to load this local `file`.
f80dba8d
MT
250
251.. option:: --idle-prof=option
252
b8f7e412 253 Report CPU idleness. `option` is one of the following:
113f0e7c
SW
254
255 **calibrate**
256 Run unit work calibration only and exit.
257
258 **system**
259 Show aggregate system idleness and unit work.
260
261 **percpu**
262 As **system** but also show per CPU idleness.
f80dba8d
MT
263
264.. option:: --inflate-log=log
265
f50fbdda 266 Inflate and output compressed `log`.
f80dba8d
MT
267
268.. option:: --trigger-file=file
269
f50fbdda 270 Execute trigger command when `file` exists.
f80dba8d 271
f50fbdda 272.. option:: --trigger-timeout=time
f80dba8d 273
f50fbdda 274 Execute trigger at this `time`.
f80dba8d 275
f50fbdda 276.. option:: --trigger=command
f80dba8d 277
f50fbdda 278 Set this `command` as local trigger.
f80dba8d 279
f50fbdda 280.. option:: --trigger-remote=command
f80dba8d 281
f50fbdda 282 Set this `command` as remote trigger.
f80dba8d
MT
283
284.. option:: --aux-path=path
285
f4401bf8
SW
286 Use the directory specified by `path` for generated state files instead
287 of the current working directory.
f80dba8d
MT
288
289Any parameters following the options will be assumed to be job files, unless
290they match a job file parameter. Multiple job files can be listed and each job
291file will be regarded as a separate group. Fio will :option:`stonewall`
292execution between each group.
293
294
295Job file format
296---------------
297
298As previously described, fio accepts one or more job files describing what it is
299supposed to do. The job file format is the classic ini file, where the names
c60ebc45 300enclosed in [] brackets define the job name. You are free to use any ASCII name
f80dba8d
MT
301you want, except *global* which has special meaning. Following the job name is
302a sequence of zero or more parameters, one per line, that define the behavior of
303the job. If the first character in a line is a ';' or a '#', the entire line is
304discarded as a comment.
305
306A *global* section sets defaults for the jobs described in that file. A job may
307override a *global* section parameter, and a job file may even have several
308*global* sections if so desired. A job is only affected by a *global* section
309residing above it.
310
f50fbdda
TK
311The :option:`--cmdhelp` option also lists all options. If used with a `command`
312argument, :option:`--cmdhelp` will detail the given `command`.
f80dba8d
MT
313
314See the `examples/` directory for inspiration on how to write job files. Note
315the copyright and license requirements currently apply to `examples/` files.
316
317So let's look at a really simple job file that defines two processes, each
318randomly reading from a 128MiB file:
319
320.. code-block:: ini
321
322 ; -- start job file --
323 [global]
324 rw=randread
325 size=128m
326
327 [job1]
328
329 [job2]
330
331 ; -- end job file --
332
333As you can see, the job file sections themselves are empty as all the described
334parameters are shared. As no :option:`filename` option is given, fio makes up a
335`filename` for each of the jobs as it sees fit. On the command line, this job
336would look as follows::
337
338$ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
339
340
341Let's look at an example that has a number of processes writing randomly to
342files:
343
344.. code-block:: ini
345
346 ; -- start job file --
347 [random-writers]
348 ioengine=libaio
349 iodepth=4
350 rw=randwrite
351 bs=32k
352 direct=0
353 size=64m
354 numjobs=4
355 ; -- end job file --
356
357Here we have no *global* section, as we only have one job defined anyway. We
358want to use async I/O here, with a depth of 4 for each file. We also increased
359the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
360jobs. The result is 4 processes each randomly writing to their own 64MiB
361file. Instead of using the above job file, you could have given the parameters
362on the command line. For this case, you would specify::
363
364$ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
365
366When fio is utilized as a basis of any reasonably large test suite, it might be
367desirable to share a set of standardized settings across multiple job files.
368Instead of copy/pasting such settings, any section may pull in an external
369:file:`filename.fio` file with *include filename* directive, as in the following
370example::
371
372 ; -- start job file including.fio --
373 [global]
374 filename=/tmp/test
375 filesize=1m
376 include glob-include.fio
377
378 [test]
379 rw=randread
380 bs=4k
381 time_based=1
382 runtime=10
383 include test-include.fio
384 ; -- end job file including.fio --
385
386.. code-block:: ini
387
388 ; -- start job file glob-include.fio --
389 thread=1
390 group_reporting=1
391 ; -- end job file glob-include.fio --
392
393.. code-block:: ini
394
395 ; -- start job file test-include.fio --
396 ioengine=libaio
397 iodepth=4
398 ; -- end job file test-include.fio --
399
400Settings pulled into a section apply to that section only (except *global*
401section). Include directives may be nested in that any included file may contain
402further include directive(s). Include files may not contain [] sections.
403
404
405Environment variables
406~~~~~~~~~~~~~~~~~~~~~
407
408Fio also supports environment variable expansion in job files. Any sub-string of
409the form ``${VARNAME}`` as part of an option value (in other words, on the right
410of the '='), will be expanded to the value of the environment variable called
411`VARNAME`. If no such environment variable is defined, or `VARNAME` is the
412empty string, the empty string will be substituted.
413
414As an example, let's look at a sample fio invocation and job file::
415
416$ SIZE=64m NUMJOBS=4 fio jobfile.fio
417
418.. code-block:: ini
419
420 ; -- start job file --
421 [random-writers]
422 rw=randwrite
423 size=${SIZE}
424 numjobs=${NUMJOBS}
425 ; -- end job file --
426
427This will expand to the following equivalent job file at runtime:
428
429.. code-block:: ini
430
431 ; -- start job file --
432 [random-writers]
433 rw=randwrite
434 size=64m
435 numjobs=4
436 ; -- end job file --
437
438Fio ships with a few example job files, you can also look there for inspiration.
439
440Reserved keywords
441~~~~~~~~~~~~~~~~~
442
443Additionally, fio has a set of reserved keywords that will be replaced
444internally with the appropriate value. Those keywords are:
445
446**$pagesize**
447
448 The architecture page size of the running system.
449
450**$mb_memory**
451
452 Megabytes of total memory in the system.
453
454**$ncpus**
455
456 Number of online available CPUs.
457
458These can be used on the command line or in the job file, and will be
459automatically substituted with the current system values when the job is
460run. Simple math is also supported on these keywords, so you can perform actions
461like::
462
b034c0dd 463 size=8*$mb_memory
f80dba8d
MT
464
465and get that properly expanded to 8 times the size of memory in the machine.
466
467
468Job file parameters
469-------------------
470
471This section describes in details each parameter associated with a job. Some
472parameters take an option of a given type, such as an integer or a
473string. Anywhere a numeric value is required, an arithmetic expression may be
474used, provided it is surrounded by parentheses. Supported operators are:
475
476 - addition (+)
477 - subtraction (-)
478 - multiplication (*)
479 - division (/)
480 - modulus (%)
481 - exponentiation (^)
482
483For time values in expressions, units are microseconds by default. This is
484different than for time values not in expressions (not enclosed in
485parentheses). The following types are used:
486
487
488Parameter types
489~~~~~~~~~~~~~~~
490
491**str**
b034c0dd 492 String: A sequence of alphanumeric characters.
f80dba8d
MT
493
494**time**
008d0feb
SW
495 Integer with possible time suffix. Without a unit value is interpreted as
496 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
497 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
498 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
f80dba8d
MT
499
500.. _int:
501
502**int**
503 Integer. A whole number value, which may contain an integer prefix
504 and an integer suffix:
505
b034c0dd 506 [*integer prefix*] **number** [*integer suffix*]
f80dba8d
MT
507
508 The optional *integer prefix* specifies the number's base. The default
509 is decimal. *0x* specifies hexadecimal.
510
511 The optional *integer suffix* specifies the number's units, and includes an
512 optional unit prefix and an optional unit. For quantities of data, the
947e0fe0
SW
513 default unit is bytes. For quantities of time, the default unit is seconds
514 unless otherwise specified.
f80dba8d 515
9207a0cb 516 With :option:`kb_base`\=1000, fio follows international standards for unit
f80dba8d
MT
517 prefixes. To specify power-of-10 decimal values defined in the
518 International System of Units (SI):
519
eccce61a
TK
520 * *K* -- means kilo (K) or 1000
521 * *M* -- means mega (M) or 1000**2
522 * *G* -- means giga (G) or 1000**3
523 * *T* -- means tera (T) or 1000**4
524 * *P* -- means peta (P) or 1000**5
f80dba8d
MT
525
526 To specify power-of-2 binary values defined in IEC 80000-13:
527
eccce61a
TK
528 * *Ki* -- means kibi (Ki) or 1024
529 * *Mi* -- means mebi (Mi) or 1024**2
530 * *Gi* -- means gibi (Gi) or 1024**3
531 * *Ti* -- means tebi (Ti) or 1024**4
532 * *Pi* -- means pebi (Pi) or 1024**5
f80dba8d 533
9207a0cb 534 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
f80dba8d
MT
535 from those specified in the SI and IEC 80000-13 standards to provide
536 compatibility with old scripts. For example, 4k means 4096.
537
538 For quantities of data, an optional unit of 'B' may be included
b8f7e412 539 (e.g., 'kB' is the same as 'k').
f80dba8d
MT
540
541 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
542 not milli). 'b' and 'B' both mean byte, not bit.
543
9207a0cb 544 Examples with :option:`kb_base`\=1000:
f80dba8d
MT
545
546 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
547 * *1 MiB*: 1048576, 1mi, 1024ki
548 * *1 MB*: 1000000, 1m, 1000k
549 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
550 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
551
9207a0cb 552 Examples with :option:`kb_base`\=1024 (default):
f80dba8d
MT
553
554 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
555 * *1 MiB*: 1048576, 1m, 1024k
556 * *1 MB*: 1000000, 1mi, 1000ki
557 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
558 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
559
560 To specify times (units are not case sensitive):
561
562 * *D* -- means days
563 * *H* -- means hours
4502cb42 564 * *M* -- means minutes
f80dba8d
MT
565 * *s* -- or sec means seconds (default)
566 * *ms* -- or *msec* means milliseconds
567 * *us* -- or *usec* means microseconds
568
569 If the option accepts an upper and lower range, use a colon ':' or
570 minus '-' to separate such values. See :ref:`irange <irange>`.
4502cb42
SW
571 If the lower value specified happens to be larger than the upper value
572 the two values are swapped.
f80dba8d
MT
573
574.. _bool:
575
576**bool**
577 Boolean. Usually parsed as an integer, however only defined for
578 true and false (1 and 0).
579
580.. _irange:
581
582**irange**
583 Integer range with suffix. Allows value range to be given, such as
c60ebc45 584 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
f80dba8d
MT
585 option allows two sets of ranges, they can be specified with a ',' or '/'
586 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
587
588**float_list**
589 A list of floating point numbers, separated by a ':' character.
590
f5c3bcf2
TK
591With the above in mind, here follows the complete list of fio job parameters.
592
f80dba8d
MT
593
594Units
595~~~~~
596
597.. option:: kb_base=int
598
599 Select the interpretation of unit prefixes in input parameters.
600
601 **1000**
602 Inputs comply with IEC 80000-13 and the International
603 System of Units (SI). Use:
604
605 - power-of-2 values with IEC prefixes (e.g., KiB)
606 - power-of-10 values with SI prefixes (e.g., kB)
607
608 **1024**
609 Compatibility mode (default). To avoid breaking old scripts:
610
611 - power-of-2 values with SI prefixes
612 - power-of-10 values with IEC prefixes
613
614 See :option:`bs` for more details on input parameters.
615
616 Outputs always use correct prefixes. Most outputs include both
617 side-by-side, like::
618
619 bw=2383.3kB/s (2327.4KiB/s)
620
621 If only one value is reported, then kb_base selects the one to use:
622
623 **1000** -- SI prefixes
624
625 **1024** -- IEC prefixes
626
627.. option:: unit_base=int
628
629 Base unit for reporting. Allowed values are:
630
631 **0**
632 Use auto-detection (default).
633 **8**
634 Byte based.
635 **1**
636 Bit based.
637
638
f80dba8d
MT
639Job description
640~~~~~~~~~~~~~~~
641
642.. option:: name=str
643
644 ASCII name of the job. This may be used to override the name printed by fio
645 for this job. Otherwise the job name is used. On the command line this
646 parameter has the special purpose of also signaling the start of a new job.
647
648.. option:: description=str
649
650 Text description of the job. Doesn't do anything except dump this text
651 description when this job is run. It's not parsed.
652
653.. option:: loops=int
654
655 Run the specified number of iterations of this job. Used to repeat the same
656 workload a given number of times. Defaults to 1.
657
658.. option:: numjobs=int
659
79591fa9
TK
660 Create the specified number of clones of this job. Each clone of job
661 is spawned as an independent thread or process. May be used to setup a
f80dba8d
MT
662 larger number of threads/processes doing the same thing. Each thread is
663 reported separately; to see statistics for all clones as a whole, use
664 :option:`group_reporting` in conjunction with :option:`new_group`.
a47b697c 665 See :option:`--max-jobs`. Default: 1.
f80dba8d
MT
666
667
668Time related parameters
669~~~~~~~~~~~~~~~~~~~~~~~
670
671.. option:: runtime=time
672
f75ede1d 673 Tell fio to terminate processing after the specified period of time. It
f80dba8d 674 can be quite hard to determine for how long a specified job will run, so
f75ede1d 675 this parameter is handy to cap the total runtime to a given time. When
804c0839 676 the unit is omitted, the value is interpreted in seconds.
f80dba8d
MT
677
678.. option:: time_based
679
680 If set, fio will run for the duration of the :option:`runtime` specified
681 even if the file(s) are completely read or written. It will simply loop over
682 the same workload as many times as the :option:`runtime` allows.
683
a881438b 684.. option:: startdelay=irange(time)
f80dba8d 685
947e0fe0
SW
686 Delay the start of job for the specified amount of time. Can be a single
687 value or a range. When given as a range, each thread will choose a value
688 randomly from within the range. Value is in seconds if a unit is omitted.
f80dba8d
MT
689
690.. option:: ramp_time=time
691
692 If set, fio will run the specified workload for this amount of time before
693 logging any performance numbers. Useful for letting performance settle
694 before logging results, thus minimizing the runtime required for stable
695 results. Note that the ``ramp_time`` is considered lead in time for a job,
696 thus it will increase the total runtime if a special timeout or
f75ede1d
SW
697 :option:`runtime` is specified. When the unit is omitted, the value is
698 given in seconds.
f80dba8d
MT
699
700.. option:: clocksource=str
701
702 Use the given clocksource as the base of timing. The supported options are:
703
704 **gettimeofday**
705 :manpage:`gettimeofday(2)`
706
707 **clock_gettime**
708 :manpage:`clock_gettime(2)`
709
710 **cpu**
711 Internal CPU clock source
712
713 cpu is the preferred clocksource if it is reliable, as it is very fast (and
714 fio is heavy on time calls). Fio will automatically use this clocksource if
715 it's supported and considered reliable on the system it is running on,
716 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
717 means supporting TSC Invariant.
718
719.. option:: gtod_reduce=bool
720
721 Enable all of the :manpage:`gettimeofday(2)` reducing options
f75ede1d 722 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
f80dba8d
MT
723 reduce precision of the timeout somewhat to really shrink the
724 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
725 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
726 time keeping was enabled.
727
728.. option:: gtod_cpu=int
729
730 Sometimes it's cheaper to dedicate a single thread of execution to just
731 getting the current time. Fio (and databases, for instance) are very
732 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
733 one CPU aside for doing nothing but logging current time to a shared memory
734 location. Then the other threads/processes that run I/O workloads need only
735 copy that segment, instead of entering the kernel with a
736 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
737 calls will be excluded from other uses. Fio will manually clear it from the
738 CPU mask of other jobs.
739
740
741Target file/device
742~~~~~~~~~~~~~~~~~~
743
744.. option:: directory=str
745
746 Prefix filenames with this directory. Used to place files in a different
747 location than :file:`./`. You can specify a number of directories by
748 separating the names with a ':' character. These directories will be
02dd2689 749 assigned equally distributed to job clones created by :option:`numjobs` as
f80dba8d
MT
750 long as they are using generated filenames. If specific `filename(s)` are
751 set fio will use the first listed directory, and thereby matching the
f4401bf8
SW
752 `filename` semantic (which generates a file for each clone if not
753 specified, but lets all clones use the same file if set).
f80dba8d 754
02dd2689
SW
755 See the :option:`filename` option for information on how to escape "``:``" and
756 "``\``" characters within the directory path itself.
f80dba8d 757
f4401bf8
SW
758 Note: To control the directory fio will use for internal state files
759 use :option:`--aux-path`.
760
f80dba8d
MT
761.. option:: filename=str
762
763 Fio normally makes up a `filename` based on the job name, thread number, and
02dd2689
SW
764 file number (see :option:`filename_format`). If you want to share files
765 between threads in a job or several
79591fa9
TK
766 jobs with fixed file paths, specify a `filename` for each of them to override
767 the default. If the ioengine is file based, you can specify a number of files
768 by separating the names with a ':' colon. So if you wanted a job to open
769 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
770 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
771 specified, :option:`nrfiles` is ignored. The size of regular files specified
02dd2689 772 by this option will be :option:`size` divided by number of files unless an
79591fa9
TK
773 explicit size is specified by :option:`filesize`.
