4 The first step in getting fio to simulate a desired I/O workload, is writing a
5 job file describing that specific setup. A job file may contain any number of
6 threads and/or files -- the typical contents of the job file is a *global*
7 section defining shared parameters, and one or more job sections describing the
8 jobs involved. When run, fio parses this file and sets everything up as
9 described. If we break down a job from top to bottom, it contains the following
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?
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.
26 How much data are we going to be reading/writing.
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
36 If the I/O engine is async, how large a queuing depth do we want to
42 How many files are we spreading the workload over.
44 `Threads, processes and job synchronization`_
46 How many threads or processes should we spread this workload over.
48 The above are the basic parameters defined for a workload, in addition there's a
49 multitude of parameters that modify other aspects of how this job behaves.
55 .. option:: --debug=type
57 Enable verbose tracing `type` of various fio actions. May be ``all`` for all types
58 or individual types separated by a comma (e.g. ``--debug=file,mem`` will
59 enable file and memory debugging). Currently, additional logging is
63 Dump info related to processes.
65 Dump info related to file actions.
67 Dump info related to I/O queuing.
69 Dump info related to memory allocations.
71 Dump info related to blktrace setup.
73 Dump info related to I/O verification.
75 Enable all debug options.
77 Dump info related to random offset generation.
79 Dump info related to option matching and parsing.
81 Dump info related to disk utilization updates.
83 Dump info only related to job number x.
85 Dump info only related to mutex up/down ops.
87 Dump info related to profile extensions.
89 Dump info related to internal time keeping.
91 Dump info related to networking connections.
93 Dump info related to I/O rate switching.
95 Dump info related to log compress/decompress.
97 Dump info related to steadystate detection.
99 Dump info related to the helper thread.
101 Dump info related to support for zoned block devices.
103 Show available debug options.
105 .. option:: --parse-only
107 Parse options only, don't start any I/O.
109 .. option:: --merge-blktrace-only
111 Merge blktraces only, don't start any I/O.
113 .. option:: --output=filename
115 Write output to file `filename`.
117 .. option:: --output-format=format
119 Set the reporting `format` to `normal`, `terse`, `json`, or `json+`. Multiple
120 formats can be selected, separated by a comma. `terse` is a CSV based
121 format. `json+` is like `json`, except it adds a full dump of the latency
124 .. option:: --bandwidth-log
126 Generate aggregate bandwidth logs.
128 .. option:: --minimal
130 Print statistics in a terse, semicolon-delimited format.
132 .. option:: --append-terse
134 Print statistics in selected mode AND terse, semicolon-delimited format.
135 **Deprecated**, use :option:`--output-format` instead to select multiple
138 .. option:: --terse-version=version
140 Set terse `version` output format (default 3, or 2 or 4 or 5).
142 .. option:: --version
144 Print version information and exit.
148 Print a summary of the command line options and exit.
150 .. option:: --cpuclock-test
152 Perform test and validation of internal CPU clock.
154 .. option:: --crctest=[test]
156 Test the speed of the built-in checksumming functions. If no argument is
157 given, all of them are tested. Alternatively, a comma separated list can
158 be passed, in which case the given ones are tested.
160 .. option:: --cmdhelp=command
162 Print help information for `command`. May be ``all`` for all commands.
164 .. option:: --enghelp=[ioengine[,command]]
166 List all commands defined by `ioengine`, or print help for `command`
167 defined by `ioengine`. If no `ioengine` is given, list all
170 .. option:: --showcmd
172 Convert given job files to a set of command-line options.
174 .. option:: --readonly
176 Turn on safety read-only checks, preventing writes and trims. The
177 ``--readonly`` option is an extra safety guard to prevent users from
178 accidentally starting a write or trim workload when that is not desired.
179 Fio will only modify the device under test if
180 `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite` is given. This
181 safety net can be used as an extra precaution.
183 .. option:: --eta=when
185 Specifies when real-time ETA estimate should be printed. `when` may be
186 `always`, `never` or `auto`. `auto` is the default, it prints ETA
187 when requested if the output is a TTY. `always` disregards the output
188 type, and prints ETA when requested. `never` never prints ETA.
190 .. option:: --eta-interval=time
192 By default, fio requests client ETA status roughly every second. With
193 this option, the interval is configurable. Fio imposes a minimum
194 allowed time to avoid flooding the console, less than 250 msec is
197 .. option:: --eta-newline=time
199 Force a new line for every `time` period passed. When the unit is omitted,
200 the value is interpreted in seconds.
202 .. option:: --status-interval=time
204 Force a full status dump of cumulative (from job start) values at `time`
205 intervals. This option does *not* provide per-period measurements. So
206 values such as bandwidth are running averages. When the time unit is omitted,
207 `time` is interpreted in seconds. Note that using this option with
208 ``--output-format=json`` will yield output that technically isn't valid
209 json, since the output will be collated sets of valid json. It will need
210 to be split into valid sets of json after the run.
212 .. option:: --section=name
214 Only run specified section `name` in job file. Multiple sections can be specified.
215 The ``--section`` option allows one to combine related jobs into one file.
216 E.g. one job file could define light, moderate, and heavy sections. Tell
217 fio to run only the "heavy" section by giving ``--section=heavy``
218 command line option. One can also specify the "write" operations in one
219 section and "verify" operation in another section. The ``--section`` option
220 only applies to job sections. The reserved *global* section is always
223 .. option:: --alloc-size=kb
225 Allocate additional internal smalloc pools of size `kb` in KiB. The
226 ``--alloc-size`` option increases shared memory set aside for use by fio.
227 If running large jobs with randommap enabled, fio can run out of memory.
228 Smalloc is an internal allocator for shared structures from a fixed size
229 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
231 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
234 .. option:: --warnings-fatal
236 All fio parser warnings are fatal, causing fio to exit with an
239 .. option:: --max-jobs=nr
241 Set the maximum number of threads/processes to support to `nr`.
242 NOTE: On Linux, it may be necessary to increase the shared-memory
243 limit (:file:`/proc/sys/kernel/shmmax`) if fio runs into errors while
246 .. option:: --server=args
248 Start a backend server, with `args` specifying what to listen to.
249 See `Client/Server`_ section.
251 .. option:: --daemonize=pidfile
253 Background a fio server, writing the pid to the given `pidfile` file.
255 .. option:: --client=hostname
257 Instead of running the jobs locally, send and run them on the given `hostname`
258 or set of `hostname`\s. See `Client/Server`_ section.
260 .. option:: --remote-config=file
262 Tell fio server to load this local `file`.
264 .. option:: --idle-prof=option
266 Report CPU idleness. `option` is one of the following:
269 Run unit work calibration only and exit.
272 Show aggregate system idleness and unit work.
275 As **system** but also show per CPU idleness.
277 .. option:: --inflate-log=log
279 Inflate and output compressed `log`.
281 .. option:: --trigger-file=file
283 Execute trigger command when `file` exists.
285 .. option:: --trigger-timeout=time
287 Execute trigger at this `time`.
289 .. option:: --trigger=command
291 Set this `command` as local trigger.
293 .. option:: --trigger-remote=command
295 Set this `command` as remote trigger.
297 .. option:: --aux-path=path
299 Use the directory specified by `path` for generated state files instead
300 of the current working directory.
302 Any parameters following the options will be assumed to be job files, unless
303 they match a job file parameter. Multiple job files can be listed and each job
304 file will be regarded as a separate group. Fio will :option:`stonewall`
305 execution between each group.
311 As previously described, fio accepts one or more job files describing what it is
312 supposed to do. The job file format is the classic ini file, where the names
313 enclosed in [] brackets define the job name. You are free to use any ASCII name
314 you want, except *global* which has special meaning. Following the job name is
315 a sequence of zero or more parameters, one per line, that define the behavior of
316 the job. If the first character in a line is a ';' or a '#', the entire line is
317 discarded as a comment.
319 A *global* section sets defaults for the jobs described in that file. A job may
320 override a *global* section parameter, and a job file may even have several
321 *global* sections if so desired. A job is only affected by a *global* section
324 The :option:`--cmdhelp` option also lists all options. If used with a `command`
325 argument, :option:`--cmdhelp` will detail the given `command`.
327 See the `examples/` directory for inspiration on how to write job files. Note
328 the copyright and license requirements currently apply to `examples/` files.
330 So let's look at a really simple job file that defines two processes, each
331 randomly reading from a 128MiB file:
335 ; -- start job file --
346 As you can see, the job file sections themselves are empty as all the described
347 parameters are shared. As no :option:`filename` option is given, fio makes up a
348 `filename` for each of the jobs as it sees fit. On the command line, this job
349 would look as follows::
351 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
354 Let's look at an example that has a number of processes writing randomly to
359 ; -- start job file --
370 Here we have no *global* section, as we only have one job defined anyway. We
371 want to use async I/O here, with a depth of 4 for each file. We also increased
372 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
373 jobs. The result is 4 processes each randomly writing to their own 64MiB
374 file. Instead of using the above job file, you could have given the parameters
375 on the command line. For this case, you would specify::
377 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
379 When fio is utilized as a basis of any reasonably large test suite, it might be
380 desirable to share a set of standardized settings across multiple job files.
381 Instead of copy/pasting such settings, any section may pull in an external
382 :file:`filename.fio` file with *include filename* directive, as in the following
385 ; -- start job file including.fio --
389 include glob-include.fio
396 include test-include.fio
397 ; -- end job file including.fio --
401 ; -- start job file glob-include.fio --
404 ; -- end job file glob-include.fio --
408 ; -- start job file test-include.fio --
411 ; -- end job file test-include.fio --
413 Settings pulled into a section apply to that section only (except *global*
414 section). Include directives may be nested in that any included file may contain
415 further include directive(s). Include files may not contain [] sections.
418 Environment variables
419 ~~~~~~~~~~~~~~~~~~~~~
421 Fio also supports environment variable expansion in job files. Any sub-string of
422 the form ``${VARNAME}`` as part of an option value (in other words, on the right
423 of the '='), will be expanded to the value of the environment variable called
424 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
425 empty string, the empty string will be substituted.
427 As an example, let's look at a sample fio invocation and job file::
429 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
433 ; -- start job file --
440 This will expand to the following equivalent job file at runtime:
444 ; -- start job file --
451 Fio ships with a few example job files, you can also look there for inspiration.
456 Additionally, fio has a set of reserved keywords that will be replaced
457 internally with the appropriate value. Those keywords are:
461 The architecture page size of the running system.
465 Megabytes of total memory in the system.
469 Number of online available CPUs.
471 These can be used on the command line or in the job file, and will be
472 automatically substituted with the current system values when the job is
473 run. Simple math is also supported on these keywords, so you can perform actions
478 and get that properly expanded to 8 times the size of memory in the machine.
484 This section describes in details each parameter associated with a job. Some
485 parameters take an option of a given type, such as an integer or a
486 string. Anywhere a numeric value is required, an arithmetic expression may be
487 used, provided it is surrounded by parentheses. Supported operators are:
496 For time values in expressions, units are microseconds by default. This is
497 different than for time values not in expressions (not enclosed in
498 parentheses). The following types are used:
505 String: A sequence of alphanumeric characters.
508 Integer with possible time suffix. Without a unit value is interpreted as
509 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
510 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
511 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
516 Integer. A whole number value, which may contain an integer prefix
517 and an integer suffix:
519 [*integer prefix*] **number** [*integer suffix*]
521 The optional *integer prefix* specifies the number's base. The default
522 is decimal. *0x* specifies hexadecimal.
524 The optional *integer suffix* specifies the number's units, and includes an
525 optional unit prefix and an optional unit. For quantities of data, the
526 default unit is bytes. For quantities of time, the default unit is seconds
527 unless otherwise specified.
529 With :option:`kb_base`\=1000, fio follows international standards for unit
530 prefixes. To specify power-of-10 decimal values defined in the
531 International System of Units (SI):
533 * *K* -- means kilo (K) or 1000
534 * *M* -- means mega (M) or 1000**2
535 * *G* -- means giga (G) or 1000**3
536 * *T* -- means tera (T) or 1000**4
537 * *P* -- means peta (P) or 1000**5
539 To specify power-of-2 binary values defined in IEC 80000-13:
541 * *Ki* -- means kibi (Ki) or 1024
542 * *Mi* -- means mebi (Mi) or 1024**2
543 * *Gi* -- means gibi (Gi) or 1024**3
544 * *Ti* -- means tebi (Ti) or 1024**4
545 * *Pi* -- means pebi (Pi) or 1024**5
547 For Zone Block Device Mode:
550 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
551 from those specified in the SI and IEC 80000-13 standards to provide
552 compatibility with old scripts. For example, 4k means 4096.
554 For quantities of data, an optional unit of 'B' may be included
555 (e.g., 'kB' is the same as 'k').
557 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
558 not milli). 'b' and 'B' both mean byte, not bit.
560 Examples with :option:`kb_base`\=1000:
562 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
563 * *1 MiB*: 1048576, 1mi, 1024ki
564 * *1 MB*: 1000000, 1m, 1000k
565 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
566 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
568 Examples with :option:`kb_base`\=1024 (default):
570 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
571 * *1 MiB*: 1048576, 1m, 1024k
572 * *1 MB*: 1000000, 1mi, 1000ki
573 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
574 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
576 To specify times (units are not case sensitive):
580 * *M* -- means minutes
581 * *s* -- or sec means seconds (default)
582 * *ms* -- or *msec* means milliseconds
583 * *us* -- or *usec* means microseconds
585 If the option accepts an upper and lower range, use a colon ':' or
586 minus '-' to separate such values. See :ref:`irange <irange>`.
587 If the lower value specified happens to be larger than the upper value
588 the two values are swapped.
593 Boolean. Usually parsed as an integer, however only defined for
594 true and false (1 and 0).
599 Integer range with suffix. Allows value range to be given, such as
600 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
601 option allows two sets of ranges, they can be specified with a ',' or '/'
602 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
605 A list of floating point numbers, separated by a ':' character.
607 With the above in mind, here follows the complete list of fio job parameters.
613 .. option:: kb_base=int
615 Select the interpretation of unit prefixes in input parameters.
618 Inputs comply with IEC 80000-13 and the International
619 System of Units (SI). Use:
621 - power-of-2 values with IEC prefixes (e.g., KiB)
622 - power-of-10 values with SI prefixes (e.g., kB)
625 Compatibility mode (default). To avoid breaking old scripts:
627 - power-of-2 values with SI prefixes
628 - power-of-10 values with IEC prefixes
630 See :option:`bs` for more details on input parameters.
632 Outputs always use correct prefixes. Most outputs include both
635 bw=2383.3kB/s (2327.4KiB/s)
637 If only one value is reported, then kb_base selects the one to use:
639 **1000** -- SI prefixes
641 **1024** -- IEC prefixes
643 .. option:: unit_base=int
645 Base unit for reporting. Allowed values are:
648 Use auto-detection (default).
660 ASCII name of the job. This may be used to override the name printed by fio
661 for this job. Otherwise the job name is used. On the command line this
662 parameter has the special purpose of also signaling the start of a new job.
664 .. option:: description=str
666 Text description of the job. Doesn't do anything except dump this text
667 description when this job is run. It's not parsed.
669 .. option:: loops=int
671 Run the specified number of iterations of this job. Used to repeat the same
672 workload a given number of times. Defaults to 1.
674 .. option:: numjobs=int
676 Create the specified number of clones of this job. Each clone of job
677 is spawned as an independent thread or process. May be used to setup a
678 larger number of threads/processes doing the same thing. Each thread is
679 reported separately; to see statistics for all clones as a whole, use
680 :option:`group_reporting` in conjunction with :option:`new_group`.