774
02dd2689
SW
775 Each colon and backslash in the wanted path must be escaped with a ``\``
776 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
777 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
778 :file:`F:\\filename` then you would use ``filename=F\:\\filename``.
779
f80dba8d
MT
780 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
781 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
782 Note: Windows and FreeBSD prevent write access to areas
02dd2689
SW
783 of the disk containing in-use data (e.g. filesystems).
784
785 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
786 of the two depends on the read/write direction set.
f80dba8d
MT
787
788.. option:: filename_format=str
789
790 If sharing multiple files between jobs, it is usually necessary to have fio
791 generate the exact names that you want. By default, fio will name a file
792 based on the default file format specification of
793 :file:`jobname.jobnumber.filenumber`. With this option, that can be
794 customized. Fio will recognize and replace the following keywords in this
795 string:
796
797 **$jobname**
798 The name of the worker thread or process.
799 **$jobnum**
800 The incremental number of the worker thread or process.
801 **$filenum**
802 The incremental number of the file for that worker thread or
803 process.
804
805 To have dependent jobs share a set of files, this option can be set to have
806 fio generate filenames that are shared between the two. For instance, if
807 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
808 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
809 will be used if no other format specifier is given.
810
645943c0
JB
811 If you specify a path then the directories will be created up to the
812 main directory for the file. So for example if you specify
813 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
814 created before the file setup part of the job. If you specify
815 :option:`directory` then the path will be relative that directory,
816 otherwise it is treated as the absolute path.
817
f80dba8d
MT
818.. option:: unique_filename=bool
819
820 To avoid collisions between networked clients, fio defaults to prefixing any
821 generated filenames (with a directory specified) with the source of the
822 client connecting. To disable this behavior, set this option to 0.
823
824.. option:: opendir=str
825
826 Recursively open any files below directory `str`.
827
828.. option:: lockfile=str
829
830 Fio defaults to not locking any files before it does I/O to them. If a file
831 or file descriptor is shared, fio can serialize I/O to that file to make the
832 end result consistent. This is usual for emulating real workloads that share
833 files. The lock modes are:
834
835 **none**
836 No locking. The default.
837 **exclusive**
838 Only one thread or process may do I/O at a time, excluding all
839 others.
840 **readwrite**
841 Read-write locking on the file. Many readers may
842 access the file at the same time, but writes get exclusive access.
843
844.. option:: nrfiles=int
845
79591fa9
TK
846 Number of files to use for this job. Defaults to 1. The size of files
847 will be :option:`size` divided by this unless explicit size is specified by
848 :option:`filesize`. Files are created for each thread separately, and each
849 file will have a file number within its name by default, as explained in
850 :option:`filename` section.
851
f80dba8d
MT
852
853.. option:: openfiles=int
854
855 Number of files to keep open at the same time. Defaults to the same as
856 :option:`nrfiles`, can be set smaller to limit the number simultaneous
857 opens.
858
859.. option:: file_service_type=str
860
861 Defines how fio decides which file from a job to service next. The following
862 types are defined:
863
864 **random**
865 Choose a file at random.
866
867 **roundrobin**
868 Round robin over opened files. This is the default.
869
870 **sequential**
871 Finish one file before moving on to the next. Multiple files can
f50fbdda 872 still be open depending on :option:`openfiles`.
f80dba8d
MT
873
874 **zipf**
c60ebc45 875 Use a *Zipf* distribution to decide what file to access.
f80dba8d
MT
876
877 **pareto**
c60ebc45 878 Use a *Pareto* distribution to decide what file to access.
f80dba8d 879
dd3503d3 880 **normal**
c60ebc45 881 Use a *Gaussian* (normal) distribution to decide what file to
f80dba8d
MT
882 access.
883
dd3503d3
SW
884 **gauss**
885 Alias for normal.
886
f80dba8d
MT
887 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
888 tell fio how many I/Os to issue before switching to a new file. For example,
889 specifying ``file_service_type=random:8`` would cause fio to issue
890 8 I/Os before selecting a new file at random. For the non-uniform
891 distributions, a floating point postfix can be given to influence how the
892 distribution is skewed. See :option:`random_distribution` for a description
893 of how that would work.
894
895.. option:: ioscheduler=str
896
897 Attempt to switch the device hosting the file to the specified I/O scheduler
898 before running.
899
900.. option:: create_serialize=bool
901
902 If true, serialize the file creation for the jobs. This may be handy to
903 avoid interleaving of data files, which may greatly depend on the filesystem
a47b697c 904 used and even the number of processors in the system. Default: true.
f80dba8d
MT
905
906.. option:: create_fsync=bool
907
22413915 908 :manpage:`fsync(2)` the data file after creation. This is the default.
f80dba8d
MT
909
910.. option:: create_on_open=bool
911
730bd7d9
SW
912 If true, don't pre-create files but allow the job's open() to create a file
913 when it's time to do I/O. Default: false -- pre-create all necessary files
914 when the job starts.
f80dba8d
MT
915
916.. option:: create_only=bool
917
918 If true, fio will only run the setup phase of the job. If files need to be
4502cb42 919 laid out or updated on disk, only that will be done -- the actual job contents
a47b697c 920 are not executed. Default: false.
f80dba8d
MT
921
922.. option:: allow_file_create=bool
923
730bd7d9
SW
924 If true, fio is permitted to create files as part of its workload. If this
925 option is false, then fio will error out if
f80dba8d
MT
926 the files it needs to use don't already exist. Default: true.
927
928.. option:: allow_mounted_write=bool
929
c60ebc45 930 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
f80dba8d
MT
931 to what appears to be a mounted device or partition. This should help catch
932 creating inadvertently destructive tests, not realizing that the test will
b1db0375
TK
933 destroy data on the mounted file system. Note that some platforms don't allow
934 writing against a mounted device regardless of this option. Default: false.
f80dba8d
MT
935
936.. option:: pre_read=bool
937
938 If this is given, files will be pre-read into memory before starting the
939 given I/O operation. This will also clear the :option:`invalidate` flag,
940 since it is pointless to pre-read and then drop the cache. This will only
941 work for I/O engines that are seek-able, since they allow you to read the
a47b697c
SW
942 same data multiple times. Thus it will not work on non-seekable I/O engines
943 (e.g. network, splice). Default: false.
f80dba8d
MT
944
945.. option:: unlink=bool
946
947 Unlink the job files when done. Not the default, as repeated runs of that
a47b697c
SW
948 job would then waste time recreating the file set again and again. Default:
949 false.
f80dba8d
MT
950
951.. option:: unlink_each_loop=bool
952
a47b697c 953 Unlink job files after each iteration or loop. Default: false.
f80dba8d 954
7b865a2f
BVA
955.. option:: zonemode=str
956
957 Accepted values are:
958
959 **none**
960 The :option:`zonerange`, :option:`zonesize` and
961 :option:`zoneskip` parameters are ignored.
962 **strided**
963 I/O happens in a single zone until
964 :option:`zonesize` bytes have been transferred.
965 After that number of bytes has been
966 transferred processing of the next zone
967 starts.
968 **zbd**
969 Zoned block device mode. I/O happens
970 sequentially in each zone, even if random I/O
971 has been selected. Random I/O happens across
972 all zones instead of being restricted to a
973 single zone. The :option:`zoneskip` parameter
974 is ignored. :option:`zonerange` and
975 :option:`zonesize` must be identical.
976
5faddc64 977.. option:: zonerange=int
f80dba8d 978
7b865a2f
BVA
979 Size of a single zone. See also :option:`zonesize` and
980 :option:`zoneskip`.
f80dba8d 981
5faddc64 982.. option:: zonesize=int
f80dba8d 983
7b865a2f
BVA
984 For :option:`zonemode` =strided, this is the number of bytes to
985 transfer before skipping :option:`zoneskip` bytes. If this parameter
986 is smaller than :option:`zonerange` then only a fraction of each zone
987 with :option:`zonerange` bytes will be accessed. If this parameter is
988 larger than :option:`zonerange` then each zone will be accessed
989 multiple times before skipping to the next zone.
990
991 For :option:`zonemode` =zbd, this is the size of a single zone. The
992 :option:`zonerange` parameter is ignored in this mode.
f80dba8d
MT
993
994.. option:: zoneskip=int
995
7b865a2f
BVA
996 For :option:`zonemode` =strided, the number of bytes to skip after
997 :option:`zonesize` bytes of data have been transferred. This parameter
998 must be zero for :option:`zonemode` =zbd.
f80dba8d 999
bfbdd35b
BVA
1000.. option:: read_beyond_wp=bool
1001
1002 This parameter applies to :option:`zonemode` =zbd only.
1003
1004 Zoned block devices are block devices that consist of multiple zones.
1005 Each zone has a type, e.g. conventional or sequential. A conventional
1006 zone can be written at any offset that is a multiple of the block
1007 size. Sequential zones must be written sequentially. The position at
1008 which a write must occur is called the write pointer. A zoned block
1009 device can be either drive managed, host managed or host aware. For
1010 host managed devices the host must ensure that writes happen
1011 sequentially. Fio recognizes host managed devices and serializes
1012 writes to sequential zones for these devices.
1013
1014 If a read occurs in a sequential zone beyond the write pointer then
1015 the zoned block device will complete the read without reading any data
1016 from the storage medium. Since such reads lead to unrealistically high
1017 bandwidth and IOPS numbers fio only reads beyond the write pointer if
1018 explicitly told to do so. Default: false.
1019
59b07544
BVA
1020.. option:: max_open_zones=int
1021
1022 When running a random write test across an entire drive many more
1023 zones will be open than in a typical application workload. Hence this
1024 command line option that allows to limit the number of open zones. The
1025 number of open zones is defined as the number of zones to which write
1026 commands are issued.
1027
a7c2b6fc
BVA
1028.. option:: zone_reset_threshold=float
1029
1030 A number between zero and one that indicates the ratio of logical
1031 blocks with data to the total number of logical blocks in the test
1032 above which zones should be reset periodically.
1033
1034.. option:: zone_reset_frequency=float
1035
1036 A number between zero and one that indicates how often a zone reset
1037 should be issued if the zone reset threshold has been exceeded. A zone
1038 reset is submitted after each (1 / zone_reset_frequency) write
1039 requests. This and the previous parameter can be used to simulate
1040 garbage collection activity.
1041
f80dba8d
MT
1042
1043I/O type
1044~~~~~~~~
1045
1046.. option:: direct=bool
1047
1048 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
8e889110 1049 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
f80dba8d
MT
1050 ioengines don't support direct I/O. Default: false.
1051
1052.. option:: atomic=bool
1053
1054 If value is true, attempt to use atomic direct I/O. Atomic writes are
1055 guaranteed to be stable once acknowledged by the operating system. Only
1056 Linux supports O_ATOMIC right now.
1057
1058.. option:: buffered=bool
1059
1060 If value is true, use buffered I/O. This is the opposite of the
1061 :option:`direct` option. Defaults to true.
1062
1063.. option:: readwrite=str, rw=str
1064
1065 Type of I/O pattern. Accepted values are:
1066
1067 **read**
1068 Sequential reads.
1069 **write**
1070 Sequential writes.
1071 **trim**
3740cfc8
VF
1072 Sequential trims (Linux block devices and SCSI
1073 character devices only).
f80dba8d
MT
1074 **randread**
1075 Random reads.
2831be97
SW
1076 **randwrite**
1077 Random writes.
f80dba8d 1078 **randtrim**
3740cfc8
VF
1079 Random trims (Linux block devices and SCSI
1080 character devices only).
f80dba8d
MT
1081 **rw,readwrite**
1082 Sequential mixed reads and writes.
1083 **randrw**
1084 Random mixed reads and writes.
1085 **trimwrite**
1086 Sequential trim+write sequences. Blocks will be trimmed first,
1087 then the same blocks will be written to.
1088
1089 Fio defaults to read if the option is not specified. For the mixed I/O
1090 types, the default is to split them 50/50. For certain types of I/O the
730bd7d9
SW
1091 result may still be skewed a bit, since the speed may be different.
1092
1093 It is possible to specify the number of I/Os to do before getting a new
1094 offset by appending ``:<nr>`` to the end of the string given. For a
f80dba8d
MT
1095 random read, it would look like ``rw=randread:8`` for passing in an offset
1096 modifier with a value of 8. If the suffix is used with a sequential I/O
730bd7d9
SW
1097 pattern, then the *<nr>* value specified will be **added** to the generated
1098 offset for each I/O turning sequential I/O into sequential I/O with holes.
1099 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1100 the :option:`rw_sequencer` option.
f80dba8d
MT
1101
1102.. option:: rw_sequencer=str
1103
1104 If an offset modifier is given by appending a number to the ``rw=<str>``
1105 line, then this option controls how that number modifies the I/O offset
1106 being generated. Accepted values are:
1107
1108 **sequential**
1109 Generate sequential offset.
1110 **identical**
1111 Generate the same offset.
1112
1113 ``sequential`` is only useful for random I/O, where fio would normally
c60ebc45 1114 generate a new random offset for every I/O. If you append e.g. 8 to randread,
007c7be9
SW
1115 you would get a new random offset for every 8 I/Os. The result would be a
1116 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
f80dba8d
MT
1117 to specify that. As sequential I/O is already sequential, setting
1118 ``sequential`` for that would not result in any differences. ``identical``
1119 behaves in a similar fashion, except it sends the same offset 8 number of
1120 times before generating a new offset.
1121
1122.. option:: unified_rw_reporting=bool
1123
1124 Fio normally reports statistics on a per data direction basis, meaning that
1125 reads, writes, and trims are accounted and reported separately. If this
1126 option is set fio sums the results and report them as "mixed" instead.
1127
1128.. option:: randrepeat=bool
1129
1130 Seed the random number generator used for random I/O patterns in a
1131 predictable way so the pattern is repeatable across runs. Default: true.
1132
1133.. option:: allrandrepeat=bool
1134
1135 Seed all random number generators in a predictable way so results are
1136 repeatable across runs. Default: false.
1137
1138.. option:: randseed=int
1139
1140 Seed the random number generators based on this seed value, to be able to
1141 control what sequence of output is being generated. If not set, the random
1142 sequence depends on the :option:`randrepeat` setting.
1143
1144.. option:: fallocate=str
1145
1146 Whether pre-allocation is performed when laying down files.
1147 Accepted values are:
1148
1149 **none**
1150 Do not pre-allocate space.
1151
2c3e17be
SW
1152 **native**
1153 Use a platform's native pre-allocation call but fall back to
1154 **none** behavior if it fails/is not implemented.
1155
f80dba8d
MT
1156 **posix**
1157 Pre-allocate via :manpage:`posix_fallocate(3)`.
1158
1159 **keep**
1160 Pre-allocate via :manpage:`fallocate(2)` with
1161 FALLOC_FL_KEEP_SIZE set.
1162
1163 **0**
1164 Backward-compatible alias for **none**.
1165
1166 **1**
1167 Backward-compatible alias for **posix**.
1168
1169 May not be available on all supported platforms. **keep** is only available
2c3e17be
SW
1170 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1171 because ZFS doesn't support pre-allocation. Default: **native** if any
1172 pre-allocation methods are available, **none** if not.
f80dba8d
MT
1173
1174.. option:: fadvise_hint=str
1175
c712c97a
JA
1176 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1177 advise the kernel on what I/O patterns are likely to be issued.
1178 Accepted values are:
f80dba8d
MT
1179
1180 **0**
1181 Backwards-compatible hint for "no hint".
1182
1183 **1**
1184 Backwards compatible hint for "advise with fio workload type". This
1185 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1186 for a sequential workload.
1187
1188 **sequential**
1189 Advise using **FADV_SEQUENTIAL**.
1190
1191 **random**
1192 Advise using **FADV_RANDOM**.
1193
8f4b9f24 1194.. option:: write_hint=str
f80dba8d 1195
8f4b9f24
JA
1196 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1197 from a write. Only supported on Linux, as of version 4.13. Accepted
1198 values are:
1199
1200 **none**
1201 No particular life time associated with this file.
1202
1203 **short**
1204 Data written to this file has a short life time.
1205
1206 **medium**
1207 Data written to this file has a medium life time.
1208
1209 **long**
1210 Data written to this file has a long life time.
1211
1212 **extreme**
1213 Data written to this file has a very long life time.
1214
1215 The values are all relative to each other, and no absolute meaning
1216 should be associated with them.
f80dba8d
MT
1217
1218.. option:: offset=int
1219
82dbb8cb 1220 Start I/O at the provided offset in the file, given as either a fixed size in
83c8b093
JF
1221 bytes or a percentage. If a percentage is given, the generated offset will be
1222 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1223 provided. Data before the given offset will not be touched. This
89978a6b
BW
1224 effectively caps the file size at `real_size - offset`. Can be combined with
1225 :option:`size` to constrain the start and end range of the I/O workload.
44bb1142
TK
1226 A percentage can be specified by a number between 1 and 100 followed by '%',
1227 for example, ``offset=20%`` to specify 20%.
f80dba8d 1228
83c8b093
JF
1229.. option:: offset_align=int
1230
1231 If set to non-zero value, the byte offset generated by a percentage ``offset``
1232 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1233 offset is aligned to the minimum block size.
1234
f80dba8d
MT
1235.. option:: offset_increment=int
1236
1237 If this is provided, then the real offset becomes `offset + offset_increment
1238 * thread_number`, where the thread number is a counter that starts at 0 and
1239 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1240 specified). This option is useful if there are several jobs which are
1241 intended to operate on a file in parallel disjoint segments, with even
1242 spacing between the starting points.
1243
1244.. option:: number_ios=int
1245
c60ebc45 1246 Fio will normally perform I/Os until it has exhausted the size of the region
f80dba8d
MT
1247 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1248 condition). With this setting, the range/size can be set independently of
c60ebc45 1249 the number of I/Os to perform. When fio reaches this number, it will exit
f80dba8d
MT
1250 normally and report status. Note that this does not extend the amount of I/O
1251 that will be done, it will only stop fio if this condition is met before
1252 other end-of-job criteria.
1253
1254.. option:: fsync=int
1255
730bd7d9
SW
1256 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1257 the dirty data for every number of blocks given. For example, if you give 32
1258 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1259 using non-buffered I/O, we may not sync the file. The exception is the sg
1260 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1261 means fio does not periodically issue and wait for a sync to complete. Also
1262 see :option:`end_fsync` and :option:`fsync_on_close`.
f80dba8d
MT
1263
1264.. option:: fdatasync=int
1265
1266 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
000a5f1c 1267 not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
730bd7d9
SW
1268 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1269 Defaults to 0, which means fio does not periodically issue and wait for a
1270 data-only sync to complete.
f80dba8d
MT
1271
1272.. option:: write_barrier=int
1273
2831be97 1274 Make every `N-th` write a barrier write.
f80dba8d 1275
f50fbdda 1276.. option:: sync_file_range=str:int
f80dba8d 1277
f50fbdda 1278 Use :manpage:`sync_file_range(2)` for every `int` number of write
f80dba8d
MT
1279 operations. Fio will track range of writes that have happened since the last
1280 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1281
1282 **wait_before**
1283 SYNC_FILE_RANGE_WAIT_BEFORE
1284 **write**
1285 SYNC_FILE_RANGE_WRITE
1286 **wait_after**
1287 SYNC_FILE_RANGE_WAIT_AFTER
1288
1289 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1290 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1291 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1292 Linux specific.
1293
1294.. option:: overwrite=bool
1295
1296 If true, writes to a file will always overwrite existing data. If the file
1297 doesn't already exist, it will be created before the write phase begins. If
1298 the file exists and is large enough for the specified write phase, nothing
a47b697c 1299 will be done. Default: false.
f80dba8d
MT
1300
1301.. option:: end_fsync=bool
1302
a47b697c
SW
1303 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1304 Default: false.
f80dba8d
MT
1305
1306.. option:: fsync_on_close=bool
1307
1308 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
a47b697c
SW
1309 from :option:`end_fsync` in that it will happen on every file close, not
1310 just at the end of the job. Default: false.
f80dba8d
MT
1311
1312.. option:: rwmixread=int
1313
1314 Percentage of a mixed workload that should be reads. Default: 50.
1315
1316.. option:: rwmixwrite=int
1317
1318 Percentage of a mixed workload that should be writes. If both
1319 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1320 add up to 100%, the latter of the two will be used to override the
1321 first. This may interfere with a given rate setting, if fio is asked to
1322 limit reads or writes to a certain rate. If that is the case, then the
1323 distribution may be skewed. Default: 50.
1324
1325.. option:: random_distribution=str:float[,str:float][,str:float]
1326
1327 By default, fio will use a completely uniform random distribution when asked
1328 to perform random I/O. Sometimes it is useful to skew the distribution in
1329 specific ways, ensuring that some parts of the data is more hot than others.
1330 fio includes the following distribution models:
1331
1332 **random**
1333 Uniform random distribution
1334
1335 **zipf**
1336 Zipf distribution
1337
1338 **pareto**
1339 Pareto distribution
1340
b2f4b559 1341 **normal**
c60ebc45 1342 Normal (Gaussian) distribution
f80dba8d
MT
1343
1344 **zoned**
1345 Zoned random distribution
1346
59466396
JA
1347 **zoned_abs**
1348 Zone absolute random distribution
1349
f80dba8d 1350 When using a **zipf** or **pareto** distribution, an input value is also
f50fbdda 1351 needed to define the access pattern. For **zipf**, this is the `Zipf
c60ebc45 1352 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
f50fbdda 1353 program, :command:`fio-genzipf`, that can be used visualize what the given input
f80dba8d
MT
1354 values will yield in terms of hit rates. If you wanted to use **zipf** with
1355 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1356 option. If a non-uniform model is used, fio will disable use of the random
b2f4b559
SW
1357 map. For the **normal** distribution, a normal (Gaussian) deviation is
1358 supplied as a value between 0 and 100.
f80dba8d
MT
1359
1360 For a **zoned** distribution, fio supports specifying percentages of I/O
1361 access that should fall within what range of the file or device. For
1362 example, given a criteria of:
1363
f50fbdda
TK
1364 * 60% of accesses should be to the first 10%
1365 * 30% of accesses should be to the next 20%
1366 * 8% of accesses should be to the next 30%
1367 * 2% of accesses should be to the next 40%
f80dba8d
MT
1368
1369 we can define that through zoning of the random accesses. For the above
1370 example, the user would do::
1371
1372 random_distribution=zoned:60/10:30/20:8/30:2/40
1373
59466396
JA
1374 A **zoned_abs** distribution works exactly like the **zoned**, except
1375 that it takes absolute sizes. For example, let's say you wanted to
1376 define access according to the following criteria:
1377
1378 * 60% of accesses should be to the first 20G
1379 * 30% of accesses should be to the next 100G
1380 * 10% of accesses should be to the next 500G
1381
1382 we can define an absolute zoning distribution with:
1383
1384 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1385
6a16ece8
JA
1386 For both **zoned** and **zoned_abs**, fio supports defining up to
1387 256 separate zones.
1388
59466396
JA
1389 Similarly to how :option:`bssplit` works for setting ranges and
1390 percentages of block sizes. Like :option:`bssplit`, it's possible to
1391 specify separate zones for reads, writes, and trims. If just one set
1392 is given, it'll apply to all of them. This goes for both **zoned**
1393 **zoned_abs** distributions.
f80dba8d
MT
1394
1395.. option:: percentage_random=int[,int][,int]
1396
1397 For a random workload, set how big a percentage should be random. This
1398 defaults to 100%, in which case the workload is fully random. It can be set
1399 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1400 sequential. Any setting in between will result in a random mix of sequential
1401 and random I/O, at the given percentages. Comma-separated values may be
1402 specified for reads, writes, and trims as described in :option:`blocksize`.
1403
1404.. option:: norandommap
1405
1406 Normally fio will cover every block of the file when doing random I/O. If
1407 this option is given, fio will just get a new random offset without looking
1408 at past I/O history. This means that some blocks may not be read or written,
1409 and that some blocks may be read/written more than once. If this option is
1410 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1411 only intact blocks are verified, i.e., partially-overwritten blocks are
47e6a6e5
SW
1412 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1413 the same block to be overwritten, which can cause verification errors. Either
1414 do not use norandommap in this case, or also use the lfsr random generator.
f80dba8d
MT
1415
1416.. option:: softrandommap=bool
1417
1418 See :option:`norandommap`. If fio runs with the random block map enabled and
1419 it fails to allocate the map, if this option is set it will continue without
1420 a random block map. As coverage will not be as complete as with random maps,
1421 this option is disabled by default.
1422
1423.. option:: random_generator=str
1424
f50fbdda 1425 Fio supports the following engines for generating I/O offsets for random I/O:
f80dba8d
MT
1426
1427 **tausworthe**
f50fbdda 1428 Strong 2^88 cycle random number generator.
f80dba8d 1429 **lfsr**
f50fbdda 1430 Linear feedback shift register generator.
f80dba8d 1431 **tausworthe64**
f50fbdda 1432 Strong 64-bit 2^258 cycle random number generator.
f80dba8d
MT
1433
1434 **tausworthe** is a strong random number generator, but it requires tracking
1435 on the side if we want to ensure that blocks are only read or written
f50fbdda 1436 once. **lfsr** guarantees that we never generate the same offset twice, and
f80dba8d 1437 it's also less computationally expensive. It's not a true random generator,
f50fbdda 1438 however, though for I/O purposes it's typically good enough. **lfsr** only
f80dba8d
MT
1439 works with single block sizes, not with workloads that use multiple block
1440 sizes. If used with such a workload, fio may read or write some blocks
1441 multiple times. The default value is **tausworthe**, unless the required
1442 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1443 selected automatically.
1444
1445
1446Block size
1447~~~~~~~~~~
1448
1449.. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1450
1451 The block size in bytes used for I/O units. Default: 4096. A single value
1452 applies to reads, writes, and trims. Comma-separated values may be
1453 specified for reads, writes, and trims. A value not terminated in a comma
1454 applies to subsequent types.
1455
1456 Examples:
1457
1458 **bs=256k**
1459 means 256k for reads, writes and trims.
1460
1461 **bs=8k,32k**
1462 means 8k for reads, 32k for writes and trims.
1463
1464 **bs=8k,32k,**
1465 means 8k for reads, 32k for writes, and default for trims.
1466
1467 **bs=,8k**
1468 means default for reads, 8k for writes and trims.
1469
1470 **bs=,8k,**
b443ae44 1471 means default for reads, 8k for writes, and default for trims.
f80dba8d
MT
1472
1473.. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1474
1475 A range of block sizes in bytes for I/O units. The issued I/O unit will
1476 always be a multiple of the minimum size, unless
1477 :option:`blocksize_unaligned` is set.
1478
1479 Comma-separated ranges may be specified for reads, writes, and trims as
1480 described in :option:`blocksize`.
1481
1482 Example: ``bsrange=1k-4k,2k-8k``.
1483
1484.. option:: bssplit=str[,str][,str]
1485
6a16ece8
JA
1486 Sometimes you want even finer grained control of the block sizes
1487 issued, not just an even split between them. This option allows you to
1488 weight various block sizes, so that you are able to define a specific
1489 amount of block sizes issued. The format for this option is::
f80dba8d
MT
1490
1491 bssplit=blocksize/percentage:blocksize/percentage
1492
6a16ece8
JA
1493 for as many block sizes as needed. So if you want to define a workload
1494 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1495 write::
f80dba8d
MT
1496
1497 bssplit=4k/10:64k/50:32k/40
1498
6a16ece8
JA
1499 Ordering does not matter. If the percentage is left blank, fio will
1500 fill in the remaining values evenly. So a bssplit option like this one::
f80dba8d
MT
1501
1502 bssplit=4k/50:1k/:32k/
1503
6a16ece8
JA
1504 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1505 add up to 100, if bssplit is given a range that adds up to more, it
1506 will error out.
f80dba8d
MT
1507
1508 Comma-separated values may be specified for reads, writes, and trims as
1509 described in :option:`blocksize`.
1510
6a16ece8
JA
1511 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1512 having 90% 4k writes and 10% 8k writes, you would specify::
f80dba8d 1513
cf04b906 1514 bssplit=2k/50:4k/50,4k/90:8k/10
f80dba8d 1515
6a16ece8
JA
1516 Fio supports defining up to 64 different weights for each data
1517 direction.
1518
f80dba8d
MT
1519.. option:: blocksize_unaligned, bs_unaligned
1520
1521 If set, fio will issue I/O units with any size within
1522 :option:`blocksize_range`, not just multiples of the minimum size. This
1523 typically won't work with direct I/O, as that normally requires sector
1524 alignment.
1525
589e88b7 1526.. option:: bs_is_seq_rand=bool
f80dba8d
MT
1527
1528 If this option is set, fio will use the normal read,write blocksize settings
1529 as sequential,random blocksize settings instead. Any random read or write
1530 will use the WRITE blocksize settings, and any sequential read or write will
1531 use the READ blocksize settings.
1532
1533.. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1534
1535 Boundary to which fio will align random I/O units. Default:
1536 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1537 I/O, though it usually depends on the hardware block size. This option is
1538 mutually exclusive with using a random map for files, so it will turn off
1539 that option. Comma-separated values may be specified for reads, writes, and
1540 trims as described in :option:`blocksize`.
1541
1542
1543Buffers and memory
1544~~~~~~~~~~~~~~~~~~
1545
1546.. option:: zero_buffers
1547
1548 Initialize buffers with all zeros. Default: fill buffers with random data.
1549
1550.. option:: refill_buffers
1551
1552 If this option is given, fio will refill the I/O buffers on every
72592780
SW
1553 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1554 naturally. Defaults to being unset i.e., the buffer is only filled at
1555 init time and the data in it is reused when possible but if any of
1556 :option:`verify`, :option:`buffer_compress_percentage` or
1557 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1558 automatically enabled.
f80dba8d
MT
1559
1560.. option:: scramble_buffers=bool
1561
1562 If :option:`refill_buffers` is too costly and the target is using data
1563 deduplication, then setting this option will slightly modify the I/O buffer
1564 contents to defeat normal de-dupe attempts. This is not enough to defeat
1565 more clever block compression attempts, but it will stop naive dedupe of
1566 blocks. Default: true.
1567
1568.. option:: buffer_compress_percentage=int
1569
72592780
SW
1570 If this is set, then fio will attempt to provide I/O buffer content
1571 (on WRITEs) that compresses to the specified level. Fio does this by
1572 providing a mix of random data followed by fixed pattern data. The
1573 fixed pattern is either zeros, or the pattern specified by
1574 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1575 might skew the compression ratio slightly. Setting
1576 `buffer_compress_percentage` to a value other than 100 will also
1577 enable :option:`refill_buffers` in order to reduce the likelihood that
1578 adjacent blocks are so similar that they over compress when seen
1579 together. See :option:`buffer_compress_chunk` for how to set a finer or
1580 coarser granularity for the random/fixed data region. Defaults to unset
1581 i.e., buffer data will not adhere to any compression level.
f80dba8d
MT
1582
1583.. option:: buffer_compress_chunk=int
1584
72592780
SW
1585 This setting allows fio to manage how big the random/fixed data region
1586 is when using :option:`buffer_compress_percentage`. When
1587 `buffer_compress_chunk` is set to some non-zero value smaller than the
1588 block size, fio can repeat the random/fixed region throughout the I/O
1589 buffer at the specified interval (which particularly useful when
1590 bigger block sizes are used for a job). When set to 0, fio will use a
1591 chunk size that matches the block size resulting in a single
1592 random/fixed region within the I/O buffer. Defaults to 512. When the
1593 unit is omitted, the value is interpreted in bytes.
f80dba8d
MT
1594
1595.. option:: buffer_pattern=str
1596
a1554f65
SB
1597 If set, fio will fill the I/O buffers with this pattern or with the contents
1598 of a file. If not set, the contents of I/O buffers are defined by the other
1599 options related to buffer contents. The setting can be any pattern of bytes,
1600 and can be prefixed with 0x for hex values. It may also be a string, where
1601 the string must then be wrapped with ``""``. Or it may also be a filename,
1602 where the filename must be wrapped with ``''`` in which case the file is
1603 opened and read. Note that not all the file contents will be read if that
1604 would cause the buffers to overflow. So, for example::
1605
1606 buffer_pattern='filename'
1607
1608 or::
f80dba8d
MT
1609
1610 buffer_pattern="abcd"
1611
1612 or::
1613
1614 buffer_pattern=-12
1615
1616 or::
1617
1618 buffer_pattern=0xdeadface
1619
1620 Also you can combine everything together in any order::
1621
a1554f65 1622 buffer_pattern=0xdeadface"abcd"-12'filename'
f80dba8d
MT
1623
1624.. option:: dedupe_percentage=int
1625
1626 If set, fio will generate this percentage of identical buffers when
1627 writing. These buffers will be naturally dedupable. The contents of the
1628 buffers depend on what other buffer compression settings have been set. It's
1629 possible to have the individual buffers either fully compressible, or not at
72592780
SW
1630 all -- this option only controls the distribution of unique buffers. Setting
1631 this option will also enable :option:`refill_buffers` to prevent every buffer
1632 being identical.
f80dba8d
MT
1633
1634.. option:: invalidate=bool
1635
730bd7d9
SW
1636 Invalidate the buffer/page cache parts of the files to be used prior to
1637 starting I/O if the platform and file type support it. Defaults to true.
21c1b29e
TK
1638 This will be ignored if :option:`pre_read` is also specified for the
1639 same job.
f80dba8d
MT
1640
1641.. option:: sync=bool
1642
1643 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1644 this means using O_SYNC. Default: false.
1645
1646.. option:: iomem=str, mem=str
1647
1648 Fio can use various types of memory as the I/O unit buffer. The allowed
1649 values are:
1650
1651 **malloc**
1652 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1653 type.
1654
1655 **shm**
1656 Use shared memory as the buffers. Allocated through
1657 :manpage:`shmget(2)`.
1658
1659 **shmhuge**
1660 Same as shm, but use huge pages as backing.
1661
1662 **mmap**
22413915 1663 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
f80dba8d
MT
1664 be file backed if a filename is given after the option. The format
1665 is `mem=mmap:/path/to/file`.
1666
1667 **mmaphuge**
1668 Use a memory mapped huge file as the buffer backing. Append filename
1669 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1670
1671 **mmapshared**
1672 Same as mmap, but use a MMAP_SHARED mapping.
1673
03553853
YR
1674 **cudamalloc**
1675 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
f50fbdda 1676 The :option:`ioengine` must be `rdma`.