681 See :option:`--max-jobs`. Default: 1.
684 Time related parameters
685 ~~~~~~~~~~~~~~~~~~~~~~~
687 .. option:: runtime=time
689 Limit runtime. The test will run until it completes the configured I/O
690 workload or until it has run for this specified amount of time, whichever
691 occurs first. It can be quite hard to determine for how long a specified
692 job will run, so this parameter is handy to cap the total runtime to a
693 given time. When the unit is omitted, the value is interpreted in
696 .. option:: time_based
698 If set, fio will run for the duration of the :option:`runtime` specified
699 even if the file(s) are completely read or written. It will simply loop over
700 the same workload as many times as the :option:`runtime` allows.
702 .. option:: startdelay=irange(time)
704 Delay the start of job for the specified amount of time. Can be a single
705 value or a range. When given as a range, each thread will choose a value
706 randomly from within the range. Value is in seconds if a unit is omitted.
708 .. option:: ramp_time=time
710 If set, fio will run the specified workload for this amount of time before
711 logging any performance numbers. Useful for letting performance settle
712 before logging results, thus minimizing the runtime required for stable
713 results. Note that the ``ramp_time`` is considered lead in time for a job,
714 thus it will increase the total runtime if a special timeout or
715 :option:`runtime` is specified. When the unit is omitted, the value is
718 .. option:: clocksource=str
720 Use the given clocksource as the base of timing. The supported options are:
723 :manpage:`gettimeofday(2)`
726 :manpage:`clock_gettime(2)`
729 Internal CPU clock source
731 cpu is the preferred clocksource if it is reliable, as it is very fast (and
732 fio is heavy on time calls). Fio will automatically use this clocksource if
733 it's supported and considered reliable on the system it is running on,
734 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
735 means supporting TSC Invariant.
737 .. option:: gtod_reduce=bool
739 Enable all of the :manpage:`gettimeofday(2)` reducing options
740 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
741 reduce precision of the timeout somewhat to really shrink the
742 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
743 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
744 time keeping was enabled.
746 .. option:: gtod_cpu=int
748 Sometimes it's cheaper to dedicate a single thread of execution to just
749 getting the current time. Fio (and databases, for instance) are very
750 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
751 one CPU aside for doing nothing but logging current time to a shared memory
752 location. Then the other threads/processes that run I/O workloads need only
753 copy that segment, instead of entering the kernel with a
754 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
755 calls will be excluded from other uses. Fio will manually clear it from the
756 CPU mask of other jobs.
758 .. option:: job_start_clock_id=int
759 The clock_id passed to the call to `clock_gettime` used to record job_start
760 in the `json` output format. Default is 0, or CLOCK_REALTIME.
766 .. option:: directory=str
768 Prefix filenames with this directory. Used to place files in a different
769 location than :file:`./`. You can specify a number of directories by
770 separating the names with a ':' character. These directories will be
771 assigned equally distributed to job clones created by :option:`numjobs` as
772 long as they are using generated filenames. If specific `filename(s)` are
773 set fio will use the first listed directory, and thereby matching the
774 `filename` semantic (which generates a file for each clone if not
775 specified, but lets all clones use the same file if set).
777 See the :option:`filename` option for information on how to escape "``:``"
778 characters within the directory path itself.
780 Note: To control the directory fio will use for internal state files
781 use :option:`--aux-path`.
783 .. option:: filename=str
785 Fio normally makes up a `filename` based on the job name, thread number, and
786 file number (see :option:`filename_format`). If you want to share files
787 between threads in a job or several
788 jobs with fixed file paths, specify a `filename` for each of them to override
789 the default. If the ioengine is file based, you can specify a number of files
790 by separating the names with a ':' colon. So if you wanted a job to open
791 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
792 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
793 specified, :option:`nrfiles` is ignored. The size of regular files specified
794 by this option will be :option:`size` divided by number of files unless an
795 explicit size is specified by :option:`filesize`.
797 Each colon in the wanted path must be escaped with a ``\``
798 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
799 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
800 :file:`F:\\filename` then you would use ``filename=F\:\filename``.
802 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
803 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
804 Note: Windows and FreeBSD prevent write access to areas
805 of the disk containing in-use data (e.g. filesystems).
807 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
808 of the two depends on the read/write direction set.
810 .. option:: filename_format=str
812 If sharing multiple files between jobs, it is usually necessary to have fio
813 generate the exact names that you want. By default, fio will name a file
814 based on the default file format specification of
815 :file:`jobname.jobnumber.filenumber`. With this option, that can be
816 customized. Fio will recognize and replace the following keywords in this
820 The name of the worker thread or process.
822 IP of the fio process when using client/server mode.
824 The incremental number of the worker thread or process.
826 The incremental number of the file for that worker thread or
829 To have dependent jobs share a set of files, this option can be set to have
830 fio generate filenames that are shared between the two. For instance, if
831 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
832 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
833 will be used if no other format specifier is given.
835 If you specify a path then the directories will be created up to the
836 main directory for the file. So for example if you specify
837 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
838 created before the file setup part of the job. If you specify
839 :option:`directory` then the path will be relative that directory,
840 otherwise it is treated as the absolute path.
842 .. option:: unique_filename=bool
844 To avoid collisions between networked clients, fio defaults to prefixing any
845 generated filenames (with a directory specified) with the source of the
846 client connecting. To disable this behavior, set this option to 0.
848 .. option:: opendir=str
850 Recursively open any files below directory `str`. This accepts only a
851 single directory and unlike related options, colons appearing in the
852 path must not be escaped.
854 .. option:: lockfile=str
856 Fio defaults to not locking any files before it does I/O to them. If a file
857 or file descriptor is shared, fio can serialize I/O to that file to make the
858 end result consistent. This is usual for emulating real workloads that share
859 files. The lock modes are:
862 No locking. The default.
864 Only one thread or process may do I/O at a time, excluding all
867 Read-write locking on the file. Many readers may
868 access the file at the same time, but writes get exclusive access.
870 .. option:: nrfiles=int
872 Number of files to use for this job. Defaults to 1. The size of files
873 will be :option:`size` divided by this unless explicit size is specified by
874 :option:`filesize`. Files are created for each thread separately, and each
875 file will have a file number within its name by default, as explained in
876 :option:`filename` section.
879 .. option:: openfiles=int
881 Number of files to keep open at the same time. Defaults to the same as
882 :option:`nrfiles`, can be set smaller to limit the number simultaneous
885 .. option:: file_service_type=str
887 Defines how fio decides which file from a job to service next. The following
891 Choose a file at random.
894 Round robin over opened files. This is the default.
897 Finish one file before moving on to the next. Multiple files can
898 still be open depending on :option:`openfiles`.
901 Use a *Zipf* distribution to decide what file to access.
904 Use a *Pareto* distribution to decide what file to access.
907 Use a *Gaussian* (normal) distribution to decide what file to
913 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
914 tell fio how many I/Os to issue before switching to a new file. For example,
915 specifying ``file_service_type=random:8`` would cause fio to issue
916 8 I/Os before selecting a new file at random. For the non-uniform
917 distributions, a floating point postfix can be given to influence how the
918 distribution is skewed. See :option:`random_distribution` for a description
919 of how that would work.
921 .. option:: ioscheduler=str
923 Attempt to switch the device hosting the file to the specified I/O scheduler
926 .. option:: create_serialize=bool
928 If true, serialize the file creation for the jobs. This may be handy to
929 avoid interleaving of data files, which may greatly depend on the filesystem
930 used and even the number of processors in the system. Default: true.
932 .. option:: create_fsync=bool
934 :manpage:`fsync(2)` the data file after creation. This is the default.
936 .. option:: create_on_open=bool
938 If true, don't pre-create files but allow the job's open() to create a file
939 when it's time to do I/O. Default: false -- pre-create all necessary files
942 .. option:: create_only=bool
944 If true, fio will only run the setup phase of the job. If files need to be
945 laid out or updated on disk, only that will be done -- the actual job contents
946 are not executed. Default: false.
948 .. option:: allow_file_create=bool
950 If true, fio is permitted to create files as part of its workload. If this
951 option is false, then fio will error out if
952 the files it needs to use don't already exist. Default: true.
954 .. option:: allow_mounted_write=bool
956 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
957 to what appears to be a mounted device or partition. This should help catch
958 creating inadvertently destructive tests, not realizing that the test will
959 destroy data on the mounted file system. Note that some platforms don't allow
960 writing against a mounted device regardless of this option. Default: false.
962 .. option:: pre_read=bool
964 If this is given, files will be pre-read into memory before starting the
965 given I/O operation. This will also clear the :option:`invalidate` flag,
966 since it is pointless to pre-read and then drop the cache. This will only
967 work for I/O engines that are seek-able, since they allow you to read the
968 same data multiple times. Thus it will not work on non-seekable I/O engines
969 (e.g. network, splice). Default: false.
971 .. option:: unlink=bool
973 Unlink the job files when done. Not the default, as repeated runs of that
974 job would then waste time recreating the file set again and again. Default:
977 .. option:: unlink_each_loop=bool
979 Unlink job files after each iteration or loop. Default: false.
981 .. option:: zonemode=str
986 The :option:`zonerange`, :option:`zonesize`,
987 :option `zonecapacity` and option:`zoneskip`
988 parameters are ignored.
990 I/O happens in a single zone until
991 :option:`zonesize` bytes have been transferred.
992 After that number of bytes has been
993 transferred processing of the next zone
994 starts. :option `zonecapacity` is ignored.
996 Zoned block device mode. I/O happens
997 sequentially in each zone, even if random I/O
998 has been selected. Random I/O happens across
999 all zones instead of being restricted to a
1000 single zone. The :option:`zoneskip` parameter
1001 is ignored. :option:`zonerange` and
1002 :option:`zonesize` must be identical.
1003 Trim is handled using a zone reset operation.
1004 Trim only considers non-empty sequential write
1005 required and sequential write preferred zones.
1007 .. option:: zonerange=int
1009 Size of a single zone. See also :option:`zonesize` and
1012 .. option:: zonesize=int
1014 For :option:`zonemode` =strided, this is the number of bytes to
1015 transfer before skipping :option:`zoneskip` bytes. If this parameter
1016 is smaller than :option:`zonerange` then only a fraction of each zone
1017 with :option:`zonerange` bytes will be accessed. If this parameter is
1018 larger than :option:`zonerange` then each zone will be accessed
1019 multiple times before skipping to the next zone.
1021 For :option:`zonemode` =zbd, this is the size of a single zone. The
1022 :option:`zonerange` parameter is ignored in this mode.
1025 .. option:: zonecapacity=int
1027 For :option:`zonemode` =zbd, this defines the capacity of a single zone,
1028 which is the accessible area starting from the zone start address.
1029 This parameter only applies when using :option:`zonemode` =zbd in
1030 combination with regular block devices. If not specified it defaults to
1031 the zone size. If the target device is a zoned block device, the zone
1032 capacity is obtained from the device information and this option is
1035 .. option:: zoneskip=int
1037 For :option:`zonemode` =strided, the number of bytes to skip after
1038 :option:`zonesize` bytes of data have been transferred. This parameter
1039 must be zero for :option:`zonemode` =zbd.
1041 .. option:: read_beyond_wp=bool
1043 This parameter applies to :option:`zonemode` =zbd only.
1045 Zoned block devices are block devices that consist of multiple zones.
1046 Each zone has a type, e.g. conventional or sequential. A conventional
1047 zone can be written at any offset that is a multiple of the block
1048 size. Sequential zones must be written sequentially. The position at
1049 which a write must occur is called the write pointer. A zoned block
1050 device can be either drive managed, host managed or host aware. For
1051 host managed devices the host must ensure that writes happen
1052 sequentially. Fio recognizes host managed devices and serializes
1053 writes to sequential zones for these devices.
1055 If a read occurs in a sequential zone beyond the write pointer then
1056 the zoned block device will complete the read without reading any data
1057 from the storage medium. Since such reads lead to unrealistically high
1058 bandwidth and IOPS numbers fio only reads beyond the write pointer if
1059 explicitly told to do so. Default: false.
1061 .. option:: max_open_zones=int
1063 When a zone of a zoned block device is partially written (i.e. not all
1064 sectors of the zone have been written), the zone is in one of three
1065 conditions: 'implicit open', 'explicit open' or 'closed'. Zoned block
1066 devices may have a limit called 'max_open_zones' (same name as the
1067 parameter) on the total number of zones that can simultaneously be in
1068 the 'implicit open' or 'explicit open' conditions. Zoned block devices
1069 may have another limit called 'max_active_zones', on the total number of
1070 zones that can simultaneously be in the three conditions. The
1071 :option:`max_open_zones` parameter limits the number of zones to which
1072 write commands are issued by all fio jobs, that is, limits the number of
1073 zones that will be in the conditions. When the device has the
1074 max_open_zones limit and does not have the max_active_zones limit, the
1075 :option:`max_open_zones` parameter limits the number of zones in the two
1076 open conditions up to the limit. In this case, fio includes zones in the
1077 two open conditions to the write target zones at fio start. When the
1078 device has both the max_open_zones and the max_active_zones limits, the
1079 :option:`max_open_zones` parameter limits the number of zones in the
1080 three conditions up to the limit. In this case, fio includes zones in
1081 the three conditions to the write target zones at fio start.
1083 This parameter is relevant only if the :option:`zonemode` =zbd is used.
1084 The default value is always equal to the max_open_zones limit of the
1085 target zoned block device and a value higher than this limit cannot be
1086 specified by users unless the option :option:`ignore_zone_limits` is
1087 specified. When :option:`ignore_zone_limits` is specified or the target
1088 device does not have the max_open_zones limit, :option:`max_open_zones`
1089 can specify 0 to disable any limit on the number of zones that can be
1090 simultaneously written to by all jobs.
1092 .. option:: job_max_open_zones=int
1094 In the same manner as :option:`max_open_zones`, limit the number of open
1095 zones per fio job, that is, the number of zones that a single job can
1096 simultaneously write to. A value of zero indicates no limit.
1099 .. option:: ignore_zone_limits=bool
1101 If this option is used, fio will ignore the maximum number of open
1102 zones limit of the zoned block device in use, thus allowing the
1103 option :option:`max_open_zones` value to be larger than the device
1104 reported limit. Default: false.
1106 .. option:: zone_reset_threshold=float
1108 A number between zero and one that indicates the ratio of written bytes
1109 in the zones with write pointers in the IO range to the size of the IO
1110 range. When current ratio is above this ratio, zones are reset
1111 periodically as :option:`zone_reset_frequency` specifies. If there are
1112 multiple jobs when using this option, the IO range for all write jobs
1115 .. option:: zone_reset_frequency=float
1117 A number between zero and one that indicates how often a zone reset
1118 should be issued if the zone reset threshold has been exceeded. A zone
1119 reset is submitted after each (1 / zone_reset_frequency) write
1120 requests. This and the previous parameter can be used to simulate
1121 garbage collection activity.
1127 .. option:: direct=bool
1129 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
1130 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
1131 ioengines don't support direct I/O. Default: false.
1133 .. option:: buffered=bool
1135 If value is true, use buffered I/O. This is the opposite of the
1136 :option:`direct` option. Defaults to true.
1138 .. option:: readwrite=str, rw=str
1140 Type of I/O pattern. Accepted values are:
1147 Sequential trims (Linux block devices and SCSI
1148 character devices only).
1154 Random trims (Linux block devices and SCSI
1155 character devices only).