03553853 1677
f80dba8d
MT
1678 The area allocated is a function of the maximum allowed bs size for the job,
1679 multiplied by the I/O depth given. Note that for **shmhuge** and
1680 **mmaphuge** to work, the system must have free huge pages allocated. This
1681 can normally be checked and set by reading/writing
1682 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1683 is 4MiB in size. So to calculate the number of huge pages you need for a
1684 given job file, add up the I/O depth of all jobs (normally one unless
1685 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1686 that number by the huge page size. You can see the size of the huge pages in
1687 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1688 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1689 see :option:`hugepage-size`.
1690
1691 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1692 should point there. So if it's mounted in :file:`/huge`, you would use
1693 `mem=mmaphuge:/huge/somefile`.
1694
f50fbdda 1695.. option:: iomem_align=int, mem_align=int
f80dba8d
MT
1696
1697 This indicates the memory alignment of the I/O memory buffers. Note that
1698 the given alignment is applied to the first I/O unit buffer, if using
1699 :option:`iodepth` the alignment of the following buffers are given by the
1700 :option:`bs` used. In other words, if using a :option:`bs` that is a
1701 multiple of the page sized in the system, all buffers will be aligned to
1702 this value. If using a :option:`bs` that is not page aligned, the alignment
1703 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1704 :option:`bs` used.
1705
1706.. option:: hugepage-size=int
1707
1708 Defines the size of a huge page. Must at least be equal to the system
1709 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1710 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1711 preferred way to set this to avoid setting a non-pow-2 bad value.
1712
1713.. option:: lockmem=int
1714
1715 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1716 simulate a smaller amount of memory. The amount specified is per worker.
1717
1718
1719I/O size
1720~~~~~~~~
1721
1722.. option:: size=int
1723
79591fa9
TK
1724 The total size of file I/O for each thread of this job. Fio will run until
1725 this many bytes has been transferred, unless runtime is limited by other options
1726 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1727 Fio will divide this size between the available files determined by options
1728 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1729 specified by the job. If the result of division happens to be 0, the size is
c4aa2d08 1730 set to the physical size of the given files or devices if they exist.
79591fa9 1731 If this option is not specified, fio will use the full size of the given
f80dba8d
MT
1732 files or devices. If the files do not exist, size must be given. It is also
1733 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1734 given, fio will use 20% of the full size of the given files or devices.
9d25d068
SW
1735 Can be combined with :option:`offset` to constrain the start and end range
1736 that I/O will be done within.
f80dba8d
MT
1737
1738.. option:: io_size=int, io_limit=int
1739
1740 Normally fio operates within the region set by :option:`size`, which means
1741 that the :option:`size` option sets both the region and size of I/O to be
1742 performed. Sometimes that is not what you want. With this option, it is
1743 possible to define just the amount of I/O that fio should do. For instance,
1744 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1745 will perform I/O within the first 20GiB but exit when 5GiB have been
1746 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1747 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1748 the 0..20GiB region.
1749
7fdd97ca 1750.. option:: filesize=irange(int)
f80dba8d
MT
1751
1752 Individual file sizes. May be a range, in which case fio will select sizes
1753 for files at random within the given range and limited to :option:`size` in
1754 total (if that is given). If not given, each created file is the same size.
79591fa9
TK
1755 This option overrides :option:`size` in terms of file size, which means
1756 this value is used as a fixed size or possible range of each file.
f80dba8d
MT
1757
1758.. option:: file_append=bool
1759
1760 Perform I/O after the end of the file. Normally fio will operate within the
1761 size of a file. If this option is set, then fio will append to the file
1762 instead. This has identical behavior to setting :option:`offset` to the size
1763 of a file. This option is ignored on non-regular files.
1764
1765.. option:: fill_device=bool, fill_fs=bool
1766
1767 Sets size to something really large and waits for ENOSPC (no space left on
1768 device) as the terminating condition. Only makes sense with sequential
1769 write. For a read workload, the mount point will be filled first then I/O
1770 started on the result. This option doesn't make sense if operating on a raw
1771 device node, since the size of that is already known by the file system.
1772 Additionally, writing beyond end-of-device will not return ENOSPC there.
1773
1774
1775I/O engine
1776~~~~~~~~~~
1777
1778.. option:: ioengine=str
1779
1780 Defines how the job issues I/O to the file. The following types are defined:
1781
1782 **sync**
1783 Basic :manpage:`read(2)` or :manpage:`write(2)`
1784 I/O. :manpage:`lseek(2)` is used to position the I/O location.
54227e6b 1785 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
f80dba8d
MT
1786
1787 **psync**
1788 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1789 all supported operating systems except for Windows.
1790
1791 **vsync**
1792 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
c60ebc45 1793 queuing by coalescing adjacent I/Os into a single submission.
f80dba8d
MT
1794
1795 **pvsync**
1796 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1797
1798 **pvsync2**
1799 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1800
1801 **libaio**
1802 Linux native asynchronous I/O. Note that Linux may only support
22413915 1803 queued behavior with non-buffered I/O (set ``direct=1`` or
f80dba8d
MT
1804 ``buffered=0``).
1805 This engine defines engine specific options.
1806
1807 **posixaio**
1808 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1809 :manpage:`aio_write(3)`.
1810
1811 **solarisaio**
1812 Solaris native asynchronous I/O.
1813
1814 **windowsaio**
1815 Windows native asynchronous I/O. Default on Windows.
1816
1817 **mmap**
1818 File is memory mapped with :manpage:`mmap(2)` and data copied
1819 to/from using :manpage:`memcpy(3)`.
1820
1821 **splice**
1822 :manpage:`splice(2)` is used to transfer the data and
1823 :manpage:`vmsplice(2)` to transfer data from user space to the
1824 kernel.
1825
1826 **sg**
1827 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1828 ioctl, or if the target is an sg character device we use
1829 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
f50fbdda 1830 I/O. Requires :option:`filename` option to specify either block or
3740cfc8 1831 character devices. This engine supports trim operations.
52b81b7c 1832 The sg engine includes engine specific options.
f80dba8d
MT
1833
1834 **null**
1835 Doesn't transfer any data, just pretends to. This is mainly used to
1836 exercise fio itself and for debugging/testing purposes.
1837
1838 **net**
1839 Transfer over the network to given ``host:port``. Depending on the
1840 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1841 :option:`listen` and :option:`filename` options are used to specify
1842 what sort of connection to make, while the :option:`protocol` option
1843 determines which protocol will be used. This engine defines engine
1844 specific options.
1845
1846 **netsplice**
1847 Like **net**, but uses :manpage:`splice(2)` and
1848 :manpage:`vmsplice(2)` to map data and send/receive.
1849 This engine defines engine specific options.
1850
1851 **cpuio**
1852 Doesn't transfer any data, but burns CPU cycles according to the
1853 :option:`cpuload` and :option:`cpuchunks` options. Setting
9207a0cb 1854 :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
71aa48eb 1855 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
f50fbdda 1856 to get desired CPU usage, as the cpuload only loads a
f80dba8d
MT
1857 single CPU at the desired rate. A job never finishes unless there is
1858 at least one non-cpuio job.
1859
1860 **guasi**
804c0839 1861 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall
f80dba8d
MT
1862 Interface approach to async I/O. See
1863
1864 http://www.xmailserver.org/guasi-lib.html
1865
1866 for more info on GUASI.
1867
1868 **rdma**
1869 The RDMA I/O engine supports both RDMA memory semantics
1870 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
609ac152
SB
1871 InfiniBand, RoCE and iWARP protocols. This engine defines engine
1872 specific options.
f80dba8d
MT
1873
1874 **falloc**
1875 I/O engine that does regular fallocate to simulate data transfer as
1876 fio ioengine.
1877
1878 DDIR_READ
1879 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1880
1881 DDIR_WRITE
1882 does fallocate(,mode = 0).
1883
1884 DDIR_TRIM
1885 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1886
761cd093
SW
1887 **ftruncate**
1888 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1889 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
f50fbdda 1890 size to the current block offset. :option:`blocksize` is ignored.
761cd093 1891
f80dba8d
MT
1892 **e4defrag**
1893 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1894 defragment activity in request to DDIR_WRITE event.
1895
f3f96717
IF
1896 **rados**
1897 I/O engine supporting direct access to Ceph Reliable Autonomic
1898 Distributed Object Store (RADOS) via librados. This ioengine
1899 defines engine specific options.
1900
f80dba8d
MT
1901 **rbd**
1902 I/O engine supporting direct access to Ceph Rados Block Devices
1903 (RBD) via librbd without the need to use the kernel rbd driver. This
1904 ioengine defines engine specific options.
1905
c2f6a13d
LMB
1906 **http**
1907 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
1908 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
1909
1910 This engine only supports direct IO of iodepth=1; you need to scale this
1911 via numjobs. blocksize defines the size of the objects to be created.
1912
1913 TRIM is translated to object deletion.
1914
f80dba8d 1915 **gfapi**
ac8ca2af
SW
1916 Using GlusterFS libgfapi sync interface to direct access to
1917 GlusterFS volumes without having to go through FUSE. This ioengine
f80dba8d
MT
1918 defines engine specific options.
1919
1920 **gfapi_async**
ac8ca2af
SW
1921 Using GlusterFS libgfapi async interface to direct access to
1922 GlusterFS volumes without having to go through FUSE. This ioengine
f80dba8d
MT
1923 defines engine specific options.
1924
1925 **libhdfs**
f50fbdda 1926 Read and write through Hadoop (HDFS). The :option:`filename` option
f80dba8d
MT
1927 is used to specify host,port of the hdfs name-node to connect. This
1928 engine interprets offsets a little differently. In HDFS, files once
e25c0c91
SW
1929 created cannot be modified so random writes are not possible. To
1930 imitate this the libhdfs engine expects a bunch of small files to be
1931 created over HDFS and will randomly pick a file from them
1932 based on the offset generated by fio backend (see the example
f80dba8d 1933 job file to create such files, use ``rw=write`` option). Please
e25c0c91
SW
1934 note, it may be necessary to set environment variables to work
1935 with HDFS/libhdfs properly. Each job uses its own connection to
f80dba8d
MT
1936 HDFS.
1937
1938 **mtd**
1939 Read, write and erase an MTD character device (e.g.,
1940 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1941 underlying device type, the I/O may have to go in a certain pattern,
1942 e.g., on NAND, writing sequentially to erase blocks and discarding
c298ee71 1943 before overwriting. The `trimwrite` mode works well for this
f80dba8d
MT
1944 constraint.
1945
1946 **pmemblk**
1947 Read and write using filesystem DAX to a file on a filesystem
363a5f65 1948 mounted with DAX on a persistent memory device through the PMDK
f80dba8d
MT
1949 libpmemblk library.
1950
1951 **dev-dax**
1952 Read and write using device DAX to a persistent memory device (e.g.,
363a5f65 1953 /dev/dax0.0) through the PMDK libpmem library.
f80dba8d
MT
1954
1955 **external**
1956 Prefix to specify loading an external I/O engine object file. Append
c60ebc45 1957 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
d243fd6d
TK
1958 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
1959 absolute or relative. See :file:`engines/skeleton_external.c` for
1960 details of writing an external I/O engine.
f80dba8d 1961
1216cc5a 1962 **filecreate**
b71968b1 1963 Simply create the files and do no I/O to them. You still need to
1216cc5a 1964 set `filesize` so that all the accounting still occurs, but no
b71968b1 1965 actual I/O will be done other than creating the file.
f80dba8d 1966
ae0db592
TI
1967 **libpmem**
1968 Read and write using mmap I/O to a file on a filesystem
363a5f65 1969 mounted with DAX on a persistent memory device through the PMDK
ae0db592
TI
1970 libpmem library.
1971
a40e7a59
GB
1972 **ime_psync**
1973 Synchronous read and write using DDN's Infinite Memory Engine (IME).
1974 This engine is very basic and issues calls to IME whenever an IO is
1975 queued.
1976
1977 **ime_psyncv**
1978 Synchronous read and write using DDN's Infinite Memory Engine (IME).
1979 This engine uses iovecs and will try to stack as much IOs as possible
1980 (if the IOs are "contiguous" and the IO depth is not exceeded)
1981 before issuing a call to IME.
1982
1983 **ime_aio**
1984 Asynchronous read and write using DDN's Infinite Memory Engine (IME).
1985 This engine will try to stack as much IOs as possible by creating
1986 requests for IME. FIO will then decide when to commit these requests.
1987
f80dba8d
MT
1988I/O engine specific parameters
1989~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1990
1991In addition, there are some parameters which are only valid when a specific
f50fbdda
TK
1992:option:`ioengine` is in use. These are used identically to normal parameters,
1993with the caveat that when used on the command line, they must come after the
f80dba8d
MT
1994:option:`ioengine` that defines them is selected.
1995
1996.. option:: userspace_reap : [libaio]
1997
1998 Normally, with the libaio engine in use, fio will use the
1999 :manpage:`io_getevents(2)` system call to reap newly returned events. With
2000 this flag turned on, the AIO ring will be read directly from user-space to
2001 reap events. The reaping mode is only enabled when polling for a minimum of
c60ebc45 2002 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
f80dba8d 2003
9d25d068 2004.. option:: hipri : [pvsync2]
f80dba8d
MT
2005
2006 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2007 than normal.
2008
a0679ce5
SB
2009.. option:: hipri_percentage : [pvsync2]
2010
f50fbdda 2011 When hipri is set this determines the probability of a pvsync2 I/O being high
a0679ce5
SB
2012 priority. The default is 100%.
2013
f80dba8d
MT
2014.. option:: cpuload=int : [cpuio]
2015
da19cdb4
TK
2016 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2017 option when using cpuio I/O engine.
f80dba8d
MT
2018
2019.. option:: cpuchunks=int : [cpuio]
2020
2021 Split the load into cycles of the given time. In microseconds.
2022
2023.. option:: exit_on_io_done=bool : [cpuio]
2024
2025 Detect when I/O threads are done, then exit.
2026
f80dba8d
MT
2027.. option:: namenode=str : [libhdfs]
2028
22413915 2029 The hostname or IP address of a HDFS cluster namenode to contact.
f80dba8d
MT
2030
2031.. option:: port=int
2032
f50fbdda
TK
2033 [libhdfs]
2034
2035 The listening port of the HFDS cluster namenode.
2036
f80dba8d
MT
2037 [netsplice], [net]
2038
2039 The TCP or UDP port to bind to or connect to. If this is used with
2040 :option:`numjobs` to spawn multiple instances of the same job type, then
2041 this will be the starting port number since fio will use a range of
2042 ports.
2043
609ac152
SB
2044 [rdma]
2045
2046 The port to use for RDMA-CM communication. This should be the same value
2047 on the client and the server side.
2048
2049.. option:: hostname=str : [netsplice] [net] [rdma]
f80dba8d 2050
609ac152
SB
2051 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
2052 is a TCP listener or UDP reader, the hostname is not used and must be omitted
f50fbdda 2053 unless it is a valid UDP multicast address.
f80dba8d
MT
2054
2055.. option:: interface=str : [netsplice] [net]
2056
2057 The IP address of the network interface used to send or receive UDP
2058 multicast.
2059
2060.. option:: ttl=int : [netsplice] [net]
2061
2062 Time-to-live value for outgoing UDP multicast packets. Default: 1.
2063
2064.. option:: nodelay=bool : [netsplice] [net]
2065
2066 Set TCP_NODELAY on TCP connections.
2067
f50fbdda 2068.. option:: protocol=str, proto=str : [netsplice] [net]
f80dba8d
MT
2069
2070 The network protocol to use. Accepted values are:
2071
2072 **tcp**
2073 Transmission control protocol.
2074 **tcpv6**
2075 Transmission control protocol V6.
2076 **udp**
2077 User datagram protocol.
2078 **udpv6**
2079 User datagram protocol V6.
2080 **unix**
2081 UNIX domain socket.
2082
2083 When the protocol is TCP or UDP, the port must also be given, as well as the
2084 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
f50fbdda 2085 normal :option:`filename` option should be used and the port is invalid.
f80dba8d 2086
e9184ec1 2087.. option:: listen : [netsplice] [net]
f80dba8d
MT
2088
2089 For TCP network connections, tell fio to listen for incoming connections
2090 rather than initiating an outgoing connection. The :option:`hostname` must
2091 be omitted if this option is used.
2092
e9184ec1 2093.. option:: pingpong : [netsplice] [net]
f80dba8d
MT
2094
2095 Normally a network writer will just continue writing data, and a network
2096 reader will just consume packages. If ``pingpong=1`` is set, a writer will
2097 send its normal payload to the reader, then wait for the reader to send the
2098 same payload back. This allows fio to measure network latencies. The
2099 submission and completion latencies then measure local time spent sending or
2100 receiving, and the completion latency measures how long it took for the
2101 other end to receive and send back. For UDP multicast traffic
2102 ``pingpong=1`` should only be set for a single reader when multiple readers
2103 are listening to the same address.
2104
e9184ec1 2105.. option:: window_size : [netsplice] [net]
f80dba8d
MT
2106
2107 Set the desired socket buffer size for the connection.
2108
e9184ec1 2109.. option:: mss : [netsplice] [net]
f80dba8d
MT
2110
2111 Set the TCP maximum segment size (TCP_MAXSEG).
2112
2113.. option:: donorname=str : [e4defrag]
2114
730bd7d9 2115 File will be used as a block donor (swap extents between files).
f80dba8d
MT
2116
2117.. option:: inplace=int : [e4defrag]
2118
2119 Configure donor file blocks allocation strategy:
2120
2121 **0**
2122 Default. Preallocate donor's file on init.
2123 **1**
2b455dbf 2124 Allocate space immediately inside defragment event, and free right
f80dba8d
MT
2125 after event.
2126
f3f96717 2127.. option:: clustername=str : [rbd,rados]
f80dba8d
MT
2128
2129 Specifies the name of the Ceph cluster.
2130
2131.. option:: rbdname=str : [rbd]
2132
2133 Specifies the name of the RBD.
2134
f3f96717 2135.. option:: pool=str : [rbd,rados]
f80dba8d 2136
f3f96717 2137 Specifies the name of the Ceph pool containing RBD or RADOS data.
f80dba8d 2138
f3f96717 2139.. option:: clientname=str : [rbd,rados]
f80dba8d
MT
2140
2141 Specifies the username (without the 'client.' prefix) used to access the
2142 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2143 the full *type.id* string. If no type. prefix is given, fio will add
2144 'client.' by default.
2145
f3f96717
IF
2146.. option:: busy_poll=bool : [rbd,rados]
2147
2148 Poll store instead of waiting for completion. Usually this provides better
2149 throughput at cost of higher(up to 100%) CPU utilization.
2150
f80dba8d
MT
2151.. option:: skip_bad=bool : [mtd]
2152
2153 Skip operations against known bad blocks.
2154
2155.. option:: hdfsdirectory : [libhdfs]
2156
2157 libhdfs will create chunk in this HDFS directory.
2158
2159.. option:: chunk_size : [libhdfs]
2160
2b455dbf 2161 The size of the chunk to use for each file.
f80dba8d 2162
609ac152
SB
2163.. option:: verb=str : [rdma]
2164
2165 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2166 values are write, read, send and recv. These correspond to the equivalent
2167 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2168 specified on the client side of the connection. See the examples folder.
2169
2170.. option:: bindname=str : [rdma]
2171
2172 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2173 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2174 will be passed into the rdma_bind_addr() function and on the client site it
2175 will be used in the rdma_resolve_add() function. This can be useful when
2176 multiple paths exist between the client and the server or in certain loopback
2177 configurations.
f80dba8d 2178
52b81b7c
KD
2179.. option:: readfua=bool : [sg]
2180
2181 With readfua option set to 1, read operations include
2182 the force unit access (fua) flag. Default is 0.
2183
2184.. option:: writefua=bool : [sg]
2185
2186 With writefua option set to 1, write operations include
2187 the force unit access (fua) flag. Default is 0.
2188
2c3a9150 2189.. option:: sg_write_mode=str : [sg]
3740cfc8 2190
2c3a9150
VF
2191 Specify the type of write commands to issue. This option can take three values:
2192
2193 **write**
2194 This is the default where write opcodes are issued as usual.
2195 **verify**
2196 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2197 directs the device to carry out a medium verification with no data
2198 comparison. The writefua option is ignored with this selection.
2199 **same**
2200 Issue WRITE SAME commands. This transfers a single block to the device
2201 and writes this same block of data to a contiguous sequence of LBAs
2202 beginning at the specified offset. fio's block size parameter specifies
2203 the amount of data written with each command. However, the amount of data
2204 actually transferred to the device is equal to the device's block
2205 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2206 write 16 sectors with each command. fio will still generate 8k of data
2207 for each command but only the first 512 bytes will be used and
2208 transferred to the device. The writefua option is ignored with this
2209 selection.
52b81b7c 2210
c2f6a13d
LMB
2211.. option:: http_host=str : [http]
2212
2213 Hostname to connect to. For S3, this could be the bucket hostname.
2214 Default is **localhost**
2215
2216.. option:: http_user=str : [http]
2217
2218 Username for HTTP authentication.
2219
2220.. option:: http_pass=str : [http]
2221
2222 Password for HTTP authentication.
2223
09fd2966 2224.. option:: https=str : [http]
c2f6a13d 2225
09fd2966
LMB
2226 Enable HTTPS instead of http. *on* enables HTTPS; *insecure*
2227 will enable HTTPS, but disable SSL peer verification (use with
2228 caution!). Default is **off**
c2f6a13d 2229
09fd2966 2230.. option:: http_mode=str : [http]
c2f6a13d 2231
09fd2966
LMB
2232 Which HTTP access mode to use: *webdav*, *swift*, or *s3*.
2233 Default is **webdav**
c2f6a13d
LMB
2234
2235.. option:: http_s3_region=str : [http]
2236
2237 The S3 region/zone string.
2238 Default is **us-east-1**
2239
2240.. option:: http_s3_key=str : [http]
2241
2242 The S3 secret key.
2243
2244.. option:: http_s3_keyid=str : [http]
2245
2246 The S3 key/access id.
2247
09fd2966
LMB
2248.. option:: http_swift_auth_token=str : [http]
2249
2250 The Swift auth token. See the example configuration file on how
2251 to retrieve this.
2252
c2f6a13d
LMB
2253.. option:: http_verbose=int : [http]
2254
2255 Enable verbose requests from libcurl. Useful for debugging. 1
2256 turns on verbose logging from libcurl, 2 additionally enables
2257 HTTP IO tracing. Default is **0**
2258
f80dba8d
MT
2259I/O depth
2260~~~~~~~~~
2261
2262.. option:: iodepth=int
2263
2264 Number of I/O units to keep in flight against the file. Note that
2265 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
c60ebc45 2266 for small degrees when :option:`verify_async` is in use). Even async
f80dba8d
MT
2267 engines may impose OS restrictions causing the desired depth not to be
2268 achieved. This may happen on Linux when using libaio and not setting
9207a0cb 2269 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
f80dba8d
MT
2270 eye on the I/O depth distribution in the fio output to verify that the
2271 achieved depth is as expected. Default: 1.
2272
2273.. option:: iodepth_batch_submit=int, iodepth_batch=int
2274
2275 This defines how many pieces of I/O to submit at once. It defaults to 1
2276 which means that we submit each I/O as soon as it is available, but can be
2277 raised to submit bigger batches of I/O at the time. If it is set to 0 the
2278 :option:`iodepth` value will be used.
2279
2280.. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
2281
2282 This defines how many pieces of I/O to retrieve at once. It defaults to 1
2283 which means that we'll ask for a minimum of 1 I/O in the retrieval process
2284 from the kernel. The I/O retrieval will go on until we hit the limit set by
2285 :option:`iodepth_low`. If this variable is set to 0, then fio will always
2286 check for completed events before queuing more I/O. This helps reduce I/O
2287 latency, at the cost of more retrieval system calls.
2288
2289.. option:: iodepth_batch_complete_max=int
2290
2291 This defines maximum pieces of I/O to retrieve at once. This variable should
9207a0cb 2292 be used along with :option:`iodepth_batch_complete_min`\=int variable,
f80dba8d 2293 specifying the range of min and max amount of I/O which should be
730bd7d9 2294 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
f80dba8d
MT
2295 value.
2296
2297 Example #1::
2298
2299 iodepth_batch_complete_min=1
2300 iodepth_batch_complete_max=<iodepth>
2301
2302 which means that we will retrieve at least 1 I/O and up to the whole
2303 submitted queue depth. If none of I/O has been completed yet, we will wait.
2304
2305 Example #2::
2306
2307 iodepth_batch_complete_min=0
2308 iodepth_batch_complete_max=<iodepth>
2309
2310 which means that we can retrieve up to the whole submitted queue depth, but
2311 if none of I/O has been completed yet, we will NOT wait and immediately exit
2312 the system call. In this example we simply do polling.
2313
2314.. option:: iodepth_low=int
2315
2316 The low water mark indicating when to start filling the queue
2317 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2318 attempt to keep the queue full at all times. If :option:`iodepth` is set to
c60ebc45 2319 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
f80dba8d
MT
2320 16 requests, it will let the depth drain down to 4 before starting to fill
2321 it again.
2322
997b5680
SW
2323.. option:: serialize_overlap=bool
2324
2325 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2326 When two or more I/Os are submitted simultaneously, there is no guarantee that
2327 the I/Os will be processed or completed in the submitted order. Further, if
2328 two or more of those I/Os are writes, any overlapping region between them can
2329 become indeterminate/undefined on certain storage. These issues can cause
2330 verification to fail erratically when at least one of the racing I/Os is
2331 changing data and the overlapping region has a non-zero size. Setting
2332 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2333 serializing in-flight I/Os that have a non-zero overlap. Note that setting
ee21ebee 2334 this option can reduce both performance and the :option:`iodepth` achieved.
997b5680
SW
2335 Additionally this option does not work when :option:`io_submit_mode` is set to
2336 offload. Default: false.
2337
f80dba8d
MT
2338.. option:: io_submit_mode=str
2339
2340 This option controls how fio submits the I/O to the I/O engine. The default
2341 is `inline`, which means that the fio job threads submit and reap I/O
2342 directly. If set to `offload`, the job threads will offload I/O submission
2343 to a dedicated pool of I/O threads. This requires some coordination and thus
2344 has a bit of extra overhead, especially for lower queue depth I/O where it
2345 can increase latencies. The benefit is that fio can manage submission rates
2346 independently of the device completion rates. This avoids skewed latency
730bd7d9 2347 reporting if I/O gets backed up on the device side (the coordinated omission
f80dba8d
MT
2348 problem).
2349
2350
2351I/O rate
2352~~~~~~~~
2353
a881438b 2354.. option:: thinktime=time
f80dba8d 2355
f75ede1d
SW
2356 Stall the job for the specified period of time after an I/O has completed before issuing the
2357 next. May be used to simulate processing being done by an application.
947e0fe0 2358 When the unit is omitted, the value is interpreted in microseconds. See
f80dba8d
MT
2359 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2360
a881438b 2361.. option:: thinktime_spin=time
f80dba8d
MT
2362
2363 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2364 something with the data received, before falling back to sleeping for the
f75ede1d 2365 rest of the period specified by :option:`thinktime`. When the unit is
947e0fe0 2366 omitted, the value is interpreted in microseconds.
f80dba8d
MT
2367
2368.. option:: thinktime_blocks=int
2369
2370 Only valid if :option:`thinktime` is set - control how many blocks to issue,
f50fbdda
TK
2371 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
2372 fio wait :option:`thinktime` usecs after every block. This effectively makes any
f80dba8d 2373 queue depth setting redundant, since no more than 1 I/O will be queued
f50fbdda 2374 before we have to complete it and do our :option:`thinktime`. In other words, this
f80dba8d 2375 setting effectively caps the queue depth if the latter is larger.
71bfa161 2376
f80dba8d 2377.. option:: rate=int[,int][,int]
71bfa161 2378
f80dba8d
MT
2379 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2380 suffix rules apply. Comma-separated values may be specified for reads,
2381 writes, and trims as described in :option:`blocksize`.
71bfa161 2382
b25b3464
SW
2383 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2384 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2385 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2386 latter will only limit reads.
2387
f80dba8d 2388.. option:: rate_min=int[,int][,int]
71bfa161 2389
f80dba8d
MT
2390 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2391 to meet this requirement will cause the job to exit. Comma-separated values
2392 may be specified for reads, writes, and trims as described in
2393 :option:`blocksize`.
71bfa161 2394
f80dba8d 2395.. option:: rate_iops=int[,int][,int]
71bfa161 2396
f80dba8d
MT
2397 Cap the bandwidth to this number of IOPS. Basically the same as
2398 :option:`rate`, just specified independently of bandwidth. If the job is
2399 given a block size range instead of a fixed value, the smallest block size
2400 is used as the metric. Comma-separated values may be specified for reads,
2401 writes, and trims as described in :option:`blocksize`.
71bfa161 2402
f80dba8d 2403.. option:: rate_iops_min=int[,int][,int]
71bfa161 2404
f80dba8d
MT
2405 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2406 Comma-separated values may be specified for reads, writes, and trims as
2407 described in :option:`blocksize`.
71bfa161 2408
f80dba8d 2409.. option:: rate_process=str
66c098b8 2410
f80dba8d
MT
2411 This option controls how fio manages rated I/O submissions. The default is
2412 `linear`, which submits I/O in a linear fashion with fixed delays between
c60ebc45 2413 I/Os that gets adjusted based on I/O completion rates. If this is set to
f80dba8d
MT
2414 `poisson`, fio will submit I/O based on a more real world random request
2415 flow, known as the Poisson process
2416 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2417 10^6 / IOPS for the given workload.
71bfa161 2418
1a9bf814
JA
2419.. option:: rate_ignore_thinktime=bool
2420
2421 By default, fio will attempt to catch up to the specified rate setting,
2422 if any kind of thinktime setting was used. If this option is set, then
2423 fio will ignore the thinktime and continue doing IO at the specified
2424 rate, instead of entering a catch-up mode after thinktime is done.
2425
71bfa161 2426
f80dba8d
MT
2427I/O latency
2428~~~~~~~~~~~
71bfa161 2429
a881438b 2430.. option:: latency_target=time
71bfa161 2431
f80dba8d 2432 If set, fio will attempt to find the max performance point that the given
f75ede1d 2433 workload will run at while maintaining a latency below this target. When
947e0fe0 2434 the unit is omitted, the value is interpreted in microseconds. See
f75ede1d 2435 :option:`latency_window` and :option:`latency_percentile`.
71bfa161 2436
a881438b 2437.. option:: latency_window=time
71bfa161 2438
f80dba8d 2439 Used with :option:`latency_target` to specify the sample window that the job
f75ede1d 2440 is run at varying queue depths to test the performance. When the unit is
947e0fe0 2441 omitted, the value is interpreted in microseconds.
b4692828 2442
f80dba8d 2443.. option:: latency_percentile=float
71bfa161 2444
c60ebc45 2445 The percentage of I/Os that must fall within the criteria specified by
f80dba8d 2446 :option:`latency_target` and :option:`latency_window`. If not set, this
c60ebc45 2447 defaults to 100.0, meaning that all I/Os must be equal or below to the value
f80dba8d 2448 set by :option:`latency_target`.
71bfa161 2449
a881438b 2450.. option:: max_latency=time
71bfa161 2451
f75ede1d 2452 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
947e0fe0 2453 maximum latency. When the unit is omitted, the value is interpreted in
f75ede1d 2454 microseconds.
71bfa161 2455
f80dba8d 2456.. option:: rate_cycle=int
71bfa161 2457
f80dba8d 2458 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
a47b697c 2459 of milliseconds. Defaults to 1000.
71bfa161 2460
71bfa161 2461
f80dba8d
MT
2462I/O replay
2463~~~~~~~~~~
71bfa161 2464
f80dba8d 2465.. option:: write_iolog=str
c2b1e753 2466
f80dba8d
MT
2467 Write the issued I/O patterns to the specified file. See
2468 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2469 iologs will be interspersed and the file may be corrupt.
c2b1e753 2470
f80dba8d 2471.. option:: read_iolog=str
71bfa161 2472
22413915 2473 Open an iolog with the specified filename and replay the I/O patterns it
f80dba8d
MT
2474 contains. This can be used to store a workload and replay it sometime
2475 later. The iolog given may also be a blktrace binary file, which allows fio
2476 to replay a workload captured by :command:`blktrace`. See
2477 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2478 replay, the file needs to be turned into a blkparse binary data file first
2479 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
78439a18
JA
2480 You can specify a number of files by separating the names with a ':'
2481 character. See the :option:`filename` option for information on how to
2482 escape ':' and '\' characters within the file names. These files will
2483 be sequentially assigned to job clones created by :option:`numjobs`.
71bfa161 2484
77be374d
AK
2485.. option:: read_iolog_chunked=bool
2486
2487 Determines how iolog is read. If false(default) entire :option:`read_iolog`
2488 will be read at once. If selected true, input from iolog will be read
2489 gradually. Useful when iolog is very large, or it is generated.
2490
589e88b7 2491.. option:: replay_no_stall=bool
71bfa161 2492
f80dba8d 2493 When replaying I/O with :option:`read_iolog` the default behavior is to
22413915 2494 attempt to respect the timestamps within the log and replay them with the
f80dba8d
MT
2495 appropriate delay between IOPS. By setting this variable fio will not
2496 respect the timestamps and attempt to replay them as fast as possible while
2497 still respecting ordering. The result is the same I/O pattern to a given
2498 device, but different timings.
71bfa161 2499
6dd7fa77
JA
2500.. option:: replay_time_scale=int
2501
2502 When replaying I/O with :option:`read_iolog`, fio will honor the
2503 original timing in the trace. With this option, it's possible to scale
2504 the time. It's a percentage option, if set to 50 it means run at 50%
2505 the original IO rate in the trace. If set to 200, run at twice the
2506 original IO rate. Defaults to 100.
2507
f80dba8d 2508.. option:: replay_redirect=str
b4692828 2509
f80dba8d
MT
2510 While replaying I/O patterns using :option:`read_iolog` the default behavior
2511 is to replay the IOPS onto the major/minor device that each IOP was recorded
2512 from. This is sometimes undesirable because on a different machine those
2513 major/minor numbers can map to a different device. Changing hardware on the
2514 same system can also result in a different major/minor mapping.
730bd7d9 2515 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
f80dba8d 2516 device regardless of the device it was recorded
9207a0cb 2517 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
f80dba8d
MT
2518 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2519 multiple devices will be replayed onto a single device, if the trace
2520 contains multiple devices. If you want multiple devices to be replayed
2521 concurrently to multiple redirected devices you must blkparse your trace
2522 into separate traces and replay them with independent fio invocations.