1157 Sequential mixed reads and writes.
1159 Random mixed reads and writes.
1161 Sequential trim+write sequences. Blocks will be trimmed first,
1162 then the same blocks will be written to. So if ``io_size=64K``
1163 is specified, Fio will trim a total of 64K bytes and also
1164 write 64K bytes on the same trimmed blocks. This behaviour
1165 will be consistent with ``number_ios`` or other Fio options
1166 limiting the total bytes or number of I/O's.
1168 Like trimwrite, but uses random offsets rather
1169 than sequential writes.
1171 Fio defaults to read if the option is not specified. For the mixed I/O
1172 types, the default is to split them 50/50. For certain types of I/O the
1173 result may still be skewed a bit, since the speed may be different.
1175 It is possible to specify the number of I/Os to do before getting a new
1176 offset by appending ``:<nr>`` to the end of the string given. For a
1177 random read, it would look like ``rw=randread:8`` for passing in an offset
1178 modifier with a value of 8. If the suffix is used with a sequential I/O
1179 pattern, then the *<nr>* value specified will be **added** to the generated
1180 offset for each I/O turning sequential I/O into sequential I/O with holes.
1181 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1182 the :option:`rw_sequencer` option.
1184 .. option:: rw_sequencer=str
1186 If an offset modifier is given by appending a number to the ``rw=<str>``
1187 line, then this option controls how that number modifies the I/O offset
1188 being generated. Accepted values are:
1191 Generate sequential offset.
1193 Generate the same offset.
1195 ``sequential`` is only useful for random I/O, where fio would normally
1196 generate a new random offset for every I/O. If you append e.g. 8 to
1197 randread, i.e. ``rw=randread:8`` you would get a new random offset for
1198 every 8 I/Os. The result would be a sequence of 8 sequential offsets
1199 with a random starting point. However this behavior may change if a
1200 sequential I/O reaches end of the file. As sequential I/O is already
1201 sequential, setting ``sequential`` for that would not result in any
1202 difference. ``identical`` behaves in a similar fashion, except it sends
1203 the same offset 8 number of times before generating a new offset.
1208 rw_sequencer=sequential
1211 The generated sequence of offsets will look like this:
1212 4k, 8k, 12k, 16k, 20k, 24k, 28k, 32k, 92k, 96k, 100k, 104k, 108k,
1213 112k, 116k, 120k, 48k, 52k ...
1218 rw_sequencer=identical
1221 The generated sequence of offsets will look like this:
1222 4k, 4k, 4k, 4k, 4k, 4k, 4k, 4k, 92k, 92k, 92k, 92k, 92k, 92k, 92k, 92k,
1225 .. option:: unified_rw_reporting=str
1227 Fio normally reports statistics on a per data direction basis, meaning that
1228 reads, writes, and trims are accounted and reported separately. This option
1229 determines whether fio reports the results normally, summed together, or as
1231 Accepted values are:
1234 Normal statistics reporting.
1237 Statistics are summed per data direction and reported together.
1240 Statistics are reported normally, followed by the mixed statistics.
1243 Backward-compatible alias for **none**.
1246 Backward-compatible alias for **mixed**.
1251 .. option:: randrepeat=bool
1253 Seed all random number generators in a predictable way so the pattern
1254 is repeatable across runs. Default: true.
1256 .. option:: allrandrepeat=bool
1258 Alias for :option:`randrepeat`. Default: true.
1260 .. option:: randseed=int
1262 Seed the random number generators based on this seed value, to be able to
1263 control what sequence of output is being generated. If not set, the random
1264 sequence depends on the :option:`randrepeat` setting.
1266 .. option:: fallocate=str
1268 Whether pre-allocation is performed when laying down files.
1269 Accepted values are:
1272 Do not pre-allocate space.
1275 Use a platform's native pre-allocation call but fall back to
1276 **none** behavior if it fails/is not implemented.
1279 Pre-allocate via :manpage:`posix_fallocate(3)`.
1282 Pre-allocate via :manpage:`fallocate(2)` with
1283 FALLOC_FL_KEEP_SIZE set.
1286 Extend file to final size via :manpage:`ftruncate(2)`
1287 instead of allocating.
1290 Backward-compatible alias for **none**.
1293 Backward-compatible alias for **posix**.
1295 May not be available on all supported platforms. **keep** is only available
1296 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1297 because ZFS doesn't support pre-allocation. Default: **native** if any
1298 pre-allocation methods except **truncate** are available, **none** if not.
1300 Note that using **truncate** on Windows will interact surprisingly
1301 with non-sequential write patterns. When writing to a file that has
1302 been extended by setting the end-of-file information, Windows will
1303 backfill the unwritten portion of the file up to that offset with
1304 zeroes before issuing the new write. This means that a single small
1305 write to the end of an extended file will stall until the entire
1306 file has been filled with zeroes.
1308 .. option:: fadvise_hint=str
1310 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1311 advise the kernel on what I/O patterns are likely to be issued.
1312 Accepted values are:
1315 Backwards-compatible hint for "no hint".
1318 Backwards compatible hint for "advise with fio workload type". This
1319 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1320 for a sequential workload.
1323 Advise using **FADV_SEQUENTIAL**.
1326 Advise using **FADV_RANDOM**.
1329 Advise using **FADV_NOREUSE**. This may be a no-op on older Linux
1330 kernels. Since Linux 6.3, it provides a hint to the LRU algorithm.
1331 See the :manpage:`posix_fadvise(2)` man page.
1333 .. option:: write_hint=str
1335 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1336 from a write. Only supported on Linux, as of version 4.13. Accepted
1340 No particular life time associated with this file.
1343 Data written to this file has a short life time.
1346 Data written to this file has a medium life time.
1349 Data written to this file has a long life time.
1352 Data written to this file has a very long life time.
1354 The values are all relative to each other, and no absolute meaning
1355 should be associated with them.
1357 .. option:: offset=int
1359 Start I/O at the provided offset in the file, given as either a fixed size in
1360 bytes, zones or a percentage. If a percentage is given, the generated offset will be
1361 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1362 provided. Data before the given offset will not be touched. This
1363 effectively caps the file size at `real_size - offset`. Can be combined with
1364 :option:`size` to constrain the start and end range of the I/O workload.
1365 A percentage can be specified by a number between 1 and 100 followed by '%',
1366 for example, ``offset=20%`` to specify 20%. In ZBD mode, value can be set as
1367 number of zones using 'z'.
1369 .. option:: offset_align=int
1371 If set to non-zero value, the byte offset generated by a percentage ``offset``
1372 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1373 offset is aligned to the minimum block size.
1375 .. option:: offset_increment=int
1377 If this is provided, then the real offset becomes `offset + offset_increment
1378 * thread_number`, where the thread number is a counter that starts at 0 and
1379 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1380 specified). This option is useful if there are several jobs which are
1381 intended to operate on a file in parallel disjoint segments, with even
1382 spacing between the starting points. Percentages can be used for this option.
1383 If a percentage is given, the generated offset will be aligned to the minimum
1384 ``blocksize`` or to the value of ``offset_align`` if provided. In ZBD mode, value can
1385 also be set as number of zones using 'z'.
1387 .. option:: number_ios=int
1389 Fio will normally perform I/Os until it has exhausted the size of the region
1390 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1391 condition). With this setting, the range/size can be set independently of
1392 the number of I/Os to perform. When fio reaches this number, it will exit
1393 normally and report status. Note that this does not extend the amount of I/O
1394 that will be done, it will only stop fio if this condition is met before
1395 other end-of-job criteria.
1397 .. option:: fsync=int
1399 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1400 the dirty data for every number of blocks given. For example, if you give 32
1401 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1402 using non-buffered I/O, we may not sync the file. The exception is the sg
1403 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1404 means fio does not periodically issue and wait for a sync to complete. Also
1405 see :option:`end_fsync` and :option:`fsync_on_close`.
1407 .. option:: fdatasync=int
1409 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1410 not metadata blocks. In Windows, DragonFlyBSD or OSX there is no
1411 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1412 Defaults to 0, which means fio does not periodically issue and wait for a
1413 data-only sync to complete.
1415 .. option:: write_barrier=int
1417 Make every `N-th` write a barrier write.
1419 .. option:: sync_file_range=str:int
1421 Use :manpage:`sync_file_range(2)` for every `int` number of write
1422 operations. Fio will track range of writes that have happened since the last
1423 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1426 SYNC_FILE_RANGE_WAIT_BEFORE
1428 SYNC_FILE_RANGE_WRITE
1430 SYNC_FILE_RANGE_WAIT_AFTER
1432 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1433 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1434 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1437 .. option:: overwrite=bool
1439 If true, writes to a file will always overwrite existing data. If the file
1440 doesn't already exist, it will be created before the write phase begins. If
1441 the file exists and is large enough for the specified write phase, nothing
1442 will be done. Default: false.
1444 .. option:: end_fsync=bool
1446 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1449 .. option:: fsync_on_close=bool
1451 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1452 from :option:`end_fsync` in that it will happen on every file close, not
1453 just at the end of the job. Default: false.
1455 .. option:: rwmixread=int
1457 Percentage of a mixed workload that should be reads. Default: 50.
1459 .. option:: rwmixwrite=int
1461 Percentage of a mixed workload that should be writes. If both
1462 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1463 add up to 100%, the latter of the two will be used to override the
1464 first. This may interfere with a given rate setting, if fio is asked to
1465 limit reads or writes to a certain rate. If that is the case, then the
1466 distribution may be skewed. Default: 50.
1468 .. option:: random_distribution=str:float[:float][,str:float][,str:float]
1470 By default, fio will use a completely uniform random distribution when asked
1471 to perform random I/O. Sometimes it is useful to skew the distribution in
1472 specific ways, ensuring that some parts of the data is more hot than others.
1473 fio includes the following distribution models:
1476 Uniform random distribution
1485 Normal (Gaussian) distribution
1488 Zoned random distribution
1491 Zone absolute random distribution
1493 When using a **zipf** or **pareto** distribution, an input value is also
1494 needed to define the access pattern. For **zipf**, this is the `Zipf
1495 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1496 program, :command:`fio-genzipf`, that can be used visualize what the given input
1497 values will yield in terms of hit rates. If you wanted to use **zipf** with
1498 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1499 option. If a non-uniform model is used, fio will disable use of the random
1500 map. For the **normal** distribution, a normal (Gaussian) deviation is
1501 supplied as a value between 0 and 100.
1503 The second, optional float is allowed for **pareto**, **zipf** and **normal** distributions.
1504 It allows one to set base of distribution in non-default place, giving more control
1505 over most probable outcome. This value is in range [0-1] which maps linearly to
1506 range of possible random values.
1507 Defaults are: random for **pareto** and **zipf**, and 0.5 for **normal**.
1508 If you wanted to use **zipf** with a `theta` of 1.2 centered on 1/4 of allowed value range,
1509 you would use ``random_distribution=zipf:1.2:0.25``.
1511 For a **zoned** distribution, fio supports specifying percentages of I/O
1512 access that should fall within what range of the file or device. For
1513 example, given a criteria of:
1515 * 60% of accesses should be to the first 10%
1516 * 30% of accesses should be to the next 20%
1517 * 8% of accesses should be to the next 30%
1518 * 2% of accesses should be to the next 40%
1520 we can define that through zoning of the random accesses. For the above
1521 example, the user would do::
1523 random_distribution=zoned:60/10:30/20:8/30:2/40
1525 A **zoned_abs** distribution works exactly like the **zoned**, except
1526 that it takes absolute sizes. For example, let's say you wanted to
1527 define access according to the following criteria:
1529 * 60% of accesses should be to the first 20G
1530 * 30% of accesses should be to the next 100G
1531 * 10% of accesses should be to the next 500G
1533 we can define an absolute zoning distribution with:
1535 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1537 For both **zoned** and **zoned_abs**, fio supports defining up to
1540 Similarly to how :option:`bssplit` works for setting ranges and
1541 percentages of block sizes. Like :option:`bssplit`, it's possible to
1542 specify separate zones for reads, writes, and trims. If just one set
1543 is given, it'll apply to all of them. This goes for both **zoned**
1544 **zoned_abs** distributions.
1546 .. option:: percentage_random=int[,int][,int]
1548 For a random workload, set how big a percentage should be random. This
1549 defaults to 100%, in which case the workload is fully random. It can be set
1550 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1551 sequential. Any setting in between will result in a random mix of sequential
1552 and random I/O, at the given percentages. Comma-separated values may be
1553 specified for reads, writes, and trims as described in :option:`blocksize`.
1555 .. option:: norandommap
1557 Normally fio will cover every block of the file when doing random I/O. If
1558 this option is given, fio will just get a new random offset without looking
1559 at past I/O history. This means that some blocks may not be read or written,
1560 and that some blocks may be read/written more than once. If this option is
1561 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1562 only intact blocks are verified, i.e., partially-overwritten blocks are
1563 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1564 the same block to be overwritten, which can cause verification errors. Either
1565 do not use norandommap in this case, or also use the lfsr random generator.
1567 .. option:: softrandommap=bool
1569 See :option:`norandommap`. If fio runs with the random block map enabled and
1570 it fails to allocate the map, if this option is set it will continue without
1571 a random block map. As coverage will not be as complete as with random maps,
1572 this option is disabled by default.
1574 .. option:: random_generator=str
1576 Fio supports the following engines for generating I/O offsets for random I/O:
1579 Strong 2^88 cycle random number generator.
1581 Linear feedback shift register generator.
1583 Strong 64-bit 2^258 cycle random number generator.
1585 **tausworthe** is a strong random number generator, but it requires tracking
1586 on the side if we want to ensure that blocks are only read or written
1587 once. **lfsr** guarantees that we never generate the same offset twice, and
1588 it's also less computationally expensive. It's not a true random generator,
1589 however, though for I/O purposes it's typically good enough. **lfsr** only
1590 works with single block sizes, not with workloads that use multiple block
1591 sizes. If used with such a workload, fio may read or write some blocks
1592 multiple times. The default value is **tausworthe**, unless the required
1593 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1594 selected automatically.
1600 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1602 The block size in bytes used for I/O units. Default: 4096. A single value
1603 applies to reads, writes, and trims. Comma-separated values may be
1604 specified for reads, writes, and trims. A value not terminated in a comma
1605 applies to subsequent types.
1610 means 256k for reads, writes and trims.
1613 means 8k for reads, 32k for writes and trims.
1616 means 8k for reads, 32k for writes, and default for trims.
1619 means default for reads, 8k for writes and trims.
1622 means default for reads, 8k for writes, and default for trims.
1624 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1626 A range of block sizes in bytes for I/O units. The issued I/O unit will
1627 always be a multiple of the minimum size, unless
1628 :option:`blocksize_unaligned` is set.
1630 Comma-separated ranges may be specified for reads, writes, and trims as
1631 described in :option:`blocksize`.
1633 Example: ``bsrange=1k-4k,2k-8k``.
1635 .. option:: bssplit=str[,str][,str]
1637 Sometimes you want even finer grained control of the block sizes
1638 issued, not just an even split between them. This option allows you to
1639 weight various block sizes, so that you are able to define a specific
1640 amount of block sizes issued. The format for this option is::
1642 bssplit=blocksize/percentage:blocksize/percentage
1644 for as many block sizes as needed. So if you want to define a workload
1645 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1648 bssplit=4k/10:64k/50:32k/40
1650 Ordering does not matter. If the percentage is left blank, fio will
1651 fill in the remaining values evenly. So a bssplit option like this one::
1653 bssplit=4k/50:1k/:32k/
1655 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1656 add up to 100, if bssplit is given a range that adds up to more, it
1659 Comma-separated values may be specified for reads, writes, and trims as
1660 described in :option:`blocksize`.