2523 Unfortunately this also breaks the strict time ordering between multiple
2524 device accesses.
71bfa161 2525
f80dba8d 2526.. option:: replay_align=int
74929ac2 2527
f80dba8d
MT
2528 Force alignment of I/O offsets and lengths in a trace to this power of 2
2529 value.
3c54bc46 2530
f80dba8d 2531.. option:: replay_scale=int
3c54bc46 2532
f80dba8d 2533 Scale sector offsets down by this factor when replaying traces.
3c54bc46 2534
38f68906
JA
2535.. option:: replay_skip=str
2536
2537 Sometimes it's useful to skip certain IO types in a replay trace.
2538 This could be, for instance, eliminating the writes in the trace.
2539 Or not replaying the trims/discards, if you are redirecting to
2540 a device that doesn't support them. This option takes a comma
2541 separated list of read, write, trim, sync.
2542
3c54bc46 2543
f80dba8d
MT
2544Threads, processes and job synchronization
2545~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3c54bc46 2546
f80dba8d 2547.. option:: thread
3c54bc46 2548
730bd7d9
SW
2549 Fio defaults to creating jobs by using fork, however if this option is
2550 given, fio will create jobs by using POSIX Threads' function
2551 :manpage:`pthread_create(3)` to create threads instead.
71bfa161 2552
f80dba8d 2553.. option:: wait_for=str
74929ac2 2554
730bd7d9
SW
2555 If set, the current job won't be started until all workers of the specified
2556 waitee job are done.
74929ac2 2557
f80dba8d
MT
2558 ``wait_for`` operates on the job name basis, so there are a few
2559 limitations. First, the waitee must be defined prior to the waiter job
2560 (meaning no forward references). Second, if a job is being referenced as a
2561 waitee, it must have a unique name (no duplicate waitees).
74929ac2 2562
f80dba8d 2563.. option:: nice=int
892a6ffc 2564
f80dba8d 2565 Run the job with the given nice value. See man :manpage:`nice(2)`.
892a6ffc 2566
f80dba8d
MT
2567 On Windows, values less than -15 set the process class to "High"; -1 through
2568 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2569 priority class.
74929ac2 2570
f80dba8d 2571.. option:: prio=int
71bfa161 2572
f80dba8d
MT
2573 Set the I/O priority value of this job. Linux limits us to a positive value
2574 between 0 and 7, with 0 being the highest. See man
2575 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2576 systems since meaning of priority may differ.
71bfa161 2577
f80dba8d 2578.. option:: prioclass=int
d59aa780 2579
f80dba8d 2580 Set the I/O priority class. See man :manpage:`ionice(1)`.
d59aa780 2581
f80dba8d 2582.. option:: cpus_allowed=str
6d500c2e 2583
730bd7d9 2584 Controls the same options as :option:`cpumask`, but accepts a textual
b570e037
SW
2585 specification of the permitted CPUs instead and CPUs are indexed from 0. So
2586 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
2587 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
2588 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
2589
2590 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
2591 processor group will be used and affinity settings are inherited from the
2592 system. An fio build configured to target Windows 7 makes options that set
2593 CPUs processor group aware and values will set both the processor group
2594 and a CPU from within that group. For example, on a system where processor
2595 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
2596 values between 0 and 39 will bind CPUs from processor group 0 and
2597 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
2598 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
2599 single ``cpus_allowed`` option must be from the same processor group. For
2600 Windows fio builds not built for Windows 7, CPUs will only be selected from
2601 (and be relative to) whatever processor group fio happens to be running in
2602 and CPUs from other processor groups cannot be used.
6d500c2e 2603
f80dba8d 2604.. option:: cpus_allowed_policy=str
6d500c2e 2605
f80dba8d 2606 Set the policy of how fio distributes the CPUs specified by
730bd7d9 2607 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
6d500c2e 2608
f80dba8d
MT
2609 **shared**
2610 All jobs will share the CPU set specified.
2611 **split**
2612 Each job will get a unique CPU from the CPU set.
6d500c2e 2613
22413915 2614 **shared** is the default behavior, if the option isn't specified. If
f80dba8d
MT
2615 **split** is specified, then fio will will assign one cpu per job. If not
2616 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2617 in the set.
6d500c2e 2618
b570e037
SW
2619.. option:: cpumask=int
2620
2621 Set the CPU affinity of this job. The parameter given is a bit mask of
2622 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2623 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2624 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2625 operating systems or kernel versions. This option doesn't work well for a
2626 higher CPU count than what you can store in an integer mask, so it can only
2627 control cpus 1-32. For boxes with larger CPU counts, use
2628 :option:`cpus_allowed`.
2629
f80dba8d 2630.. option:: numa_cpu_nodes=str
6d500c2e 2631
f80dba8d
MT
2632 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2633 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
ac8ca2af 2634 NUMA options support, fio must be built on a system with libnuma-dev(el)
f80dba8d 2635 installed.
61b9861d 2636
f80dba8d 2637.. option:: numa_mem_policy=str
61b9861d 2638
f80dba8d
MT
2639 Set this job's memory policy and corresponding NUMA nodes. Format of the
2640 arguments::
5c94b008 2641
f80dba8d 2642 <mode>[:<nodelist>]
ce35b1ec 2643
804c0839 2644 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
730bd7d9
SW
2645 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2646 policies, no node needs to be specified. For ``prefer``, only one node is
2647 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2648 follows: a comma delimited list of numbers, A-B ranges, or `all`.
71bfa161 2649
f80dba8d 2650.. option:: cgroup=str
390b1537 2651
f80dba8d
MT
2652 Add job to this control group. If it doesn't exist, it will be created. The
2653 system must have a mounted cgroup blkio mount point for this to work. If
2654 your system doesn't have it mounted, you can do so with::
5af1c6f3 2655
f80dba8d 2656 # mount -t cgroup -o blkio none /cgroup
5af1c6f3 2657
f80dba8d 2658.. option:: cgroup_weight=int
5af1c6f3 2659
f80dba8d
MT
2660 Set the weight of the cgroup to this value. See the documentation that comes
2661 with the kernel, allowed values are in the range of 100..1000.
a086c257 2662
f80dba8d 2663.. option:: cgroup_nodelete=bool
8c07860d 2664
f80dba8d
MT
2665 Normally fio will delete the cgroups it has created after the job
2666 completion. To override this behavior and to leave cgroups around after the
2667 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2668 to inspect various cgroup files after job completion. Default: false.
8c07860d 2669
f80dba8d 2670.. option:: flow_id=int
8c07860d 2671
f80dba8d
MT
2672 The ID of the flow. If not specified, it defaults to being a global
2673 flow. See :option:`flow`.
1907dbc6 2674
f80dba8d 2675.. option:: flow=int
71bfa161 2676
f80dba8d
MT
2677 Weight in token-based flow control. If this value is used, then there is a
2678 'flow counter' which is used to regulate the proportion of activity between
2679 two or more jobs. Fio attempts to keep this flow counter near zero. The
2680 ``flow`` parameter stands for how much should be added or subtracted to the
2681 flow counter on each iteration of the main I/O loop. That is, if one job has
2682 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2683 ratio in how much one runs vs the other.
71bfa161 2684
f80dba8d 2685.. option:: flow_watermark=int
a31041ea 2686
f80dba8d
MT
2687 The maximum value that the absolute value of the flow counter is allowed to
2688 reach before the job must wait for a lower value of the counter.
82407585 2689
f80dba8d 2690.. option:: flow_sleep=int
82407585 2691
f80dba8d
MT
2692 The period of time, in microseconds, to wait after the flow watermark has
2693 been exceeded before retrying operations.
82407585 2694
f80dba8d 2695.. option:: stonewall, wait_for_previous
82407585 2696
f80dba8d
MT
2697 Wait for preceding jobs in the job file to exit, before starting this
2698 one. Can be used to insert serialization points in the job file. A stone
2699 wall also implies starting a new reporting group, see
2700 :option:`group_reporting`.
2701
2702.. option:: exitall
2703
730bd7d9
SW
2704 By default, fio will continue running all other jobs when one job finishes
2705 but sometimes this is not the desired action. Setting ``exitall`` will
2706 instead make fio terminate all other jobs when one job finishes.
f80dba8d
MT
2707
2708.. option:: exec_prerun=str
2709
2710 Before running this job, issue the command specified through
2711 :manpage:`system(3)`. Output is redirected in a file called
2712 :file:`jobname.prerun.txt`.
2713
2714.. option:: exec_postrun=str
2715
2716 After the job completes, issue the command specified though
2717 :manpage:`system(3)`. Output is redirected in a file called
2718 :file:`jobname.postrun.txt`.
2719
2720.. option:: uid=int
2721
2722 Instead of running as the invoking user, set the user ID to this value
2723 before the thread/process does any work.
2724
2725.. option:: gid=int
2726
2727 Set group ID, see :option:`uid`.
2728
2729
2730Verification
2731~~~~~~~~~~~~
2732
2733.. option:: verify_only
2734
2735 Do not perform specified workload, only verify data still matches previous
2736 invocation of this workload. This option allows one to check data multiple
2737 times at a later date without overwriting it. This option makes sense only
2738 for workloads that write data, and does not support workloads with the
2739 :option:`time_based` option set.
2740
2741.. option:: do_verify=bool
2742
2743 Run the verify phase after a write phase. Only valid if :option:`verify` is
2744 set. Default: true.
2745
2746.. option:: verify=str
2747
2748 If writing to a file, fio can verify the file contents after each iteration
2749 of the job. Each verification method also implies verification of special
2750 header, which is written to the beginning of each block. This header also
2751 includes meta information, like offset of the block, block number, timestamp
2752 when block was written, etc. :option:`verify` can be combined with
2753 :option:`verify_pattern` option. The allowed values are:
2754
2755 **md5**
2756 Use an md5 sum of the data area and store it in the header of
2757 each block.
2758
2759 **crc64**
2760 Use an experimental crc64 sum of the data area and store it in the
2761 header of each block.
2762
2763 **crc32c**
a5896300
SW
2764 Use a crc32c sum of the data area and store it in the header of
2765 each block. This will automatically use hardware acceleration
2766 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
2767 fall back to software crc32c if none is found. Generally the
804c0839 2768 fastest checksum fio supports when hardware accelerated.
f80dba8d
MT
2769
2770 **crc32c-intel**
a5896300 2771 Synonym for crc32c.
f80dba8d
MT
2772
2773 **crc32**
2774 Use a crc32 sum of the data area and store it in the header of each
2775 block.
2776
2777 **crc16**
2778 Use a crc16 sum of the data area and store it in the header of each
2779 block.
2780
2781 **crc7**
2782 Use a crc7 sum of the data area and store it in the header of each
2783 block.
2784
2785 **xxhash**
2786 Use xxhash as the checksum function. Generally the fastest software
2787 checksum that fio supports.
2788
2789 **sha512**
2790 Use sha512 as the checksum function.
2791
2792 **sha256**
2793 Use sha256 as the checksum function.
2794
2795 **sha1**
2796 Use optimized sha1 as the checksum function.
82407585 2797
ae3a5acc
JA
2798 **sha3-224**
2799 Use optimized sha3-224 as the checksum function.
2800
2801 **sha3-256**
2802 Use optimized sha3-256 as the checksum function.
2803
2804 **sha3-384**
2805 Use optimized sha3-384 as the checksum function.
2806
2807 **sha3-512**
2808 Use optimized sha3-512 as the checksum function.
2809
f80dba8d
MT
2810 **meta**
2811 This option is deprecated, since now meta information is included in
2812 generic verification header and meta verification happens by
2813 default. For detailed information see the description of the
2814 :option:`verify` setting. This option is kept because of
2815 compatibility's sake with old configurations. Do not use it.
2816
2817 **pattern**
2818 Verify a strict pattern. Normally fio includes a header with some
2819 basic information and checksumming, but if this option is set, only
2820 the specific pattern set with :option:`verify_pattern` is verified.
2821
2822 **null**
2823 Only pretend to verify. Useful for testing internals with
9207a0cb 2824 :option:`ioengine`\=null, not for much else.
f80dba8d
MT
2825
2826 This option can be used for repeated burn-in tests of a system to make sure
2827 that the written data is also correctly read back. If the data direction
2828 given is a read or random read, fio will assume that it should verify a
2829 previously written file. If the data direction includes any form of write,
2830 the verify will be of the newly written data.
2831
47e6a6e5
SW
2832 To avoid false verification errors, do not use the norandommap option when
2833 verifying data with async I/O engines and I/O depths > 1. Or use the
2834 norandommap and the lfsr random generator together to avoid writing to the
2835 same offset with muliple outstanding I/Os.
2836
f80dba8d
MT
2837.. option:: verify_offset=int
2838
2839 Swap the verification header with data somewhere else in the block before
2840 writing. It is swapped back before verifying.
2841
2842.. option:: verify_interval=int
2843
2844 Write the verification header at a finer granularity than the
2845 :option:`blocksize`. It will be written for chunks the size of
2846 ``verify_interval``. :option:`blocksize` should divide this evenly.
2847
2848.. option:: verify_pattern=str
2849
2850 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2851 filling with totally random bytes, but sometimes it's interesting to fill
2852 with a known pattern for I/O verification purposes. Depending on the width
730bd7d9 2853 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
f80dba8d
MT
2854 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2855 a 32-bit quantity has to be a hex number that starts with either "0x" or
2856 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2857 format, which means that for each block offset will be written and then
2858 verified back, e.g.::
61b9861d
RP
2859
2860 verify_pattern=%o
2861
f80dba8d
MT
2862 Or use combination of everything::
2863
61b9861d 2864 verify_pattern=0xff%o"abcd"-12
e28218f3 2865
f80dba8d
MT
2866.. option:: verify_fatal=bool
2867
2868 Normally fio will keep checking the entire contents before quitting on a
2869 block verification failure. If this option is set, fio will exit the job on
2870 the first observed failure. Default: false.
2871
2872.. option:: verify_dump=bool
2873
2874 If set, dump the contents of both the original data block and the data block
2875 we read off disk to files. This allows later analysis to inspect just what
2876 kind of data corruption occurred. Off by default.
2877
2878.. option:: verify_async=int
2879
2880 Fio will normally verify I/O inline from the submitting thread. This option
2881 takes an integer describing how many async offload threads to create for I/O
2882 verification instead, causing fio to offload the duty of verifying I/O
2883 contents to one or more separate threads. If using this offload option, even
2884 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2885 than 1, as it allows them to have I/O in flight while verifies are running.
d7e6ea1c 2886 Defaults to 0 async threads, i.e. verification is not asynchronous.
f80dba8d
MT
2887
2888.. option:: verify_async_cpus=str
2889
2890 Tell fio to set the given CPU affinity on the async I/O verification
2891 threads. See :option:`cpus_allowed` for the format used.
2892
2893.. option:: verify_backlog=int
2894
2895 Fio will normally verify the written contents of a job that utilizes verify
2896 once that job has completed. In other words, everything is written then
2897 everything is read back and verified. You may want to verify continually
2898 instead for a variety of reasons. Fio stores the meta data associated with
2899 an I/O block in memory, so for large verify workloads, quite a bit of memory
2900 would be used up holding this meta data. If this option is enabled, fio will
2901 write only N blocks before verifying these blocks.
2902
2903.. option:: verify_backlog_batch=int
2904
2905 Control how many blocks fio will verify if :option:`verify_backlog` is
2906 set. If not set, will default to the value of :option:`verify_backlog`
2907 (meaning the entire queue is read back and verified). If
2908 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2909 blocks will be verified, if ``verify_backlog_batch`` is larger than
2910 :option:`verify_backlog`, some blocks will be verified more than once.
2911
2912.. option:: verify_state_save=bool
2913
2914 When a job exits during the write phase of a verify workload, save its
2915 current state. This allows fio to replay up until that point, if the verify
2916 state is loaded for the verify read phase. The format of the filename is,
2917 roughly::
2918
f50fbdda 2919 <type>-<jobname>-<jobindex>-verify.state.
f80dba8d
MT
2920
2921 <type> is "local" for a local run, "sock" for a client/server socket
2922 connection, and "ip" (192.168.0.1, for instance) for a networked
d7e6ea1c 2923 client/server connection. Defaults to true.
f80dba8d
MT
2924
2925.. option:: verify_state_load=bool
2926
2927 If a verify termination trigger was used, fio stores the current write state
2928 of each thread. This can be used at verification time so that fio knows how
2929 far it should verify. Without this information, fio will run a full
a47b697c
SW
2930 verification pass, according to the settings in the job file used. Default
2931 false.
f80dba8d
MT
2932
2933.. option:: trim_percentage=int
2934
2935 Number of verify blocks to discard/trim.
2936
2937.. option:: trim_verify_zero=bool
2938
22413915 2939 Verify that trim/discarded blocks are returned as zeros.
f80dba8d
MT
2940
2941.. option:: trim_backlog=int
2942
5cfd1e9a 2943 Trim after this number of blocks are written.
f80dba8d
MT
2944
2945.. option:: trim_backlog_batch=int
2946
2947 Trim this number of I/O blocks.
2948
2949.. option:: experimental_verify=bool
2950
2951 Enable experimental verification.
2952
f80dba8d
MT
2953Steady state
2954~~~~~~~~~~~~
2955
2956.. option:: steadystate=str:float, ss=str:float
2957
2958 Define the criterion and limit for assessing steady state performance. The
2959 first parameter designates the criterion whereas the second parameter sets
2960 the threshold. When the criterion falls below the threshold for the
2961 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2962 direct fio to terminate the job when the least squares regression slope
2963 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2964 this will apply to all jobs in the group. Below is the list of available
2965 steady state assessment criteria. All assessments are carried out using only
2966 data from the rolling collection window. Threshold limits can be expressed
2967 as a fixed value or as a percentage of the mean in the collection window.