1662 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1663 having 90% 4k writes and 10% 8k writes, you would specify::
1665 bssplit=2k/50:4k/50,4k/90:8k/10
1667 Fio supports defining up to 64 different weights for each data
1670 .. option:: blocksize_unaligned, bs_unaligned
1672 If set, fio will issue I/O units with any size within
1673 :option:`blocksize_range`, not just multiples of the minimum size. This
1674 typically won't work with direct I/O, as that normally requires sector
1677 .. option:: bs_is_seq_rand=bool
1679 If this option is set, fio will use the normal read,write blocksize settings
1680 as sequential,random blocksize settings instead. Any random read or write
1681 will use the WRITE blocksize settings, and any sequential read or write will
1682 use the READ blocksize settings.
1684 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1686 Boundary to which fio will align random I/O units. Default:
1687 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1688 I/O, though it usually depends on the hardware block size. This option is
1689 mutually exclusive with using a random map for files, so it will turn off
1690 that option. Comma-separated values may be specified for reads, writes, and
1691 trims as described in :option:`blocksize`.
1697 .. option:: zero_buffers
1699 Initialize buffers with all zeros. Default: fill buffers with random data.
1701 .. option:: refill_buffers
1703 If this option is given, fio will refill the I/O buffers on every
1704 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1705 naturally. Defaults to being unset i.e., the buffer is only filled at
1706 init time and the data in it is reused when possible but if any of
1707 :option:`verify`, :option:`buffer_compress_percentage` or
1708 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1709 automatically enabled.
1711 .. option:: scramble_buffers=bool
1713 If :option:`refill_buffers` is too costly and the target is using data
1714 deduplication, then setting this option will slightly modify the I/O buffer
1715 contents to defeat normal de-dupe attempts. This is not enough to defeat
1716 more clever block compression attempts, but it will stop naive dedupe of
1717 blocks. Default: true.
1719 .. option:: buffer_compress_percentage=int
1721 If this is set, then fio will attempt to provide I/O buffer content
1722 (on WRITEs) that compresses to the specified level. Fio does this by
1723 providing a mix of random data followed by fixed pattern data. The
1724 fixed pattern is either zeros, or the pattern specified by
1725 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1726 might skew the compression ratio slightly. Setting
1727 `buffer_compress_percentage` to a value other than 100 will also
1728 enable :option:`refill_buffers` in order to reduce the likelihood that
1729 adjacent blocks are so similar that they over compress when seen
1730 together. See :option:`buffer_compress_chunk` for how to set a finer or
1731 coarser granularity for the random/fixed data region. Defaults to unset
1732 i.e., buffer data will not adhere to any compression level.
1734 .. option:: buffer_compress_chunk=int
1736 This setting allows fio to manage how big the random/fixed data region
1737 is when using :option:`buffer_compress_percentage`. When
1738 `buffer_compress_chunk` is set to some non-zero value smaller than the
1739 block size, fio can repeat the random/fixed region throughout the I/O
1740 buffer at the specified interval (which particularly useful when
1741 bigger block sizes are used for a job). When set to 0, fio will use a
1742 chunk size that matches the block size resulting in a single
1743 random/fixed region within the I/O buffer. Defaults to 512. When the
1744 unit is omitted, the value is interpreted in bytes.
1746 .. option:: buffer_pattern=str
1748 If set, fio will fill the I/O buffers with this pattern or with the contents
1749 of a file. If not set, the contents of I/O buffers are defined by the other
1750 options related to buffer contents. The setting can be any pattern of bytes,
1751 and can be prefixed with 0x for hex values. It may also be a string, where
1752 the string must then be wrapped with ``""``. Or it may also be a filename,
1753 where the filename must be wrapped with ``''`` in which case the file is
1754 opened and read. Note that not all the file contents will be read if that
1755 would cause the buffers to overflow. So, for example::
1757 buffer_pattern='filename'
1761 buffer_pattern="abcd"
1769 buffer_pattern=0xdeadface
1771 Also you can combine everything together in any order::
1773 buffer_pattern=0xdeadface"abcd"-12'filename'
1775 .. option:: dedupe_percentage=int
1777 If set, fio will generate this percentage of identical buffers when
1778 writing. These buffers will be naturally dedupable. The contents of the
1779 buffers depend on what other buffer compression settings have been set. It's
1780 possible to have the individual buffers either fully compressible, or not at
1781 all -- this option only controls the distribution of unique buffers. Setting
1782 this option will also enable :option:`refill_buffers` to prevent every buffer
1785 .. option:: dedupe_mode=str
1787 If ``dedupe_percentage=<int>`` is given, then this option controls how fio
1788 generates the dedupe buffers.
1791 Generate dedupe buffers by repeating previous writes
1793 Generate dedupe buffers from working set
1795 ``repeat`` is the default option for fio. Dedupe buffers are generated
1796 by repeating previous unique write.
1798 ``working_set`` is a more realistic workload.
1799 With ``working_set``, ``dedupe_working_set_percentage=<int>`` should be provided.
1800 Given that, fio will use the initial unique write buffers as its working set.
1801 Upon deciding to dedupe, fio will randomly choose a buffer from the working set.
1802 Note that by using ``working_set`` the dedupe percentage will converge
1803 to the desired over time while ``repeat`` maintains the desired percentage
1806 .. option:: dedupe_working_set_percentage=int
1808 If ``dedupe_mode=<str>`` is set to ``working_set``, then this controls
1809 the percentage of size of the file or device used as the buffers
1810 fio will choose to generate the dedupe buffers from
1812 Note that size needs to be explicitly provided and only 1 file per
1815 .. option:: dedupe_global=bool
1817 This controls whether the deduplication buffers will be shared amongst
1818 all jobs that have this option set. The buffers are spread evenly between
1821 .. option:: invalidate=bool
1823 Invalidate the buffer/page cache parts of the files to be used prior to
1824 starting I/O if the platform and file type support it. Defaults to true.
1825 This will be ignored if :option:`pre_read` is also specified for the
1828 .. option:: sync=str
1830 Whether, and what type, of synchronous I/O to use for writes. The allowed
1834 Do not use synchronous IO, the default.
1840 Use synchronous file IO. For the majority of I/O engines,
1841 this means using O_SYNC.
1847 Use synchronous data IO. For the majority of I/O engines,
1848 this means using O_DSYNC.
1851 .. option:: iomem=str, mem=str
1853 Fio can use various types of memory as the I/O unit buffer. The allowed
1857 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1861 Use shared memory as the buffers. Allocated through
1862 :manpage:`shmget(2)`.
1865 Same as shm, but use huge pages as backing.
1868 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1869 be file backed if a filename is given after the option. The format
1870 is `mem=mmap:/path/to/file`.
1873 Use a memory mapped huge file as the buffer backing. Append filename
1874 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1877 Same as mmap, but use a MMAP_SHARED mapping.
1880 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1881 The :option:`ioengine` must be `rdma`.
1883 The area allocated is a function of the maximum allowed bs size for the job,
1884 multiplied by the I/O depth given. Note that for **shmhuge** and
1885 **mmaphuge** to work, the system must have free huge pages allocated. This
1886 can normally be checked and set by reading/writing
1887 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1888 is 2 or 4MiB in size depending on the platform. So to calculate the
1889 number of huge pages you need for a given job file, add up the I/O
1890 depth of all jobs (normally one unless :option:`iodepth` is used) and
1891 multiply by the maximum bs set. Then divide that number by the huge
1892 page size. You can see the size of the huge pages in
1893 :file:`/proc/meminfo`. If no huge pages are allocated by having a
1894 non-zero number in `nr_hugepages`, using **mmaphuge** or **shmhuge**
1895 will fail. Also see :option:`hugepage-size`.
1897 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1898 should point there. So if it's mounted in :file:`/huge`, you would use
1899 `mem=mmaphuge:/huge/somefile`.
1901 .. option:: iomem_align=int, mem_align=int
1903 This indicates the memory alignment of the I/O memory buffers. Note that
1904 the given alignment is applied to the first I/O unit buffer, if using
1905 :option:`iodepth` the alignment of the following buffers are given by the
1906 :option:`bs` used. In other words, if using a :option:`bs` that is a
1907 multiple of the page sized in the system, all buffers will be aligned to
1908 this value. If using a :option:`bs` that is not page aligned, the alignment
1909 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1912 .. option:: hugepage-size=int
1914 Defines the size of a huge page. Must at least be equal to the system
1915 setting, see :file:`/proc/meminfo` and
1916 :file:`/sys/kernel/mm/hugepages/`. Defaults to 2 or 4MiB depending on
1917 the platform. Should probably always be a multiple of megabytes, so
1918 using ``hugepage-size=Xm`` is the preferred way to set this to avoid
1919 setting a non-pow-2 bad value.
1921 .. option:: lockmem=int
1923 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1924 simulate a smaller amount of memory. The amount specified is per worker.
1930 .. option:: size=int
1932 The total size of file I/O for each thread of this job. Fio will run until
1933 this many bytes has been transferred, unless runtime is altered by other means
1934 such as (1) :option:`runtime`, (2) :option:`io_size` (3) :option:`number_ios`,
1935 (4) gaps/holes while doing I/O's such as ``rw=read:16K``, or (5) sequential
1936 I/O reaching end of the file which is possible when :option:`percentage_random`
1938 Fio will divide this size between the available files determined by options
1939 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1940 specified by the job. If the result of division happens to be 0, the size is
1941 set to the physical size of the given files or devices if they exist.
1942 If this option is not specified, fio will use the full size of the given
1943 files or devices. If the files do not exist, size must be given. It is also
1944 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1945 given, fio will use 20% of the full size of the given files or devices.
1946 In ZBD mode, value can also be set as number of zones using 'z'.
1947 Can be combined with :option:`offset` to constrain the start and end range
1948 that I/O will be done within.
1950 .. option:: io_size=int, io_limit=int
1952 Normally fio operates within the region set by :option:`size`, which means
1953 that the :option:`size` option sets both the region and size of I/O to be
1954 performed. Sometimes that is not what you want. With this option, it is
1955 possible to define just the amount of I/O that fio should do. For instance,
1956 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1957 will perform I/O within the first 20GiB but exit when 5GiB have been
1958 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1959 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1960 the 0..20GiB region.
1962 .. option:: filesize=irange(int)
1964 Individual file sizes. May be a range, in which case fio will select sizes for
1965 files at random within the given range. If not given, each created file is the
1966 same size. This option overrides :option:`size` in terms of file size, i.e. if
1967 :option:`filesize` is specified then :option:`size` becomes merely the default
1968 for :option:`io_size` and has no effect at all if :option:`io_size` is set
1971 .. option:: file_append=bool
1973 Perform I/O after the end of the file. Normally fio will operate within the
1974 size of a file. If this option is set, then fio will append to the file
1975 instead. This has identical behavior to setting :option:`offset` to the size
1976 of a file. This option is ignored on non-regular files.
1978 .. option:: fill_device=bool, fill_fs=bool
1980 Sets size to something really large and waits for ENOSPC (no space left on
1981 device) or EDQUOT (disk quota exceeded)
1982 as the terminating condition. Only makes sense with sequential
1983 write. For a read workload, the mount point will be filled first then I/O
1984 started on the result. This option doesn't make sense if operating on a raw
1985 device node, since the size of that is already known by the file system.
1986 Additionally, writing beyond end-of-device will not return ENOSPC there.
1992 .. option:: ioengine=str
1994 Defines how the job issues I/O to the file. The following types are defined:
1997 Basic :manpage:`read(2)` or :manpage:`write(2)`
1998 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1999 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
2002 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
2003 all supported operating systems except for Windows.
2006 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
2007 queuing by coalescing adjacent I/Os into a single submission.
2010 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
2013 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
2016 Fast Linux native asynchronous I/O. Supports async IO
2017 for both direct and buffered IO.
2018 This engine defines engine specific options.
2021 Fast Linux native asynchronous I/O for pass through commands.
2022 This engine defines engine specific options.
2025 Linux native asynchronous I/O. Note that Linux may only support
2026 queued behavior with non-buffered I/O (set ``direct=1`` or
2028 This engine defines engine specific options.
2031 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
2032 :manpage:`aio_write(3)`.
2035 Solaris native asynchronous I/O.
2038 Windows native asynchronous I/O. Default on Windows.
2041 File is memory mapped with :manpage:`mmap(2)` and data copied
2042 to/from using :manpage:`memcpy(3)`.
2045 :manpage:`splice(2)` is used to transfer the data and
2046 :manpage:`vmsplice(2)` to transfer data from user space to the
2050 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
2051 ioctl, or if the target is an sg character device we use
2052 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
2053 I/O. Requires :option:`filename` option to specify either block or
2054 character devices. This engine supports trim operations.
2055 The sg engine includes engine specific options.
2058 Read, write, trim and ZBC/ZAC operations to a zoned
2059 block device using libzbc library. The target can be
2060 either an SG character device or a block device file.
2063 Doesn't transfer any data, just pretends to. This is mainly used to
2064 exercise fio itself and for debugging/testing purposes.
2067 Transfer over the network to given ``host:port``. Depending on the
2068 :option:`protocol` used, the :option:`hostname`, :option:`port`,
2069 :option:`listen` and :option:`filename` options are used to specify
2070 what sort of connection to make, while the :option:`protocol` option
2071 determines which protocol will be used. This engine defines engine
2075 Like **net**, but uses :manpage:`splice(2)` and
2076 :manpage:`vmsplice(2)` to map data and send/receive.
2077 This engine defines engine specific options.
2080 Doesn't transfer any data, but burns CPU cycles according to the
2081 :option:`cpuload`, :option:`cpuchunks` and :option:`cpumode` options.
2082 Setting :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
2083 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
2084 to get desired CPU usage, as the cpuload only loads a
2085 single CPU at the desired rate. A job never finishes unless there is
2086 at least one non-cpuio job.
2087 Setting :option:`cpumode`\=qsort replace the default noop instructions loop
2088 by a qsort algorithm to consume more energy.
2091 The RDMA I/O engine supports both RDMA memory semantics
2092 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
2093 InfiniBand, RoCE and iWARP protocols. This engine defines engine
2097 I/O engine that does regular fallocate to simulate data transfer as
2101 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
2104 does fallocate(,mode = 0).
2107 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
2110 I/O engine that sends :manpage:`ftruncate(2)` operations in response
2111 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
2112 size to the current block offset. :option:`blocksize` is ignored.
2115 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
2116 defragment activity in request to DDIR_WRITE event.
2119 I/O engine supporting direct access to Ceph Reliable Autonomic
2120 Distributed Object Store (RADOS) via librados. This ioengine
2121 defines engine specific options.
2124 I/O engine supporting direct access to Ceph Rados Block Devices
2125 (RBD) via librbd without the need to use the kernel rbd driver. This
2126 ioengine defines engine specific options.
2129 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
2130 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
2132 This engine only supports direct IO of iodepth=1; you need to scale this
2133 via numjobs. blocksize defines the size of the objects to be created.
2135 TRIM is translated to object deletion.
2138 Using GlusterFS libgfapi sync interface to direct access to
2139 GlusterFS volumes without having to go through FUSE. This ioengine
2140 defines engine specific options.
2143 Using GlusterFS libgfapi async interface to direct access to
2144 GlusterFS volumes without having to go through FUSE. This ioengine
2145 defines engine specific options.
2148 Read and write through Hadoop (HDFS). The :option:`filename` option
2149 is used to specify host,port of the hdfs name-node to connect. This
2150 engine interprets offsets a little differently. In HDFS, files once
2151 created cannot be modified so random writes are not possible. To
2152 imitate this the libhdfs engine expects a bunch of small files to be
2153 created over HDFS and will randomly pick a file from them
2154 based on the offset generated by fio backend (see the example
2155 job file to create such files, use ``rw=write`` option). Please
2156 note, it may be necessary to set environment variables to work
2157 with HDFS/libhdfs properly. Each job uses its own connection to
2161 Read, write and erase an MTD character device (e.g.,
2162 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
2163 underlying device type, the I/O may have to go in a certain pattern,
2164 e.g., on NAND, writing sequentially to erase blocks and discarding
2165 before overwriting. The `trimwrite` mode works well for this
2169 Read and write using device DAX to a persistent memory device (e.g.,
2170 /dev/dax0.0) through the PMDK libpmem library.
2173 Prefix to specify loading an external I/O engine object file. Append
2174 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
2175 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
2176 absolute or relative. See :file:`engines/skeleton_external.c` for
2177 details of writing an external I/O engine.
2180 Simply create the files and do no I/O to them. You still need to
2181 set `filesize` so that all the accounting still occurs, but no
2182 actual I/O will be done other than creating the file.
2185 Simply do stat() and do no I/O to the file. You need to set 'filesize'
2186 and 'nrfiles', so that files will be created.
2187 This engine is to measure file lookup and meta data access.
2190 Simply delete the files by unlink() and do no I/O to them. You need to set 'filesize'
2191 and 'nrfiles', so that the files will be created.
2192 This engine is to measure file delete.
2195 Read and write using mmap I/O to a file on a filesystem
2196 mounted with DAX on a persistent memory device through the PMDK
2200 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2201 This engine is very basic and issues calls to IME whenever an IO is
2205 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2206 This engine uses iovecs and will try to stack as much IOs as possible
2207 (if the IOs are "contiguous" and the IO depth is not exceeded)
2208 before issuing a call to IME.
2211 Asynchronous read and write using DDN's Infinite Memory Engine (IME).
2212 This engine will try to stack as much IOs as possible by creating
2213 requests for IME. FIO will then decide when to commit these requests.
2216 Read and write iscsi lun with libiscsi.
2219 Read and write a Network Block Device (NBD).
2222 I/O engine supporting libcufile synchronous access to nvidia-fs and a
2223 GPUDirect Storage-supported filesystem. This engine performs
2224 I/O without transferring buffers between user-space and the kernel,
2225 unless :option:`verify` is set or :option:`cuda_io` is `posix`.
2226 :option:`iomem` must not be `cudamalloc`. This ioengine defines
2227 engine specific options.
2230 I/O engine supporting asynchronous read and write operations to the
2231 DAOS File System (DFS) via libdfs.
2234 I/O engine supporting asynchronous read and write operations to
2235 NFS filesystems from userspace via libnfs. This is useful for
2236 achieving higher concurrency and thus throughput than is possible
2240 Execute 3rd party tools. Could be used to perform monitoring during jobs runtime.
2243 I/O engine using the xNVMe C API, for NVMe devices. The xnvme engine provides
2244 flexibility to access GNU/Linux Kernel NVMe driver via libaio, IOCTLs, io_uring,
2245 the SPDK NVMe driver, or your own custom NVMe driver. The xnvme engine includes
2246 engine specific options. (See https://xnvme.io).
2249 Use the libblkio library
2250 (https://gitlab.com/libblkio/libblkio). The specific
2251 *driver* to use must be set using
2252 :option:`libblkio_driver`. If
2253 :option:`mem`/:option:`iomem` is not specified, memory
2254 allocation is delegated to libblkio (and so is
2255 guaranteed to work with the selected *driver*). One
2256 libblkio instance is used per process, so all jobs
2257 setting option :option:`thread` will share a single
2258 instance (with one queue per thread) and must specify
2259 compatible options. Note that some drivers don't allow
2260 several instances to access the same device or file
2261 simultaneously, but allow it for threads.
2263 I/O engine specific parameters
2264 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2266 In addition, there are some parameters which are only valid when a specific
2267 :option:`ioengine` is in use. These are used identically to normal parameters,
2268 with the caveat that when used on the command line, they must come after the
2269 :option:`ioengine` that defines them is selected.
2271 .. option:: cmdprio_percentage=int[,int] : [io_uring] [libaio]
2273 Set the percentage of I/O that will be issued with the highest priority.
2274 Default: 0. A single value applies to reads and writes. Comma-separated
2275 values may be specified for reads and writes. For this option to be
2276 effective, NCQ priority must be supported and enabled, and the :option:`direct`
2277 option must be set. fio must also be run as the root user. Unlike
2278 slat/clat/lat stats, which can be tracked and reported independently, per
2279 priority stats only track and report a single type of latency. By default,
2280 completion latency (clat) will be reported, if :option:`lat_percentiles` is
2281 set, total latency (lat) will be reported.
2283 .. option:: cmdprio_class=int[,int] : [io_uring] [libaio]
2285 Set the I/O priority class to use for I/Os that must be issued with
2286 a priority when :option:`cmdprio_percentage` or
2287 :option:`cmdprio_bssplit` is set. If not specified when
2288 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2289 this defaults to the highest priority class. A single value applies
2290 to reads and writes. Comma-separated values may be specified for
2291 reads and writes. See :manpage:`ionice(1)`. See also the
2292 :option:`prioclass` option.
2294 .. option:: cmdprio_hint=int[,int] : [io_uring] [libaio]
2296 Set the I/O priority hint to use for I/Os that must be issued with
2297 a priority when :option:`cmdprio_percentage` or
2298 :option:`cmdprio_bssplit` is set. If not specified when
2299 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2300 this defaults to 0 (no hint). A single value applies to reads and
2301 writes. Comma-separated values may be specified for reads and writes.
2302 See also the :option:`priohint` option.
2304 .. option:: cmdprio=int[,int] : [io_uring] [libaio]
2306 Set the I/O priority value to use for I/Os that must be issued with
2307 a priority when :option:`cmdprio_percentage` or
2308 :option:`cmdprio_bssplit` is set. If not specified when
2309 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2311 Linux limits us to a positive value between 0 and 7, with 0 being the
2312 highest. A single value applies to reads and writes. Comma-separated
2313 values may be specified for reads and writes. See :manpage:`ionice(1)`.
2314 Refer to an appropriate manpage for other operating systems since
2315 meaning of priority may differ. See also the :option:`prio` option.
2317 .. option:: cmdprio_bssplit=str[,str] : [io_uring] [libaio]
2319 To get a finer control over I/O priority, this option allows
2320 specifying the percentage of IOs that must have a priority set
2321 depending on the block size of the IO. This option is useful only
2322 when used together with the :option:`bssplit` option, that is,
2323 multiple different block sizes are used for reads and writes.
2325 The first accepted format for this option is the same as the format of
2326 the :option:`bssplit` option:
2328 cmdprio_bssplit=blocksize/percentage:blocksize/percentage
2330 In this case, each entry will use the priority class, priority hint
2331 and priority level defined by the options :option:`cmdprio_class`,
2332 :option:`cmdprio` and :option:`cmdprio_hint` respectively.
2334 The second accepted format for this option is:
2336 cmdprio_bssplit=blocksize/percentage/class/level:blocksize/percentage/class/level
2338 In this case, the priority class and priority level is defined inside
2339 each entry. In comparison with the first accepted format, the second
2340 accepted format does not restrict all entries to have the same priority
2341 class and priority level.
2343 The third accepted format for this option is:
2345 cmdprio_bssplit=blocksize/percentage/class/level/hint:...
2347 This is an extension of the second accepted format that allows one to
2348 also specify a priority hint.
2350 For all formats, only the read and write data directions are supported,
2351 values for trim IOs are ignored. This option is mutually exclusive with
2352 the :option:`cmdprio_percentage` option.
2354 .. option:: fixedbufs : [io_uring] [io_uring_cmd]
2356 If fio is asked to do direct IO, then Linux will map pages for each
2357 IO call, and release them when IO is done. If this option is set, the
2358 pages are pre-mapped before IO is started. This eliminates the need to
2359 map and release for each IO. This is more efficient, and reduces the
2362 .. option:: nonvectored=int : [io_uring] [io_uring_cmd]
2364 With this option, fio will use non-vectored read/write commands, where
2365 address must contain the address directly. Default is -1.
2367 .. option:: force_async=int : [io_uring] [io_uring_cmd]
2369 Normal operation for io_uring is to try and issue an sqe as
2370 non-blocking first, and if that fails, execute it in an async manner.
2371 With this option set to N, then every N request fio will ask sqe to
2372 be issued in an async manner. Default is 0.
2374 .. option:: registerfiles : [io_uring] [io_uring_cmd]
2376 With this option, fio registers the set of files being used with the
2377 kernel. This avoids the overhead of managing file counts in the kernel,
2378 making the submission and completion part more lightweight. Required
2379 for the below :option:`sqthread_poll` option.
2381 .. option:: sqthread_poll : [io_uring] [io_uring_cmd] [xnvme]
2383 Normally fio will submit IO by issuing a system call to notify the
2384 kernel of available items in the SQ ring. If this option is set, the
2385 act of submitting IO will be done by a polling thread in the kernel.
2386 This frees up cycles for fio, at the cost of using more CPU in the
2387 system. As submission is just the time it takes to fill in the sqe
2388 entries and any syscall required to wake up the idle kernel thread,
2389 fio will not report submission latencies.
2391 .. option:: sqthread_poll_cpu=int : [io_uring] [io_uring_cmd]
2393 When :option:`sqthread_poll` is set, this option provides a way to
2394 define which CPU should be used for the polling thread.
2396 .. option:: cmd_type=str : [io_uring_cmd]
2398 Specifies the type of uring passthrough command to be used. Supported
2399 value is nvme. Default is nvme.
2403 [io_uring] [io_uring_cmd] [xnvme]
2405 If this option is set, fio will attempt to use polled IO completions.
2406 Normal IO completions generate interrupts to signal the completion of
2407 IO, polled completions do not. Hence they are require active reaping
2408 by the application. The benefits are more efficient IO for high IOPS
2409 scenarios, and lower latencies for low queue depth IO.
2413 Use poll queues. This is incompatible with
2414 :option:`libblkio_wait_mode=eventfd <libblkio_wait_mode>` and
2415 :option:`libblkio_force_enable_completion_eventfd`.
2419 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2424 If this option is set, fio will attempt to use polled IO completions.
2425 This will have a similar effect as (io_uring)hipri. Only SCSI READ and
2426 WRITE commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor
2427 VERIFY). Older versions of the Linux sg driver that do not support
2428 hipri will simply ignore this flag and do normal IO. The Linux SCSI
2429 Low Level Driver (LLD) that "owns" the device also needs to support
2430 hipri (also known as iopoll and mq_poll). The MegaRAID driver is an
2431 example of a SCSI LLD. Default: clear (0) which does normal
2432 (interrupted based) IO.
2434 .. option:: userspace_reap : [libaio]
2436 Normally, with the libaio engine in use, fio will use the
2437 :manpage:`io_getevents(2)` system call to reap newly returned events. With
2438 this flag turned on, the AIO ring will be read directly from user-space to
2439 reap events. The reaping mode is only enabled when polling for a minimum of
2440 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
2442 .. option:: hipri_percentage : [pvsync2]
2444 When hipri is set this determines the probability of a pvsync2 I/O being high
2445 priority. The default is 100%.
2447 .. option:: nowait=bool : [pvsync2] [libaio] [io_uring] [io_uring_cmd]
2449 By default if a request cannot be executed immediately (e.g. resource starvation,
2450 waiting on locks) it is queued and the initiating process will be blocked until
2451 the required resource becomes free.
2453 This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and
2454 the call will return instantly with EAGAIN or a partial result rather than waiting.
2456 It is useful to also use ignore_error=EAGAIN when using this option.
2458 Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2.
2459 They return EOPNOTSUP instead of EAGAIN.
2461 For cached I/O, using this option usually means a request operates only with
2462 cached data. Currently the RWF_NOWAIT flag does not supported for cached write.
2464 For direct I/O, requests will only succeed if cache invalidation isn't required,
2465 file blocks are fully allocated and the disk request could be issued immediately.
2467 .. option:: fdp=bool : [io_uring_cmd] [xnvme]
2469 Enable Flexible Data Placement mode for write commands.
2471 .. option:: fdp_pli_select=str : [io_uring_cmd] [xnvme]
2473 Defines how fio decides which placement ID to use next. The following
2477 Choose a placement ID at random (uniform).
2480 Round robin over available placement IDs. This is the
2483 The available placement ID index/indices is defined by the option
2486 .. option:: fdp_pli=str : [io_uring_cmd] [xnvme]
2488 Select which Placement ID Index/Indicies this job is allowed to use for
2489 writes. By default, the job will cycle through all available Placement
2490 IDs, so use this to isolate these identifiers to specific jobs. If you
2491 want fio to use placement identifier only at indices 0, 2 and 5 specify
2494 .. option:: md_per_io_size=int : [io_uring_cmd]
2496 Size in bytes for separate metadata buffer per IO. Default: 0.
2498 .. option:: pi_act=int : [io_uring_cmd]
2500 Action to take when nvme namespace is formatted with protection
2501 information. If this is set to 1 and namespace is formatted with
2502 metadata size equal to protection information size, fio won't use
2503 separate metadata buffer or extended logical block. If this is set to
2504 1 and namespace is formatted with metadata size greater than protection
2505 information size, fio will not generate or verify the protection
2506 information portion of metadata for write or read case respectively.
2507 If this is set to 0, fio generates protection information for
2508 write case and verifies for read case. Default: 1.
2510 For 16 bit CRC generation fio will use isa-l if available otherwise
2511 it will use the default slower generator.
2512 (see: https://github.com/intel/isa-l)
2514 .. option:: pi_chk=str[,str][,str] : [io_uring_cmd]
2516 Controls the protection information check. This can take one or more
2517 of these values. Default: none.
2520 Enables protection information checking of guard field.
2522 Enables protection information checking of logical block
2523 reference tag field.
2525 Enables protection information checking of application tag field.
2527 .. option:: apptag=int : [io_uring_cmd]
2529 Specifies logical block application tag value, if namespace is
2530 formatted to use end to end protection information. Default: 0x1234.
2532 .. option:: apptag_mask=int : [io_uring_cmd]
2534 Specifies logical block application tag mask value, if namespace is
2535 formatted to use end to end protection information. Default: 0xffff.
2537 .. option:: cpuload=int : [cpuio]
2539 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2540 option when using cpuio I/O engine.
2542 .. option:: cpuchunks=int : [cpuio]
2544 Split the load into cycles of the given time. In microseconds.
2546 .. option:: cpumode=str : [cpuio]
2548 Specify how to stress the CPU. It can take these two values:
2551 This is the default where the CPU executes noop instructions.
2553 Replace the default noop instructions loop with a qsort algorithm to
2554 consume more energy.
2556 .. option:: exit_on_io_done=bool : [cpuio]
2558 Detect when I/O threads are done, then exit.
2560 .. option:: namenode=str : [libhdfs]
2562 The hostname or IP address of a HDFS cluster namenode to contact.
2564 .. option:: port=int
2568 The listening port of the HFDS cluster namenode.
2572 The TCP or UDP port to bind to or connect to. If this is used with
2573 :option:`numjobs` to spawn multiple instances of the same job type, then
2574 this will be the starting port number since fio will use a range of
2579 The port to use for RDMA-CM communication. This should be the same value
2580 on the client and the server side.
2582 .. option:: hostname=str : [netsplice] [net] [rdma]
2584 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
2585 is a TCP listener or UDP reader, the hostname is not used and must be omitted
2586 unless it is a valid UDP multicast address.