2968
2969 **iops**
2970 Collect IOPS data. Stop the job if all individual IOPS measurements
2971 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2972 means that all individual IOPS values must be within 2 of the mean,
2973 whereas ``iops:0.2%`` means that all individual IOPS values must be
2974 within 0.2% of the mean IOPS to terminate the job).
2975
2976 **iops_slope**
2977 Collect IOPS data and calculate the least squares regression
2978 slope. Stop the job if the slope falls below the specified limit.
2979
2980 **bw**
2981 Collect bandwidth data. Stop the job if all individual bandwidth
2982 measurements are within the specified limit of the mean bandwidth.
2983
2984 **bw_slope**
2985 Collect bandwidth data and calculate the least squares regression
2986 slope. Stop the job if the slope falls below the specified limit.
2987
2988.. option:: steadystate_duration=time, ss_dur=time
2989
2990 A rolling window of this duration will be used to judge whether steady state
2991 has been reached. Data will be collected once per second. The default is 0
f75ede1d 2992 which disables steady state detection. When the unit is omitted, the
947e0fe0 2993 value is interpreted in seconds.
f80dba8d
MT
2994
2995.. option:: steadystate_ramp_time=time, ss_ramp=time
2996
2997 Allow the job to run for the specified duration before beginning data
2998 collection for checking the steady state job termination criterion. The
947e0fe0 2999 default is 0. When the unit is omitted, the value is interpreted in seconds.
f80dba8d
MT
3000
3001
3002Measurements and reporting
3003~~~~~~~~~~~~~~~~~~~~~~~~~~
3004
3005.. option:: per_job_logs=bool
3006
3007 If set, this generates bw/clat/iops log with per file private filenames. If
3008 not set, jobs with identical names will share the log filename. Default:
3009 true.
3010
3011.. option:: group_reporting
3012
3013 It may sometimes be interesting to display statistics for groups of jobs as
3014 a whole instead of for each individual job. This is especially true if
3015 :option:`numjobs` is used; looking at individual thread/process output
3016 quickly becomes unwieldy. To see the final report per-group instead of
3017 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
3018 same reporting group, unless if separated by a :option:`stonewall`, or by
3019 using :option:`new_group`.
3020
3021.. option:: new_group
3022
3023 Start a new reporting group. See: :option:`group_reporting`. If not given,
3024 all jobs in a file will be part of the same reporting group, unless
3025 separated by a :option:`stonewall`.
3026
589e88b7 3027.. option:: stats=bool
8243be59
JA
3028
3029 By default, fio collects and shows final output results for all jobs
3030 that run. If this option is set to 0, then fio will ignore it in
3031 the final stat output.
3032
f80dba8d
MT
3033.. option:: write_bw_log=str
3034
3035 If given, write a bandwidth log for this job. Can be used to store data of
074f0817 3036 the bandwidth of the jobs in their lifetime.
f80dba8d 3037
074f0817
SW
3038 If no str argument is given, the default filename of
3039 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
3040 will still append the type of log. So if one specifies::
3041
3042 write_bw_log=foo
f80dba8d 3043
074f0817
SW
3044 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
3045 of the job (`1..N`, where `N` is the number of jobs). If
3046 :option:`per_job_logs` is false, then the filename will not include the
3047 `.x` job index.
e3cedca7 3048
074f0817
SW
3049 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
3050 text files into nice graphs. See `Log File Formats`_ for how data is
3051 structured within the file.
3052
3053.. option:: write_lat_log=str
e3cedca7 3054
074f0817 3055 Same as :option:`write_bw_log`, except this option creates I/O
77b7e675
SW
3056 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
3057 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
074f0817
SW
3058 latency files instead. See :option:`write_bw_log` for details about
3059 the filename format and `Log File Formats`_ for how data is structured
3060 within the files.
be4ecfdf 3061
f80dba8d 3062.. option:: write_hist_log=str
06842027 3063
074f0817 3064 Same as :option:`write_bw_log` but writes an I/O completion latency
77b7e675 3065 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
074f0817
SW
3066 file will be empty unless :option:`log_hist_msec` has also been set.
3067 See :option:`write_bw_log` for details about the filename format and
3068 `Log File Formats`_ for how data is structured within the file.
06842027 3069
f80dba8d 3070.. option:: write_iops_log=str
06842027 3071
074f0817 3072 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
15417073
SW
3073 :file:`name_iops.x.log`) instead. Because fio defaults to individual
3074 I/O logging, the value entry in the IOPS log will be 1 unless windowed
3075 logging (see :option:`log_avg_msec`) has been enabled. See
3076 :option:`write_bw_log` for details about the filename format and `Log
3077 File Formats`_ for how data is structured within the file.
06842027 3078
f80dba8d 3079.. option:: log_avg_msec=int
06842027 3080
f80dba8d
MT
3081 By default, fio will log an entry in the iops, latency, or bw log for every
3082 I/O that completes. When writing to the disk log, that can quickly grow to a
3083 very large size. Setting this option makes fio average the each log entry
3084 over the specified period of time, reducing the resolution of the log. See
3085 :option:`log_max_value` as well. Defaults to 0, logging all entries.
6fc82095 3086 Also see `Log File Formats`_.
06842027 3087
f80dba8d 3088.. option:: log_hist_msec=int
06842027 3089
f80dba8d
MT
3090 Same as :option:`log_avg_msec`, but logs entries for completion latency
3091 histograms. Computing latency percentiles from averages of intervals using
c60ebc45 3092 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
f80dba8d
MT
3093 histogram entries over the specified period of time, reducing log sizes for
3094 high IOPS devices while retaining percentile accuracy. See
074f0817
SW
3095 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
3096 Defaults to 0, meaning histogram logging is disabled.
06842027 3097
f80dba8d 3098.. option:: log_hist_coarseness=int
06842027 3099
f80dba8d
MT
3100 Integer ranging from 0 to 6, defining the coarseness of the resolution of
3101 the histogram logs enabled with :option:`log_hist_msec`. For each increment
3102 in coarseness, fio outputs half as many bins. Defaults to 0, for which
074f0817
SW
3103 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
3104 and `Log File Formats`_.
8b28bd41 3105
f80dba8d 3106.. option:: log_max_value=bool
66c098b8 3107
f80dba8d
MT
3108 If :option:`log_avg_msec` is set, fio logs the average over that window. If
3109 you instead want to log the maximum value, set this option to 1. Defaults to
3110 0, meaning that averaged values are logged.
a696fa2a 3111
589e88b7 3112.. option:: log_offset=bool
a696fa2a 3113
f80dba8d 3114 If this is set, the iolog options will include the byte offset for the I/O
5a83478f
SW
3115 entry as well as the other data values. Defaults to 0 meaning that
3116 offsets are not present in logs. Also see `Log File Formats`_.
71bfa161 3117
f80dba8d 3118.. option:: log_compression=int
7de87099 3119
f80dba8d
MT
3120 If this is set, fio will compress the I/O logs as it goes, to keep the
3121 memory footprint lower. When a log reaches the specified size, that chunk is
3122 removed and compressed in the background. Given that I/O logs are fairly
3123 highly compressible, this yields a nice memory savings for longer runs. The
3124 downside is that the compression will consume some background CPU cycles, so
3125 it may impact the run. This, however, is also true if the logging ends up
3126 consuming most of the system memory. So pick your poison. The I/O logs are
3127 saved normally at the end of a run, by decompressing the chunks and storing
3128 them in the specified log file. This feature depends on the availability of
3129 zlib.
e0b0d892 3130
f80dba8d 3131.. option:: log_compression_cpus=str
e0b0d892 3132
f80dba8d
MT
3133 Define the set of CPUs that are allowed to handle online log compression for
3134 the I/O jobs. This can provide better isolation between performance
0cf90a62
SW
3135 sensitive jobs, and background compression work. See
3136 :option:`cpus_allowed` for the format used.
9e684a49 3137
f80dba8d 3138.. option:: log_store_compressed=bool
9e684a49 3139
f80dba8d
MT
3140 If set, fio will store the log files in a compressed format. They can be
3141 decompressed with fio, using the :option:`--inflate-log` command line
3142 parameter. The files will be stored with a :file:`.fz` suffix.
9e684a49 3143
f80dba8d 3144.. option:: log_unix_epoch=bool
9e684a49 3145
f80dba8d
MT
3146 If set, fio will log Unix timestamps to the log files produced by enabling
3147 write_type_log for each log type, instead of the default zero-based
3148 timestamps.
3149
3150.. option:: block_error_percentiles=bool
3151
3152 If set, record errors in trim block-sized units from writes and trims and
3153 output a histogram of how many trims it took to get to errors, and what kind
3154 of error was encountered.
3155
3156.. option:: bwavgtime=int
3157
3158 Average the calculated bandwidth over the given time. Value is specified in
3159 milliseconds. If the job also does bandwidth logging through
3160 :option:`write_bw_log`, then the minimum of this option and
3161 :option:`log_avg_msec` will be used. Default: 500ms.
3162
3163.. option:: iopsavgtime=int
3164
3165 Average the calculated IOPS over the given time. Value is specified in
3166 milliseconds. If the job also does IOPS logging through
3167 :option:`write_iops_log`, then the minimum of this option and
3168 :option:`log_avg_msec` will be used. Default: 500ms.
3169
3170.. option:: disk_util=bool
3171
3172 Generate disk utilization statistics, if the platform supports it.
3173 Default: true.
3174
3175.. option:: disable_lat=bool
3176
3177 Disable measurements of total latency numbers. Useful only for cutting back
3178 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
3179 performance at really high IOPS rates. Note that to really get rid of a
3180 large amount of these calls, this option must be used with
f75ede1d 3181 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
f80dba8d
MT
3182
3183.. option:: disable_clat=bool
3184
3185 Disable measurements of completion latency numbers. See
3186 :option:`disable_lat`.
3187
3188.. option:: disable_slat=bool
3189
3190 Disable measurements of submission latency numbers. See
f50fbdda 3191 :option:`disable_lat`.
f80dba8d 3192
f75ede1d 3193.. option:: disable_bw_measurement=bool, disable_bw=bool
f80dba8d
MT
3194
3195 Disable measurements of throughput/bandwidth numbers. See
3196 :option:`disable_lat`.
3197
3198.. option:: clat_percentiles=bool
3199
b599759b
JA
3200 Enable the reporting of percentiles of completion latencies. This
3201 option is mutually exclusive with :option:`lat_percentiles`.
3202
3203.. option:: lat_percentiles=bool
3204
b71968b1 3205 Enable the reporting of percentiles of I/O latencies. This is similar
b599759b
JA
3206 to :option:`clat_percentiles`, except that this includes the
3207 submission latency. This option is mutually exclusive with
3208 :option:`clat_percentiles`.
f80dba8d
MT
3209
3210.. option:: percentile_list=float_list
3211
c32ba107
JA
3212 Overwrite the default list of percentiles for completion latencies and
3213 the block error histogram. Each number is a floating number in the
3214 range (0,100], and the maximum length of the list is 20. Use ``:`` to
3215 separate the numbers, and list the numbers in ascending order. For
3216 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
3217 values of completion latency below which 99.5% and 99.9% of the observed
3218 latencies fell, respectively.
f80dba8d 3219
e883cb35
JF
3220.. option:: significant_figures=int
3221
c32ba107
JA
3222 If using :option:`--output-format` of `normal`, set the significant
3223 figures to this value. Higher values will yield more precise IOPS and
3224 throughput units, while lower values will round. Requires a minimum
3225 value of 1 and a maximum value of 10. Defaults to 4.
e883cb35 3226
f80dba8d
MT
3227
3228Error handling
3229~~~~~~~~~~~~~~
3230
3231.. option:: exitall_on_error
3232
3233 When one job finishes in error, terminate the rest. The default is to wait
3234 for each job to finish.
3235
3236.. option:: continue_on_error=str
3237
3238 Normally fio will exit the job on the first observed failure. If this option
3239 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
3240 EILSEQ) until the runtime is exceeded or the I/O size specified is
3241 completed. If this option is used, there are two more stats that are
3242 appended, the total error count and the first error. The error field given
3243 in the stats is the first error that was hit during the run.
3244
3245 The allowed values are:
3246
3247 **none**
3248 Exit on any I/O or verify errors.
3249
3250 **read**
3251 Continue on read errors, exit on all others.
3252
3253 **write**
3254 Continue on write errors, exit on all others.
3255
3256 **io**
3257 Continue on any I/O error, exit on all others.
3258
3259 **verify**
3260 Continue on verify errors, exit on all others.
3261
3262 **all**
3263 Continue on all errors.
3264
3265 **0**
3266 Backward-compatible alias for 'none'.
3267
3268 **1**
3269 Backward-compatible alias for 'all'.
3270
3271.. option:: ignore_error=str
3272
3273 Sometimes you want to ignore some errors during test in that case you can
a35ef7cb
TK
3274 specify error list for each error type, instead of only being able to
3275 ignore the default 'non-fatal error' using :option:`continue_on_error`.
f80dba8d
MT
3276 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
3277 given error type is separated with ':'. Error may be symbol ('ENOSPC',
3278 'ENOMEM') or integer. Example::
3279
3280 ignore_error=EAGAIN,ENOSPC:122
3281
3282 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
a35ef7cb
TK
3283 WRITE. This option works by overriding :option:`continue_on_error` with
3284 the list of errors for each error type if any.
f80dba8d
MT
3285
3286.. option:: error_dump=bool
3287
3288 If set dump every error even if it is non fatal, true by default. If
3289 disabled only fatal error will be dumped.
3290
f75ede1d
SW
3291Running predefined workloads
3292----------------------------
3293
3294Fio includes predefined profiles that mimic the I/O workloads generated by
3295other tools.
3296
3297.. option:: profile=str
3298
3299 The predefined workload to run. Current profiles are:
3300
3301 **tiobench**
3302 Threaded I/O bench (tiotest/tiobench) like workload.
3303
3304 **act**
3305 Aerospike Certification Tool (ACT) like workload.
3306
3307To view a profile's additional options use :option:`--cmdhelp` after specifying
3308the profile. For example::
3309
f50fbdda 3310 $ fio --profile=act --cmdhelp
f75ede1d
SW
3311
3312Act profile options
3313~~~~~~~~~~~~~~~~~~~
3314
3315.. option:: device-names=str
3316 :noindex:
3317
3318 Devices to use.
3319
3320.. option:: load=int
3321 :noindex:
3322
3323 ACT load multiplier. Default: 1.
3324
3325.. option:: test-duration=time
3326 :noindex:
3327
947e0fe0
SW
3328 How long the entire test takes to run. When the unit is omitted, the value
3329 is given in seconds. Default: 24h.
f75ede1d
SW
3330
3331.. option:: threads-per-queue=int
3332 :noindex:
3333
f50fbdda 3334 Number of read I/O threads per device. Default: 8.
f75ede1d
SW
3335
3336.. option:: read-req-num-512-blocks=int
3337 :noindex:
3338
3339 Number of 512B blocks to read at the time. Default: 3.
3340
3341.. option:: large-block-op-kbytes=int
3342 :noindex:
3343
3344 Size of large block ops in KiB (writes). Default: 131072.
3345
3346.. option:: prep
3347 :noindex:
3348
3349 Set to run ACT prep phase.
3350
3351Tiobench profile options
3352~~~~~~~~~~~~~~~~~~~~~~~~
3353
3354.. option:: size=str
3355 :noindex:
3356
f50fbdda 3357 Size in MiB.
f75ede1d
SW
3358
3359.. option:: block=int
3360 :noindex:
3361
3362 Block size in bytes. Default: 4096.
3363
3364.. option:: numruns=int
3365 :noindex:
3366
3367 Number of runs.
3368
3369.. option:: dir=str
3370 :noindex:
3371
3372 Test directory.
3373
3374.. option:: threads=int
3375 :noindex:
3376
3377 Number of threads.
f80dba8d
MT
3378
3379Interpreting the output
3380-----------------------
3381
36214730
SW
3382..