2588 .. option:: serverip=str : [librpma_*]
2590 The IP address to be used for RDMA-CM based I/O.
2592 .. option:: direct_write_to_pmem=bool : [librpma_*]
2594 Set to 1 only when Direct Write to PMem from the remote host is possible.
2595 Otherwise, set to 0.
2597 .. option:: busy_wait_polling=bool : [librpma_*_server]
2599 Set to 0 to wait for completion instead of busy-wait polling completion.
2602 .. option:: interface=str : [netsplice] [net]
2604 The IP address of the network interface used to send or receive UDP
2607 .. option:: ttl=int : [netsplice] [net]
2609 Time-to-live value for outgoing UDP multicast packets. Default: 1.
2611 .. option:: nodelay=bool : [netsplice] [net]
2613 Set TCP_NODELAY on TCP connections.
2615 .. option:: protocol=str, proto=str : [netsplice] [net]
2617 The network protocol to use. Accepted values are:
2620 Transmission control protocol.
2622 Transmission control protocol V6.
2624 User datagram protocol.
2626 User datagram protocol V6.
2630 When the protocol is TCP or UDP, the port must also be given, as well as the
2631 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
2632 normal :option:`filename` option should be used and the port is invalid.
2634 .. option:: listen : [netsplice] [net]
2636 For TCP network connections, tell fio to listen for incoming connections
2637 rather than initiating an outgoing connection. The :option:`hostname` must
2638 be omitted if this option is used.
2640 .. option:: pingpong : [netsplice] [net]
2642 Normally a network writer will just continue writing data, and a network
2643 reader will just consume packages. If ``pingpong=1`` is set, a writer will
2644 send its normal payload to the reader, then wait for the reader to send the
2645 same payload back. This allows fio to measure network latencies. The
2646 submission and completion latencies then measure local time spent sending or
2647 receiving, and the completion latency measures how long it took for the
2648 other end to receive and send back. For UDP multicast traffic
2649 ``pingpong=1`` should only be set for a single reader when multiple readers
2650 are listening to the same address.
2652 .. option:: window_size : [netsplice] [net]
2654 Set the desired socket buffer size for the connection.
2656 .. option:: mss : [netsplice] [net]
2658 Set the TCP maximum segment size (TCP_MAXSEG).
2660 .. option:: donorname=str : [e4defrag]
2662 File will be used as a block donor (swap extents between files).
2664 .. option:: inplace=int : [e4defrag]
2666 Configure donor file blocks allocation strategy:
2669 Default. Preallocate donor's file on init.
2671 Allocate space immediately inside defragment event, and free right
2674 .. option:: clustername=str : [rbd,rados]
2676 Specifies the name of the Ceph cluster.
2678 .. option:: rbdname=str : [rbd]
2680 Specifies the name of the RBD.
2682 .. option:: clientname=str : [rbd,rados]
2684 Specifies the username (without the 'client.' prefix) used to access the
2685 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2686 the full *type.id* string. If no type. prefix is given, fio will add
2687 'client.' by default.
2689 .. option:: conf=str : [rados]
2691 Specifies the configuration path of ceph cluster, so conf file does not
2692 have to be /etc/ceph/ceph.conf.
2694 .. option:: busy_poll=bool : [rbd,rados]
2696 Poll store instead of waiting for completion. Usually this provides better
2697 throughput at cost of higher(up to 100%) CPU utilization.
2699 .. option:: touch_objects=bool : [rados]
2701 During initialization, touch (create if do not exist) all objects (files).
2702 Touching all objects affects ceph caches and likely impacts test results.
2705 .. option:: pool=str :
2709 Specifies the name of the Ceph pool containing RBD or RADOS data.
2713 Specify the label or UUID of the DAOS pool to connect to.
2715 .. option:: cont=str : [dfs]
2717 Specify the label or UUID of the DAOS container to open.
2719 .. option:: chunk_size=int
2723 Specify a different chunk size (in bytes) for the dfs file.
2724 Use DAOS container's chunk size by default.
2728 The size of the chunk to use for each file.
2730 .. option:: object_class=str : [dfs]
2732 Specify a different object class for the dfs file.
2733 Use DAOS container's object class by default.
2735 .. option:: skip_bad=bool : [mtd]
2737 Skip operations against known bad blocks.
2739 .. option:: hdfsdirectory : [libhdfs]
2741 libhdfs will create chunk in this HDFS directory.
2743 .. option:: verb=str : [rdma]
2745 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2746 values are write, read, send and recv. These correspond to the equivalent
2747 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2748 specified on the client side of the connection. See the examples folder.
2750 .. option:: bindname=str : [rdma]
2752 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2753 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2754 will be passed into the rdma_bind_addr() function and on the client site it
2755 will be used in the rdma_resolve_add() function. This can be useful when
2756 multiple paths exist between the client and the server or in certain loopback
2759 .. option:: stat_type=str : [filestat]
2761 Specify stat system call type to measure lookup/getattr performance.
2762 Default is **stat** for :manpage:`stat(2)`.
2764 .. option:: readfua=bool : [sg]
2766 With readfua option set to 1, read operations include
2767 the force unit access (fua) flag. Default is 0.
2769 .. option:: writefua=bool : [sg]
2771 With writefua option set to 1, write operations include
2772 the force unit access (fua) flag. Default is 0.
2774 .. option:: sg_write_mode=str : [sg]
2776 Specify the type of write commands to issue. This option can take three values:
2779 This is the default where write opcodes are issued as usual.
2780 **write_and_verify**
2781 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2782 directs the device to carry out a medium verification with no data
2783 comparison. The writefua option is ignored with this selection.
2785 This option is deprecated. Use write_and_verify instead.
2787 Issue WRITE SAME commands. This transfers a single block to the device
2788 and writes this same block of data to a contiguous sequence of LBAs
2789 beginning at the specified offset. fio's block size parameter specifies
2790 the amount of data written with each command. However, the amount of data
2791 actually transferred to the device is equal to the device's block
2792 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2793 write 16 sectors with each command. fio will still generate 8k of data
2794 for each command but only the first 512 bytes will be used and
2795 transferred to the device. The writefua option is ignored with this
2798 This option is deprecated. Use write_same instead.
2800 Issue WRITE SAME(16) commands as above but with the No Data Output
2801 Buffer (NDOB) bit set. No data will be transferred to the device with
2802 this bit set. Data written will be a pre-determined pattern such as
2805 Issue WRITE STREAM(16) commands. Use the **stream_id** option to specify
2806 the stream identifier.
2807 **verify_bytchk_00**
2808 Issue VERIFY commands with BYTCHK set to 00. This directs the
2809 device to carry out a medium verification with no data comparison.
2810 **verify_bytchk_01**
2811 Issue VERIFY commands with BYTCHK set to 01. This directs the device to
2812 compare the data on the device with the data transferred to the device.
2813 **verify_bytchk_11**
2814 Issue VERIFY commands with BYTCHK set to 11. This transfers a
2815 single block to the device and compares the contents of this block with the
2816 data on the device beginning at the specified offset. fio's block size
2817 parameter specifies the total amount of data compared with this command.
2818 However, only one block (sector) worth of data is transferred to the device.
2819 This is similar to the WRITE SAME command except that data is compared instead
2822 .. option:: stream_id=int : [sg]
2824 Set the stream identifier for WRITE STREAM commands. If this is set to 0 (which is not
2825 a valid stream identifier) fio will open a stream and then close it when done. Default
2828 .. option:: http_host=str : [http]
2830 Hostname to connect to. For S3, this could be the bucket hostname.
2831 Default is **localhost**
2833 .. option:: http_user=str : [http]
2835 Username for HTTP authentication.
2837 .. option:: http_pass=str : [http]
2839 Password for HTTP authentication.
2841 .. option:: https=str : [http]
2843 Enable HTTPS instead of http. *on* enables HTTPS; *insecure*
2844 will enable HTTPS, but disable SSL peer verification (use with
2845 caution!). Default is **off**
2847 .. option:: http_mode=str : [http]
2849 Which HTTP access mode to use: *webdav*, *swift*, or *s3*.
2850 Default is **webdav**
2852 .. option:: http_s3_region=str : [http]
2854 The S3 region/zone string.
2855 Default is **us-east-1**
2857 .. option:: http_s3_key=str : [http]
2861 .. option:: http_s3_keyid=str : [http]
2863 The S3 key/access id.
2865 .. option:: http_s3_sse_customer_key=str : [http]
2867 The encryption customer key in SSE server side.
2869 .. option:: http_s3_sse_customer_algorithm=str : [http]
2871 The encryption customer algorithm in SSE server side.
2872 Default is **AES256**
2874 .. option:: http_s3_storage_class=str : [http]
2876 Which storage class to access. User-customizable settings.
2877 Default is **STANDARD**
2879 .. option:: http_swift_auth_token=str : [http]
2881 The Swift auth token. See the example configuration file on how
2884 .. option:: http_verbose=int : [http]
2886 Enable verbose requests from libcurl. Useful for debugging. 1
2887 turns on verbose logging from libcurl, 2 additionally enables
2888 HTTP IO tracing. Default is **0**
2890 .. option:: uri=str : [nbd]
2892 Specify the NBD URI of the server to test. The string
2893 is a standard NBD URI
2894 (see https://github.com/NetworkBlockDevice/nbd/tree/master/doc).
2895 Example URIs: nbd://localhost:10809
2896 nbd+unix:///?socket=/tmp/socket
2897 nbds://tlshost/exportname
2899 .. option:: gpu_dev_ids=str : [libcufile]
2901 Specify the GPU IDs to use with CUDA. This is a colon-separated list of
2902 int. GPUs are assigned to workers roundrobin. Default is 0.
2904 .. option:: cuda_io=str : [libcufile]
2906 Specify the type of I/O to use with CUDA. Default is **cufile**.
2909 Use libcufile and nvidia-fs. This option performs I/O directly
2910 between a GPUDirect Storage filesystem and GPU buffers,
2911 avoiding use of a bounce buffer. If :option:`verify` is set,
2912 cudaMemcpy is used to copy verificaton data between RAM and GPU.
2913 Verification data is copied from RAM to GPU before a write
2914 and from GPU to RAM after a read. :option:`direct` must be 1.
2916 Use POSIX to perform I/O with a RAM buffer, and use cudaMemcpy
2917 to transfer data between RAM and the GPUs. Data is copied from
2918 GPU to RAM before a write and copied from RAM to GPU after a
2919 read. :option:`verify` does not affect use of cudaMemcpy.
2921 .. option:: nfs_url=str : [nfs]
2923 URL in libnfs format, eg nfs://<server|ipv4|ipv6>/path[?arg=val[&arg=val]*]
2924 Refer to the libnfs README for more details.
2926 .. option:: program=str : [exec]
2928 Specify the program to execute.
2930 .. option:: arguments=str : [exec]
2932 Specify arguments to pass to program.
2933 Some special variables can be expanded to pass fio's job details to the program.
2936 Replaced by the duration of the job in seconds.
2938 Replaced by the name of the job.
2940 .. option:: grace_time=int : [exec]
2942 Specify the time between the SIGTERM and SIGKILL signals. Default is 1 second.
2944 .. option:: std_redirect=bool : [exec]
2946 If set, stdout and stderr streams are redirected to files named from the job name. Default is true.
2948 .. option:: xnvme_async=str : [xnvme]
2950 Select the xnvme async command interface. This can take these values.
2953 This is default and use to emulate asynchronous I/O by using a
2954 single thread to create a queue pair on top of a synchronous
2955 I/O interface using the NVMe driver IOCTL.
2957 Emulate an asynchronous I/O interface with a pool of userspace
2958 threads on top of a synchronous I/O interface using the NVMe
2959 driver IOCTL. By default four threads are used.
2961 Linux native asynchronous I/O interface which supports both
2962 direct and buffered I/O.
2964 Fast Linux native asynchronous I/O interface for NVMe pass
2965 through commands. This only works with NVMe character device
2968 Use Linux aio for Asynchronous I/O.
2970 Use the posix asynchronous I/O interface to perform one or
2971 more I/O operations asynchronously.
2973 Use the user-space VFIO-based backend, implemented using
2974 libvfn instead of SPDK.
2976 Do not transfer any data; just pretend to. This is mainly used
2977 for introspective performance evaluation.
2979 .. option:: xnvme_sync=str : [xnvme]
2981 Select the xnvme synchronous command interface. This can take these values.
2984 This is default and uses Linux NVMe Driver ioctl() for
2987 This supports regular as well as vectored pread() and pwrite()
2990 This is the same as psync except that it also supports zone
2991 management commands using Linux block layer IOCTLs.
2993 .. option:: xnvme_admin=str : [xnvme]
2995 Select the xnvme admin command interface. This can take these values.
2998 This is default and uses linux NVMe Driver ioctl() for admin
3001 Use Linux Block Layer ioctl() and sysfs for admin commands.
3003 .. option:: xnvme_dev_nsid=int : [xnvme]
3005 xnvme namespace identifier for userspace NVMe driver, SPDK or vfio.
3007 .. option:: xnvme_dev_subnqn=str : [xnvme]
3009 Sets the subsystem NQN for fabrics. This is for xNVMe to utilize a
3010 fabrics target with multiple systems.
3012 .. option:: xnvme_mem=str : [xnvme]
3014 Select the xnvme memory backend. This can take these values.
3017 This is the default posix memory backend for linux NVMe driver.
3019 Use hugepages, instead of existing posix memory backend. The
3020 memory backend uses hugetlbfs. This require users to allocate
3021 hugepages, mount hugetlbfs and set an environment variable for
3024 Uses SPDK's memory allocator.
3026 Uses libvfn's memory allocator. This also specifies the use
3027 of libvfn backend instead of SPDK.
3029 .. option:: xnvme_iovec=int : [xnvme]
3031 If this option is set. xnvme will use vectored read/write commands.
3033 .. option:: libblkio_driver=str : [libblkio]
3035 The libblkio *driver* to use. Different drivers access devices through
3036 different underlying interfaces. Available drivers depend on the
3037 libblkio version in use and are listed at
3038 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3040 .. option:: libblkio_path=str : [libblkio]
3042 Sets the value of the driver-specific "path" property before connecting
3043 the libblkio instance, which identifies the target device or file on
3044 which to perform I/O. Its exact semantics are driver-dependent and not
3045 all drivers may support it; see
3046 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3048 .. option:: libblkio_pre_connect_props=str : [libblkio]
3050 A colon-separated list of additional libblkio properties to be set after
3051 creating but before connecting the libblkio instance. Each property must
3052 have the format ``<name>=<value>``. Colons can be escaped as ``\:``.
3053 These are set after the engine sets any other properties, so those can
3054 be overridden. Available properties depend on the libblkio version in use
3056 https://libblkio.gitlab.io/libblkio/blkio.html#properties
3058 .. option:: libblkio_num_entries=int : [libblkio]
3060 Sets the value of the driver-specific "num-entries" property before
3061 starting the libblkio instance. Its exact semantics are driver-dependent
3062 and not all drivers may support it; see
3063 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3065 .. option:: libblkio_queue_size=int : [libblkio]
3067 Sets the value of the driver-specific "queue-size" property before
3068 starting the libblkio instance. Its exact semantics are driver-dependent
3069 and not all drivers may support it; see
3070 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3072 .. option:: libblkio_pre_start_props=str : [libblkio]
3074 A colon-separated list of additional libblkio properties to be set after
3075 connecting but before starting the libblkio instance. Each property must
3076 have the format ``<name>=<value>``. Colons can be escaped as ``\:``.
3077 These are set after the engine sets any other properties, so those can
3078 be overridden. Available properties depend on the libblkio version in use
3080 https://libblkio.gitlab.io/libblkio/blkio.html#properties
3082 .. option:: libblkio_vectored : [libblkio]
3084 Submit vectored read and write requests.
3086 .. option:: libblkio_write_zeroes_on_trim : [libblkio]
3088 Submit trims as "write zeroes" requests instead of discard requests.
3090 .. option:: libblkio_wait_mode=str : [libblkio]
3092 How to wait for completions:
3095 Use a blocking call to ``blkioq_do_io()``.
3097 Use a blocking call to ``read()`` on the completion eventfd.
3099 Use a busy loop with a non-blocking call to ``blkioq_do_io()``.
3101 .. option:: libblkio_force_enable_completion_eventfd : [libblkio]
3103 Enable the queue's completion eventfd even when unused. This may impact
3104 performance. The default is to enable it only if
3105 :option:`libblkio_wait_mode=eventfd <libblkio_wait_mode>`.
3107 .. option:: no_completion_thread : [windowsaio]
3109 Avoid using a separate thread for completion polling.
3114 .. option:: iodepth=int
3116 Number of I/O units to keep in flight against the file. Note that
3117 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
3118 for small degrees when :option:`verify_async` is in use). Even async
3119 engines may impose OS restrictions causing the desired depth not to be
3120 achieved. This may happen on Linux when using libaio and not setting
3121 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
3122 eye on the I/O depth distribution in the fio output to verify that the
3123 achieved depth is as expected. Default: 1.
3125 .. option:: iodepth_batch_submit=int, iodepth_batch=int
3127 This defines how many pieces of I/O to submit at once. It defaults to 1
3128 which means that we submit each I/O as soon as it is available, but can be
3129 raised to submit bigger batches of I/O at the time. If it is set to 0 the
3130 :option:`iodepth` value will be used.
3132 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
3134 This defines how many pieces of I/O to retrieve at once. It defaults to 1
3135 which means that we'll ask for a minimum of 1 I/O in the retrieval process
3136 from the kernel. The I/O retrieval will go on until we hit the limit set by
3137 :option:`iodepth_low`. If this variable is set to 0, then fio will always
3138 check for completed events before queuing more I/O. This helps reduce I/O
3139 latency, at the cost of more retrieval system calls.
3141 .. option:: iodepth_batch_complete_max=int
3143 This defines maximum pieces of I/O to retrieve at once. This variable should
3144 be used along with :option:`iodepth_batch_complete_min`\=int variable,
3145 specifying the range of min and max amount of I/O which should be
3146 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
3151 iodepth_batch_complete_min=1
3152 iodepth_batch_complete_max=<iodepth>
3154 which means that we will retrieve at least 1 I/O and up to the whole
3155 submitted queue depth. If none of I/O has been completed yet, we will wait.
3159 iodepth_batch_complete_min=0
3160 iodepth_batch_complete_max=<iodepth>
3162 which means that we can retrieve up to the whole submitted queue depth, but
3163 if none of I/O has been completed yet, we will NOT wait and immediately exit
3164 the system call. In this example we simply do polling.
3166 .. option:: iodepth_low=int
3168 The low water mark indicating when to start filling the queue
3169 again. Defaults to the same as :option:`iodepth`, meaning that fio will
3170 attempt to keep the queue full at all times. If :option:`iodepth` is set to
3171 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
3172 16 requests, it will let the depth drain down to 4 before starting to fill
3175 .. option:: serialize_overlap=bool
3177 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
3178 When two or more I/Os are submitted simultaneously, there is no guarantee that
3179 the I/Os will be processed or completed in the submitted order. Further, if
3180 two or more of those I/Os are writes, any overlapping region between them can
3181 become indeterminate/undefined on certain storage. These issues can cause
3182 verification to fail erratically when at least one of the racing I/Os is
3183 changing data and the overlapping region has a non-zero size. Setting
3184 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
3185 serializing in-flight I/Os that have a non-zero overlap. Note that setting
3186 this option can reduce both performance and the :option:`iodepth` achieved.
3188 This option only applies to I/Os issued for a single job except when it is
3189 enabled along with :option:`io_submit_mode`\=offload. In offload mode, fio
3190 will check for overlap among all I/Os submitted by offload jobs with :option:`serialize_overlap`
3195 .. option:: io_submit_mode=str
3197 This option controls how fio submits the I/O to the I/O engine. The default
3198 is `inline`, which means that the fio job threads submit and reap I/O
3199 directly. If set to `offload`, the job threads will offload I/O submission
3200 to a dedicated pool of I/O threads. This requires some coordination and thus
3201 has a bit of extra overhead, especially for lower queue depth I/O where it
3202 can increase latencies. The benefit is that fio can manage submission rates
3203 independently of the device completion rates. This avoids skewed latency
3204 reporting if I/O gets backed up on the device side (the coordinated omission
3205 problem). Note that this option cannot reliably be used with async IO
3212 .. option:: thinkcycles=int
3214 Stall the job for the specified number of cycles after an I/O has completed before
3215 issuing the next. May be used to simulate processing being done by an application.
3216 This is not taken into account for the time to be waited on for :option:`thinktime`.
3217 Might not have any effect on some platforms, this can be checked by trying a setting
3218 a high enough amount of thinkcycles.
3220 .. option:: thinktime=time
3222 Stall the job for the specified period of time after an I/O has completed before issuing the
3223 next. May be used to simulate processing being done by an application.
3224 When the unit is omitted, the value is interpreted in microseconds. See
3225 :option:`thinktime_blocks`, :option:`thinktime_iotime` and :option:`thinktime_spin`.
3227 .. option:: thinktime_spin=time
3229 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
3230 something with the data received, before falling back to sleeping for the
3231 rest of the period specified by :option:`thinktime`. When the unit is
3232 omitted, the value is interpreted in microseconds.
3234 .. option:: thinktime_blocks=int
3236 Only valid if :option:`thinktime` is set - control how many blocks to issue,
3237 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
3238 fio wait :option:`thinktime` usecs after every block. This effectively makes any
3239 queue depth setting redundant, since no more than 1 I/O will be queued
3240 before we have to complete it and do our :option:`thinktime`. In other words, this
3241 setting effectively caps the queue depth if the latter is larger.
3243 .. option:: thinktime_blocks_type=str
3245 Only valid if :option:`thinktime` is set - control how :option:`thinktime_blocks`
3246 triggers. The default is `complete`, which triggers thinktime when fio completes
3247 :option:`thinktime_blocks` blocks. If this is set to `issue`, then the trigger happens
3250 .. option:: thinktime_iotime=time
3252 Only valid if :option:`thinktime` is set - control :option:`thinktime`
3253 interval by time. The :option:`thinktime` stall is repeated after IOs
3254 are executed for :option:`thinktime_iotime`. For example,
3255 ``--thinktime_iotime=9s --thinktime=1s`` repeat 10-second cycle with IOs
3256 for 9 seconds and stall for 1 second. When the unit is omitted,
3257 :option:`thinktime_iotime` is interpreted as a number of seconds. If
3258 this option is used together with :option:`thinktime_blocks`, the
3259 :option:`thinktime` stall is repeated after :option:`thinktime_iotime`
3260 or after :option:`thinktime_blocks` IOs, whichever happens first.
3262 .. option:: rate=int[,int][,int]
3264 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
3265 suffix rules apply. Comma-separated values may be specified for reads,
3266 writes, and trims as described in :option:`blocksize`.
3268 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
3269 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
3270 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
3271 latter will only limit reads.
3273 .. option:: rate_min=int[,int][,int]
3275 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
3276 to meet this requirement will cause the job to exit. Comma-separated values
3277 may be specified for reads, writes, and trims as described in
3278 :option:`blocksize`.
3280 .. option:: rate_iops=int[,int][,int]
3282 Cap the bandwidth to this number of IOPS. Basically the same as
3283 :option:`rate`, just specified independently of bandwidth. If the job is
3284 given a block size range instead of a fixed value, the smallest block size
3285 is used as the metric. Comma-separated values may be specified for reads,
3286 writes, and trims as described in :option:`blocksize`.
3288 .. option:: rate_iops_min=int[,int][,int]
3290 If fio doesn't meet this rate of I/O, it will cause the job to exit.
3291 Comma-separated values may be specified for reads, writes, and trims as
3292 described in :option:`blocksize`.
3294 .. option:: rate_process=str
3296 This option controls how fio manages rated I/O submissions. The default is
3297 `linear`, which submits I/O in a linear fashion with fixed delays between
3298 I/Os that gets adjusted based on I/O completion rates. If this is set to
3299 `poisson`, fio will submit I/O based on a more real world random request
3300 flow, known as the Poisson process
3301 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
3302 10^6 / IOPS for the given workload.
3304 .. option:: rate_ignore_thinktime=bool
3306 By default, fio will attempt to catch up to the specified rate setting,
3307 if any kind of thinktime setting was used. If this option is set, then
3308 fio will ignore the thinktime and continue doing IO at the specified
3309 rate, instead of entering a catch-up mode after thinktime is done.
3311 .. option:: rate_cycle=int
3313 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
3314 of milliseconds. Defaults to 1000.
3320 .. option:: latency_target=time
3322 If set, fio will attempt to find the max performance point that the given
3323 workload will run at while maintaining a latency below this target. When
3324 the unit is omitted, the value is interpreted in microseconds. See
3325 :option:`latency_window` and :option:`latency_percentile`.
3327 .. option:: latency_window=time
3329 Used with :option:`latency_target` to specify the sample window that the job
3330 is run at varying queue depths to test the performance. When the unit is
3331 omitted, the value is interpreted in microseconds.
3333 .. option:: latency_percentile=float
3335 The percentage of I/Os that must fall within the criteria specified by
3336 :option:`latency_target` and :option:`latency_window`. If not set, this
3337 defaults to 100.0, meaning that all I/Os must be equal or below to the value
3338 set by :option:`latency_target`.
3340 .. option:: latency_run=bool
3342 Used with :option:`latency_target`. If false (default), fio will find
3343 the highest queue depth that meets :option:`latency_target` and exit. If
3344 true, fio will continue running and try to meet :option:`latency_target`
3345 by adjusting queue depth.
3347 .. option:: max_latency=time[,time][,time]
3349 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
3350 maximum latency. When the unit is omitted, the value is interpreted in
3351 microseconds. Comma-separated values may be specified for reads, writes,
3352 and trims as described in :option:`blocksize`.
3358 .. option:: write_iolog=str
3360 Write the issued I/O patterns to the specified file. See
3361 :option:`read_iolog`. Specify a separate file for each job, otherwise the
3362 iologs will be interspersed and the file may be corrupt. This file will
3363 be opened in append mode.
3365 .. option:: read_iolog=str
3367 Open an iolog with the specified filename and replay the I/O patterns it
3368 contains. This can be used to store a workload and replay it sometime
3369 later. The iolog given may also be a blktrace binary file, which allows fio
3370 to replay a workload captured by :command:`blktrace`. See
3371 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
3372 replay, the file needs to be turned into a blkparse binary data file first
3373 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
3374 You can specify a number of files by separating the names with a ':'
3375 character. See the :option:`filename` option for information on how to
3376 escape ':' characters within the file names. These files will
3377 be sequentially assigned to job clones created by :option:`numjobs`.
3378 '-' is a reserved name, meaning read from stdin, notably if
3379 :option:`filename` is set to '-' which means stdin as well, then
3380 this flag can't be set to '-'.
3382 .. option:: read_iolog_chunked=bool
3384 Determines how iolog is read. If false(default) entire :option:`read_iolog`
3385 will be read at once. If selected true, input from iolog will be read
3386 gradually. Useful when iolog is very large, or it is generated.
3388 .. option:: merge_blktrace_file=str
3390 When specified, rather than replaying the logs passed to :option:`read_iolog`,
3391 the logs go through a merge phase which aggregates them into a single
3392 blktrace. The resulting file is then passed on as the :option:`read_iolog`
3393 parameter. The intention here is to make the order of events consistent.
3394 This limits the influence of the scheduler compared to replaying multiple
3395 blktraces via concurrent jobs.
3397 .. option:: merge_blktrace_scalars=float_list
3399 This is a percentage based option that is index paired with the list of
3400 files passed to :option:`read_iolog`. When merging is performed, scale
3401 the time of each event by the corresponding amount. For example,
3402 ``--merge_blktrace_scalars="50:100"`` runs the first trace in halftime
3403 and the second trace in realtime. This knob is separately tunable from
3404 :option:`replay_time_scale` which scales the trace during runtime and
3405 does not change the output of the merge unlike this option.
3407 .. option:: merge_blktrace_iters=float_list
3409 This is a whole number option that is index paired with the list of files
3410 passed to :option:`read_iolog`. When merging is performed, run each trace
3411 for the specified number of iterations. For example,
3412 ``--merge_blktrace_iters="2:1"`` runs the first trace for two iterations
3413 and the second trace for one iteration.
3415 .. option:: replay_no_stall=bool
3417 When replaying I/O with :option:`read_iolog` the default behavior is to
3418 attempt to respect the timestamps within the log and replay them with the
3419 appropriate delay between IOPS. By setting this variable fio will not
3420 respect the timestamps and attempt to replay them as fast as possible while
3421 still respecting ordering. The result is the same I/O pattern to a given
3422 device, but different timings.
3424 .. option:: replay_time_scale=int
3426 When replaying I/O with :option:`read_iolog`, fio will honor the
3427 original timing in the trace. With this option, it's possible to scale
3428 the time. It's a percentage option, if set to 50 it means run at 50%
3429 the original IO rate in the trace. If set to 200, run at twice the
3430 original IO rate. Defaults to 100.
3432 .. option:: replay_redirect=str
3434 While replaying I/O patterns using :option:`read_iolog` the default behavior
3435 is to replay the IOPS onto the major/minor device that each IOP was recorded
3436 from. This is sometimes undesirable because on a different machine those
3437 major/minor numbers can map to a different device. Changing hardware on the
3438 same system can also result in a different major/minor mapping.
3439 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
3440 device regardless of the device it was recorded
3441 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
3442 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
3443 multiple devices will be replayed onto a single device, if the trace
3444 contains multiple devices. If you want multiple devices to be replayed
3445 concurrently to multiple redirected devices you must blkparse your trace
3446 into separate traces and replay them with independent fio invocations.
3447 Unfortunately this also breaks the strict time ordering between multiple
3450 .. option:: replay_align=int
3452 Force alignment of the byte offsets in a trace to this value. The value
3453 must be a power of 2.
3455 .. option:: replay_scale=int
3457 Scale byte offsets down by this factor when replaying traces. Should most
3458 likely use :option:`replay_align` as well.
3460 .. option:: replay_skip=str
3462 Sometimes it's useful to skip certain IO types in a replay trace.
3463 This could be, for instance, eliminating the writes in the trace.
3464 Or not replaying the trims/discards, if you are redirecting to
3465 a device that doesn't support them. This option takes a comma
3466 separated list of read, write, trim, sync.
3469 Threads, processes and job synchronization
3470 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3474 Fio defaults to creating jobs by using fork, however if this option is
3475 given, fio will create jobs by using POSIX Threads' function
3476 :manpage:`pthread_create(3)` to create threads instead.
3478 .. option:: wait_for=str
3480 If set, the current job won't be started until all workers of the specified
3481 waitee job are done.
3483 ``wait_for`` operates on the job name basis, so there are a few
3484 limitations. First, the waitee must be defined prior to the waiter job
3485 (meaning no forward references). Second, if a job is being referenced as a
3486 waitee, it must have a unique name (no duplicate waitees).
3488 .. option:: nice=int
3490 Run the job with the given nice value. See man :manpage:`nice(2)`.
3492 On Windows, values less than -15 set the process class to "High"; -1 through
3493 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
3496 .. option:: prio=int
3498 Set the I/O priority value of this job. Linux limits us to a positive value
3499 between 0 and 7, with 0 being the highest. See man
3500 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
3501 systems since meaning of priority may differ. For per-command priority
3502 setting, see I/O engine specific :option:`cmdprio_percentage` and
3503 :option:`cmdprio` options.