3383 Example output was based on the following:
3384 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
3385 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
3386 --runtime=2m --rw=rw
3387
f80dba8d
MT
3388Fio spits out a lot of output. While running, fio will display the status of the
3389jobs created. An example of that would be::
3390
9d25d068 3391 Jobs: 1 (f=1): [_(1),M(1)][24.8%][r=20.5MiB/s,w=23.5MiB/s][r=82,w=94 IOPS][eta 01m:31s]
f80dba8d 3392
36214730
SW
3393The characters inside the first set of square brackets denote the current status of
3394each thread. The first character is the first job defined in the job file, and so
3395forth. The possible values (in typical life cycle order) are:
f80dba8d
MT
3396
3397+------+-----+-----------------------------------------------------------+
3398| Idle | Run | |
3399+======+=====+===========================================================+
3400| P | | Thread setup, but not started. |
3401+------+-----+-----------------------------------------------------------+
3402| C | | Thread created. |
3403+------+-----+-----------------------------------------------------------+
3404| I | | Thread initialized, waiting or generating necessary data. |
3405+------+-----+-----------------------------------------------------------+
3406| | p | Thread running pre-reading file(s). |
3407+------+-----+-----------------------------------------------------------+
36214730
SW
3408| | / | Thread is in ramp period. |
3409+------+-----+-----------------------------------------------------------+
f80dba8d
MT
3410| | R | Running, doing sequential reads. |
3411+------+-----+-----------------------------------------------------------+
3412| | r | Running, doing random reads. |
3413+------+-----+-----------------------------------------------------------+
3414| | W | Running, doing sequential writes. |
3415+------+-----+-----------------------------------------------------------+
3416| | w | Running, doing random writes. |
3417+------+-----+-----------------------------------------------------------+
3418| | M | Running, doing mixed sequential reads/writes. |
3419+------+-----+-----------------------------------------------------------+
3420| | m | Running, doing mixed random reads/writes. |
3421+------+-----+-----------------------------------------------------------+
36214730
SW
3422| | D | Running, doing sequential trims. |
3423+------+-----+-----------------------------------------------------------+
3424| | d | Running, doing random trims. |
3425+------+-----+-----------------------------------------------------------+
3426| | F | Running, currently waiting for :manpage:`fsync(2)`. |
f80dba8d
MT
3427+------+-----+-----------------------------------------------------------+
3428| | V | Running, doing verification of written data. |
3429+------+-----+-----------------------------------------------------------+
36214730
SW
3430| f | | Thread finishing. |
3431+------+-----+-----------------------------------------------------------+
f80dba8d
MT
3432| E | | Thread exited, not reaped by main thread yet. |
3433+------+-----+-----------------------------------------------------------+
36214730 3434| _ | | Thread reaped. |
f80dba8d
MT
3435+------+-----+-----------------------------------------------------------+
3436| X | | Thread reaped, exited with an error. |
3437+------+-----+-----------------------------------------------------------+
3438| K | | Thread reaped, exited due to signal. |
3439+------+-----+-----------------------------------------------------------+
3440
36214730
SW
3441..
3442 Example output was based on the following:
3443 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3444 --time_based --rate=2512k --bs=256K --numjobs=10 \
3445 --name=readers --rw=read --name=writers --rw=write
3446
f80dba8d 3447Fio will condense the thread string as not to take up more space on the command
36214730 3448line than needed. For instance, if you have 10 readers and 10 writers running,
f80dba8d
MT
3449the output would look like this::
3450
9d25d068 3451 Jobs: 20 (f=20): [R(10),W(10)][4.0%][r=20.5MiB/s,w=23.5MiB/s][r=82,w=94 IOPS][eta 57m:36s]
f80dba8d 3452
36214730
SW
3453Note that the status string is displayed in order, so it's possible to tell which of
3454the jobs are currently doing what. In the example above this means that jobs 1--10
3455are readers and 11--20 are writers.
f80dba8d
MT
3456
3457The other values are fairly self explanatory -- number of threads currently
36214730
SW
3458running and doing I/O, the number of currently open files (f=), the estimated
3459completion percentage, the rate of I/O since last check (read speed listed first,
f50fbdda
TK
3460then write speed and optionally trim speed) in terms of bandwidth and IOPS,
3461and time to completion for the current running group. It's impossible to estimate
3462runtime of the following groups (if any).
36214730
SW
3463
3464..
3465 Example output was based on the following:
3466 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3467 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3468 --bs=7K --name=Client1 --rw=write
3469
3470When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3471each thread, group of threads, and disks in that order. For each overall thread (or
3472group) the output looks like::
3473
3474 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3475 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3476 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3477 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3478 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3479 clat percentiles (usec):
3480 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3481 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3482 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3483 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3484 | 99.99th=[78119]
3485 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3486 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
29092211
VF
3487 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3488 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3489 lat (msec) : 100=0.65%
36214730
SW
3490 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3491 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3492 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3493 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3494 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3495 latency : target=0, window=0, percentile=100.00%, depth=8
3496
3497The job name (or first job's name when using :option:`group_reporting`) is printed,
3498along with the group id, count of jobs being aggregated, last error id seen (which
3499is 0 when there are no errors), pid/tid of that thread and the time the job/group
3500completed. Below are the I/O statistics for each data direction performed (showing
3501writes in the example above). In the order listed, they denote:
3502
3503**read/write/trim**
3504 The string before the colon shows the I/O direction the statistics
3505 are for. **IOPS** is the average I/Os performed per second. **BW**
3506 is the average bandwidth rate shown as: value in power of 2 format
3507 (value in power of 10 format). The last two values show: (**total
3508 I/O performed** in power of 2 format / **runtime** of that thread).
f80dba8d
MT
3509
3510**slat**
36214730
SW
3511 Submission latency (**min** being the minimum, **max** being the
3512 maximum, **avg** being the average, **stdev** being the standard
3513 deviation). This is the time it took to submit the I/O. For
3514 sync I/O this row is not displayed as the slat is really the
3515 completion latency (since queue/complete is one operation there).
3516 This value can be in nanoseconds, microseconds or milliseconds ---
3517 fio will choose the most appropriate base and print that (in the
3518 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
0d237712 3519 latencies are always expressed in microseconds.
f80dba8d
MT
3520
3521**clat**
3522 Completion latency. Same names as slat, this denotes the time from
3523 submission to completion of the I/O pieces. For sync I/O, clat will
3524 usually be equal (or very close) to 0, as the time from submit to
3525 complete is basically just CPU time (I/O has already been done, see slat
3526 explanation).
3527
29092211
VF
3528**lat**
3529 Total latency. Same names as slat and clat, this denotes the time from
3530 when fio created the I/O unit to completion of the I/O operation.
3531
f80dba8d 3532**bw**
36214730
SW
3533 Bandwidth statistics based on samples. Same names as the xlat stats,
3534 but also includes the number of samples taken (**samples**) and an
3535 approximate percentage of total aggregate bandwidth this thread
3536 received in its group (**per**). This last value is only really
3537 useful if the threads in this group are on the same disk, since they
3538 are then competing for disk access.
3539
3540**iops**
3541 IOPS statistics based on samples. Same names as bw.
f80dba8d 3542
29092211
VF
3543**lat (nsec/usec/msec)**
3544 The distribution of I/O completion latencies. This is the time from when
3545 I/O leaves fio and when it gets completed. Unlike the separate
3546 read/write/trim sections above, the data here and in the remaining
3547 sections apply to all I/Os for the reporting group. 250=0.04% means that
3548 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
3549 of the I/Os required 250 to 499us for completion.
3550
f80dba8d
MT
3551**cpu**
3552 CPU usage. User and system time, along with the number of context
3553 switches this thread went through, usage of system and user time, and
3554 finally the number of major and minor page faults. The CPU utilization
3555 numbers are averages for the jobs in that reporting group, while the
23a8e176 3556 context and fault counters are summed.
f80dba8d
MT
3557
3558**IO depths**
a2140525
SW
3559 The distribution of I/O depths over the job lifetime. The numbers are
3560 divided into powers of 2 and each entry covers depths from that value
3561 up to those that are lower than the next entry -- e.g., 16= covers
3562 depths from 16 to 31. Note that the range covered by a depth
3563 distribution entry can be different to the range covered by the
3564 equivalent submit/complete distribution entry.
f80dba8d
MT
3565
3566**IO submit**
3567 How many pieces of I/O were submitting in a single submit call. Each
c60ebc45 3568 entry denotes that amount and below, until the previous entry -- e.g.,
a2140525
SW
3569 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3570 call. Note that the range covered by a submit distribution entry can
3571 be different to the range covered by the equivalent depth distribution
3572 entry.
f80dba8d
MT
3573
3574**IO complete**
3575 Like the above submit number, but for completions instead.
3576
36214730
SW
3577**IO issued rwt**
3578 The number of read/write/trim requests issued, and how many of them were
3579 short or dropped.
f80dba8d 3580
29092211 3581**IO latency**
ee21ebee 3582 These values are for :option:`latency_target` and related options. When
29092211
VF
3583 these options are engaged, this section describes the I/O depth required
3584 to meet the specified latency target.
71bfa161 3585
36214730
SW
3586..
3587 Example output was based on the following:
3588 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3589 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3590 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3591
71bfa161 3592After each client has been listed, the group statistics are printed. They
f80dba8d 3593will look like this::
71bfa161 3594
f80dba8d 3595 Run status group 0 (all jobs):
36214730
SW
3596 READ: bw=20.9MiB/s (21.9MB/s), 10.4MiB/s-10.8MiB/s (10.9MB/s-11.3MB/s), io=64.0MiB (67.1MB), run=2973-3069msec
3597 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
71bfa161 3598
36214730 3599For each data direction it prints:
71bfa161 3600
36214730
SW
3601**bw**
3602 Aggregate bandwidth of threads in this group followed by the
3603 minimum and maximum bandwidth of all the threads in this group.
3604 Values outside of brackets are power-of-2 format and those
3605 within are the equivalent value in a power-of-10 format.
f80dba8d 3606**io**
36214730
SW
3607 Aggregate I/O performed of all threads in this group. The
3608 format is the same as bw.
3609**run**
3610 The smallest and longest runtimes of the threads in this group.
71bfa161 3611
f50fbdda 3612And finally, the disk statistics are printed. This is Linux specific. They will look like this::
71bfa161 3613
f80dba8d
MT
3614 Disk stats (read/write):
3615 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
71bfa161
JA
3616
3617Each value is printed for both reads and writes, with reads first. The
3618numbers denote:
3619
f80dba8d 3620**ios**
c60ebc45 3621 Number of I/Os performed by all groups.
f80dba8d 3622**merge**
007c7be9 3623 Number of merges performed by the I/O scheduler.
f80dba8d
MT
3624**ticks**
3625 Number of ticks we kept the disk busy.
36214730 3626**in_queue**
f80dba8d
MT
3627 Total time spent in the disk queue.
3628**util**
3629 The disk utilization. A value of 100% means we kept the disk
71bfa161
JA
3630 busy constantly, 50% would be a disk idling half of the time.
3631
f80dba8d
MT
3632It is also possible to get fio to dump the current output while it is running,
3633without terminating the job. To do that, send fio the **USR1** signal. You can
3634also get regularly timed dumps by using the :option:`--status-interval`
3635parameter, or by creating a file in :file:`/tmp` named
3636:file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3637current output status.
8423bd11 3638
71bfa161 3639
f80dba8d
MT
3640Terse output
3641------------
71bfa161 3642
f80dba8d
MT
3643For scripted usage where you typically want to generate tables or graphs of the
3644results, fio can output the results in a semicolon separated format. The format
3645is one long line of values, such as::
71bfa161 3646
f80dba8d
MT
3647 2;card0;0;0;7139336;121836;60004;1;10109;27.932460;116.933948;220;126861;3495.446807;1085.368601;226;126864;3523.635629;1089.012448;24063;99944;50.275485%;59818.274627;5540.657370;7155060;122104;60004;1;8338;29.086342;117.839068;388;128077;5032.488518;1234.785715;391;128085;5061.839412;1236.909129;23436;100928;50.287926%;59964.832030;5644.844189;14.595833%;19.394167%;123706;0;7313;0.1%;0.1%;0.1%;0.1%;0.1%;0.1%;100.0%;0.00%;0.00%;0.00%;0.00%;0.00%;0.00%;0.01%;0.02%;0.05%;0.16%;6.04%;40.40%;52.68%;0.64%;0.01%;0.00%;0.01%;0.00%;0.00%;0.00%;0.00%;0.00%
3648 A description of this job goes here.
562c2d2f
DN
3649
3650The job description (if provided) follows on a second line.
71bfa161 3651
a7f77fa6
SW
3652To enable terse output, use the :option:`--minimal` or
3653:option:`--output-format`\=terse command line options. The
f80dba8d
MT
3654first value is the version of the terse output format. If the output has to be
3655changed for some reason, this number will be incremented by 1 to signify that
3656change.
6820cb3b 3657
a2c95580 3658Split up, the format is as follows (comments in brackets denote when a
007c7be9 3659field was introduced or whether it's specific to some terse version):
71bfa161 3660
f80dba8d
MT
3661 ::
3662
f50fbdda 3663 terse version, fio version [v3], jobname, groupid, error
f80dba8d
MT
3664
3665 READ status::
3666
3667 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3668 Submission latency: min, max, mean, stdev (usec)
3669 Completion latency: min, max, mean, stdev (usec)
3670 Completion latency percentiles: 20 fields (see below)
3671 Total latency: min, max, mean, stdev (usec)
f50fbdda
TK
3672 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3673 IOPS [v5]: min, max, mean, stdev, number of samples
f80dba8d
MT
3674
3675 WRITE status:
3676
3677 ::
3678
3679 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3680 Submission latency: min, max, mean, stdev (usec)
247823cc 3681 Completion latency: min, max, mean, stdev (usec)
f80dba8d
MT
3682 Completion latency percentiles: 20 fields (see below)
3683 Total latency: min, max, mean, stdev (usec)
f50fbdda
TK
3684 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3685 IOPS [v5]: min, max, mean, stdev, number of samples
a2c95580
AH
3686
3687 TRIM status [all but version 3]:
3688
f50fbdda 3689 Fields are similar to READ/WRITE status.
f80dba8d
MT
3690
3691 CPU usage::
3692
3693 user, system, context switches, major faults, minor faults
3694
3695 I/O depths::
3696
3697 <=1, 2, 4, 8, 16, 32, >=64
3698
3699 I/O latencies microseconds::
3700
3701 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3702
3703 I/O latencies milliseconds::
3704
3705 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3706
a2c95580 3707 Disk utilization [v3]::
f80dba8d 3708
f50fbdda
TK
3709 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
3710 time spent in queue, disk utilization percentage
f80dba8d
MT
3711
3712 Additional Info (dependent on continue_on_error, default off)::
3713
3714 total # errors, first error code
3715
3716 Additional Info (dependent on description being set)::
3717
3718 Text description
3719
3720Completion latency percentiles can be a grouping of up to 20 sets, so for the
3721terse output fio writes all of them. Each field will look like this::
1db92cb6 3722
f50fbdda 3723 1.00%=6112
1db92cb6 3724
f80dba8d 3725which is the Xth percentile, and the `usec` latency associated with it.
1db92cb6 3726
f50fbdda 3727For `Disk utilization`, all disks used by fio are shown. So for each disk there
f80dba8d 3728will be a disk utilization section.
f2f788dd 3729
2fc26c3d 3730Below is a single line containing short names for each of the fields in the
2831be97 3731minimal output v3, separated by semicolons::
2fc26c3d 3732
f50fbdda 3733 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth;read_iops;read_runtime_ms;read_slat_min;read_slat_max;read_slat_mean;read_slat_dev;read_clat_min;read_clat_max;read_clat_mean;read_clat_dev;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min;read_lat_max;read_lat_mean;read_lat_dev;read_bw_min;read_bw_max;read_bw_agg_pct;read_bw_mean;read_bw_dev;write_kb;write_bandwidth;write_iops;write_runtime_ms;write_slat_min;write_slat_max;write_slat_mean;write_slat_dev;write_clat_min;write_clat_max;write_clat_mean;write_clat_dev;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min;write_lat_max;write_lat_mean;write_lat_dev;write_bw_min;write_bw_max;write_bw_agg_pct;write_bw_mean;write_bw_dev;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
2fc26c3d 3734
25c8b9d7 3735
44c82dba
VF
3736JSON output
3737------------
3738
3739The `json` output format is intended to be both human readable and convenient
3740for automated parsing. For the most part its sections mirror those of the
3741`normal` output. The `runtime` value is reported in msec and the `bw` value is
3742reported in 1024 bytes per second units.
3743
3744
d29c4a91
VF
3745JSON+ output
3746------------
3747
3748The `json+` output format is identical to the `json` output format except that it
3749adds a full dump of the completion latency bins. Each `bins` object contains a
3750set of (key, value) pairs where keys are latency durations and values count how
3751many I/Os had completion latencies of the corresponding duration. For example,
3752consider:
3753
3754 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
3755
3756This data indicates that one I/O required 87,552ns to complete, two I/Os required
3757100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
3758
3759Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
3760json+ output and generates CSV-formatted latency data suitable for plotting.
3761
3762The latency durations actually represent the midpoints of latency intervals.
f50fbdda 3763For details refer to :file:`stat.h`.
d29c4a91
VF
3764
3765
f80dba8d
MT
3766Trace file format
3767-----------------
3768
3769There are two trace file format that you can encounter. The older (v1) format is
3770unsupported since version 1.20-rc3 (March 2008). It will still be described
25c8b9d7
PD
3771below in case that you get an old trace and want to understand it.
3772
3773In any case the trace is a simple text file with a single action per line.
3774
3775
f80dba8d
MT
3776Trace file format v1
3777~~~~~~~~~~~~~~~~~~~~
3778
3779Each line represents a single I/O action in the following format::
3780
3781 rw, offset, length
25c8b9d7 3782
f50fbdda 3783where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
25c8b9d7 3784
22413915 3785This format is not supported in fio versions >= 1.20-rc3.
25c8b9d7 3786
25c8b9d7 3787
f80dba8d
MT
3788Trace file format v2
3789~~~~~~~~~~~~~~~~~~~~
25c8b9d7 3790
f80dba8d
MT
3791The second version of the trace file format was added in fio version 1.17. It
3792allows to access more then one file per trace and has a bigger set of possible
3793file actions.
25c8b9d7 3794
f80dba8d