3505 .. option:: prioclass=int
3507 Set the I/O priority class. See man :manpage:`ionice(1)`. For per-command
3508 priority setting, see I/O engine specific :option:`cmdprio_percentage`
3509 and :option:`cmdprio_class` options.
3511 .. option:: priohint=int
3513 Set the I/O priority hint. This is only applicable to platforms that
3514 support I/O priority classes and to devices with features controlled
3515 through priority hints, e.g. block devices supporting command duration
3516 limits, or CDL. CDL is a way to indicate the desired maximum latency
3517 of I/Os so that the device can optimize its internal command scheduling
3518 according to the latency limits indicated by the user.
3520 For per-I/O priority hint setting, see the I/O engine specific
3521 :option:`cmdprio_hint` option.
3523 .. option:: cpus_allowed=str
3525 Controls the same options as :option:`cpumask`, but accepts a textual
3526 specification of the permitted CPUs instead and CPUs are indexed from 0. So
3527 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
3528 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
3529 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
3531 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
3532 processor group will be used and affinity settings are inherited from the
3533 system. An fio build configured to target Windows 7 makes options that set
3534 CPUs processor group aware and values will set both the processor group
3535 and a CPU from within that group. For example, on a system where processor
3536 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
3537 values between 0 and 39 will bind CPUs from processor group 0 and
3538 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
3539 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
3540 single ``cpus_allowed`` option must be from the same processor group. For
3541 Windows fio builds not built for Windows 7, CPUs will only be selected from
3542 (and be relative to) whatever processor group fio happens to be running in
3543 and CPUs from other processor groups cannot be used.
3545 .. option:: cpus_allowed_policy=str
3547 Set the policy of how fio distributes the CPUs specified by
3548 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
3551 All jobs will share the CPU set specified.
3553 Each job will get a unique CPU from the CPU set.
3555 **shared** is the default behavior, if the option isn't specified. If
3556 **split** is specified, then fio will assign one cpu per job. If not
3557 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
3560 .. option:: cpumask=int
3562 Set the CPU affinity of this job. The parameter given is a bit mask of
3563 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
3564 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
3565 :manpage:`sched_setaffinity(2)`. This may not work on all supported
3566 operating systems or kernel versions. This option doesn't work well for a
3567 higher CPU count than what you can store in an integer mask, so it can only
3568 control cpus 1-32. For boxes with larger CPU counts, use
3569 :option:`cpus_allowed`.
3571 .. option:: numa_cpu_nodes=str
3573 Set this job running on specified NUMA nodes' CPUs. The arguments allow
3574 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
3575 NUMA options support, fio must be built on a system with libnuma-dev(el)
3578 .. option:: numa_mem_policy=str
3580 Set this job's memory policy and corresponding NUMA nodes. Format of the
3585 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
3586 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
3587 policies, no node needs to be specified. For ``prefer``, only one node is
3588 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
3589 follows: a comma delimited list of numbers, A-B ranges, or `all`.
3591 .. option:: cgroup=str
3593 Add job to this control group. If it doesn't exist, it will be created. The
3594 system must have a mounted cgroup blkio mount point for this to work. If
3595 your system doesn't have it mounted, you can do so with::
3597 # mount -t cgroup -o blkio none /cgroup
3599 .. option:: cgroup_weight=int
3601 Set the weight of the cgroup to this value. See the documentation that comes
3602 with the kernel, allowed values are in the range of 100..1000.
3604 .. option:: cgroup_nodelete=bool
3606 Normally fio will delete the cgroups it has created after the job
3607 completion. To override this behavior and to leave cgroups around after the
3608 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
3609 to inspect various cgroup files after job completion. Default: false.
3611 .. option:: flow_id=int
3613 The ID of the flow. If not specified, it defaults to being a global
3614 flow. See :option:`flow`.
3616 .. option:: flow=int
3618 Weight in token-based flow control. If this value is used, then fio
3619 regulates the activity between two or more jobs sharing the same
3620 flow_id. Fio attempts to keep each job activity proportional to other
3621 jobs' activities in the same flow_id group, with respect to requested
3622 weight per job. That is, if one job has `flow=3', another job has
3623 `flow=2' and another with `flow=1`, then there will be a roughly 3:2:1
3624 ratio in how much one runs vs the others.
3626 .. option:: flow_sleep=int
3628 The period of time, in microseconds, to wait after the flow counter
3629 has exceeded its proportion before retrying operations.
3631 .. option:: stonewall, wait_for_previous
3633 Wait for preceding jobs in the job file to exit, before starting this
3634 one. Can be used to insert serialization points in the job file. A stone
3635 wall also implies starting a new reporting group, see
3636 :option:`group_reporting`.
3640 By default, fio will continue running all other jobs when one job finishes.
3641 Sometimes this is not the desired action. Setting ``exitall`` will instead
3642 make fio terminate all jobs in the same group, as soon as one job of that
3645 .. option:: exit_what=str
3647 By default, fio will continue running all other jobs when one job finishes.
3648 Sometimes this is not the desired action. Setting ``exitall`` will
3649 instead make fio terminate all jobs in the same group. The option
3650 ``exit_what`` allows one to control which jobs get terminated when ``exitall``
3651 is enabled. The default is ``group`` and does not change the behaviour of
3652 ``exitall``. The setting ``all`` terminates all jobs. The setting ``stonewall``
3653 terminates all currently running jobs across all groups and continues execution
3654 with the next stonewalled group.
3656 .. option:: exec_prerun=str
3658 Before running this job, issue the command specified through
3659 :manpage:`system(3)`. Output is redirected in a file called
3660 :file:`jobname.prerun.txt`.
3662 .. option:: exec_postrun=str
3664 After the job completes, issue the command specified though
3665 :manpage:`system(3)`. Output is redirected in a file called
3666 :file:`jobname.postrun.txt`.
3670 Instead of running as the invoking user, set the user ID to this value
3671 before the thread/process does any work.
3675 Set group ID, see :option:`uid`.
3681 .. option:: verify_only
3683 Do not perform specified workload, only verify data still matches previous
3684 invocation of this workload. This option allows one to check data multiple
3685 times at a later date without overwriting it. This option makes sense only
3686 for workloads that write data, and does not support workloads with the
3687 :option:`time_based` option set.
3689 .. option:: do_verify=bool
3691 Run the verify phase after a write phase. Only valid if :option:`verify` is
3694 .. option:: verify=str
3696 If writing to a file, fio can verify the file contents after each iteration
3697 of the job. Each verification method also implies verification of special
3698 header, which is written to the beginning of each block. This header also
3699 includes meta information, like offset of the block, block number, timestamp
3700 when block was written, etc. :option:`verify` can be combined with
3701 :option:`verify_pattern` option. The allowed values are:
3704 Use an md5 sum of the data area and store it in the header of
3708 Use an experimental crc64 sum of the data area and store it in the
3709 header of each block.
3712 Use a crc32c sum of the data area and store it in the header of
3713 each block. This will automatically use hardware acceleration
3714 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
3715 fall back to software crc32c if none is found. Generally the
3716 fastest checksum fio supports when hardware accelerated.
3722 Use a crc32 sum of the data area and store it in the header of each
3726 Use a crc16 sum of the data area and store it in the header of each
3730 Use a crc7 sum of the data area and store it in the header of each
3734 Use xxhash as the checksum function. Generally the fastest software
3735 checksum that fio supports.
3738 Use sha512 as the checksum function.
3741 Use sha256 as the checksum function.
3744 Use optimized sha1 as the checksum function.
3747 Use optimized sha3-224 as the checksum function.
3750 Use optimized sha3-256 as the checksum function.
3753 Use optimized sha3-384 as the checksum function.
3756 Use optimized sha3-512 as the checksum function.
3759 This option is deprecated, since now meta information is included in
3760 generic verification header and meta verification happens by
3761 default. For detailed information see the description of the
3762 :option:`verify` setting. This option is kept because of
3763 compatibility's sake with old configurations. Do not use it.
3766 Verify a strict pattern. Normally fio includes a header with some
3767 basic information and checksumming, but if this option is set, only
3768 the specific pattern set with :option:`verify_pattern` is verified.
3771 Only pretend to verify. Useful for testing internals with
3772 :option:`ioengine`\=null, not for much else.
3774 This option can be used for repeated burn-in tests of a system to make sure
3775 that the written data is also correctly read back. If the data direction
3776 given is a read or random read, fio will assume that it should verify a
3777 previously written file. If the data direction includes any form of write,
3778 the verify will be of the newly written data.
3780 To avoid false verification errors, do not use the norandommap option when
3781 verifying data with async I/O engines and I/O depths > 1. Or use the
3782 norandommap and the lfsr random generator together to avoid writing to the
3783 same offset with multiple outstanding I/Os.
3785 .. option:: verify_offset=int
3787 Swap the verification header with data somewhere else in the block before
3788 writing. It is swapped back before verifying.
3790 .. option:: verify_interval=int
3792 Write the verification header at a finer granularity than the
3793 :option:`blocksize`. It will be written for chunks the size of
3794 ``verify_interval``. :option:`blocksize` should divide this evenly.
3796 .. option:: verify_pattern=str
3798 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
3799 filling with totally random bytes, but sometimes it's interesting to fill
3800 with a known pattern for I/O verification purposes. Depending on the width
3801 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
3802 be either a decimal or a hex number). The ``verify_pattern`` if larger than
3803 a 32-bit quantity has to be a hex number that starts with either "0x" or
3804 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
3805 format, which means that for each block offset will be written and then
3806 verified back, e.g.::
3810 Or use combination of everything::
3812 verify_pattern=0xff%o"abcd"-12
3814 .. option:: verify_fatal=bool
3816 Normally fio will keep checking the entire contents before quitting on a
3817 block verification failure. If this option is set, fio will exit the job on
3818 the first observed failure. Default: false.
3820 .. option:: verify_dump=bool
3822 If set, dump the contents of both the original data block and the data block
3823 we read off disk to files. This allows later analysis to inspect just what
3824 kind of data corruption occurred. Off by default.
3826 .. option:: verify_async=int
3828 Fio will normally verify I/O inline from the submitting thread. This option
3829 takes an integer describing how many async offload threads to create for I/O
3830 verification instead, causing fio to offload the duty of verifying I/O
3831 contents to one or more separate threads. If using this offload option, even
3832 sync I/O engines can benefit from using an :option:`iodepth` setting higher
3833 than 1, as it allows them to have I/O in flight while verifies are running.
3834 Defaults to 0 async threads, i.e. verification is not asynchronous.
3836 .. option:: verify_async_cpus=str
3838 Tell fio to set the given CPU affinity on the async I/O verification
3839 threads. See :option:`cpus_allowed` for the format used.
3841 .. option:: verify_backlog=int
3843 Fio will normally verify the written contents of a job that utilizes verify
3844 once that job has completed. In other words, everything is written then
3845 everything is read back and verified. You may want to verify continually
3846 instead for a variety of reasons. Fio stores the meta data associated with
3847 an I/O block in memory, so for large verify workloads, quite a bit of memory
3848 would be used up holding this meta data. If this option is enabled, fio will
3849 write only N blocks before verifying these blocks.
3851 .. option:: verify_backlog_batch=int
3853 Control how many blocks fio will verify if :option:`verify_backlog` is
3854 set. If not set, will default to the value of :option:`verify_backlog`
3855 (meaning the entire queue is read back and verified). If
3856 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
3857 blocks will be verified, if ``verify_backlog_batch`` is larger than
3858 :option:`verify_backlog`, some blocks will be verified more than once.
3860 .. option:: verify_state_save=bool
3862 When a job exits during the write phase of a verify workload, save its
3863 current state. This allows fio to replay up until that point, if the verify
3864 state is loaded for the verify read phase. The format of the filename is,
3867 <type>-<jobname>-<jobindex>-verify.state.
3869 <type> is "local" for a local run, "sock" for a client/server socket
3870 connection, and "ip" (192.168.0.1, for instance) for a networked
3871 client/server connection. Defaults to true.
3873 .. option:: verify_state_load=bool
3875 If a verify termination trigger was used, fio stores the current write state
3876 of each thread. This can be used at verification time so that fio knows how
3877 far it should verify. Without this information, fio will run a full
3878 verification pass, according to the settings in the job file used. Default
3881 .. option:: experimental_verify=bool
3883 Enable experimental verification. Standard verify records I/O metadata
3884 for later use during the verification phase. Experimental verify
3885 instead resets the file after the write phase and then replays I/Os for
3886 the verification phase.
3888 .. option:: trim_percentage=int
3890 Number of verify blocks to discard/trim.
3892 .. option:: trim_verify_zero=bool
3894 Verify that trim/discarded blocks are returned as zeros.
3896 .. option:: trim_backlog=int
3898 Trim after this number of blocks are written.
3900 .. option:: trim_backlog_batch=int
3902 Trim this number of I/O blocks.
3907 .. option:: steadystate=str:float, ss=str:float
3909 Define the criterion and limit for assessing steady state performance. The
3910 first parameter designates the criterion whereas the second parameter sets
3911 the threshold. When the criterion falls below the threshold for the
3912 specified duration, the job will stop. For example, `iops_slope:0.1%` will
3913 direct fio to terminate the job when the least squares regression slope
3914 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
3915 this will apply to all jobs in the group. Below is the list of available
3916 steady state assessment criteria. All assessments are carried out using only
3917 data from the rolling collection window. Threshold limits can be expressed
3918 as a fixed value or as a percentage of the mean in the collection window.
3920 When using this feature, most jobs should include the :option:`time_based`
3921 and :option:`runtime` options or the :option:`loops` option so that fio does not
3922 stop running after it has covered the full size of the specified file(s) or device(s).
3925 Collect IOPS data. Stop the job if all individual IOPS measurements
3926 are within the specified limit of the mean IOPS (e.g., ``iops:2``
3927 means that all individual IOPS values must be within 2 of the mean,
3928 whereas ``iops:0.2%`` means that all individual IOPS values must be
3929 within 0.2% of the mean IOPS to terminate the job).
3932 Collect IOPS data and calculate the least squares regression
3933 slope. Stop the job if the slope falls below the specified limit.
3936 Collect bandwidth data. Stop the job if all individual bandwidth
3937 measurements are within the specified limit of the mean bandwidth.
3940 Collect bandwidth data and calculate the least squares regression
3941 slope. Stop the job if the slope falls below the specified limit.
3943 .. option:: steadystate_duration=time, ss_dur=time
3945 A rolling window of this duration will be used to judge whether steady
3946 state has been reached. Data will be collected every
3947 :option:`ss_interval`. The default is 0 which disables steady state
3948 detection. When the unit is omitted, the value is interpreted in
3951 .. option:: steadystate_ramp_time=time, ss_ramp=time
3953 Allow the job to run for the specified duration before beginning data
3954 collection for checking the steady state job termination criterion. The
3955 default is 0. When the unit is omitted, the value is interpreted in seconds.
3957 .. option:: steadystate_check_interval=time, ss_interval=time
3959 The values during the rolling window will be collected with a period of
3960 this value. If :option:`ss_interval` is 30s and :option:`ss_dur` is
3961 300s, 10 measurements will be taken. Default is 1s but that might not
3962 converge, especially for slower devices, so set this accordingly. When
3963 the unit is omitted, the value is interpreted in seconds.
3966 Measurements and reporting
3967 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3969 .. option:: per_job_logs=bool
3971 If set to true, fio generates bw/clat/iops logs with per job unique
3972 filenames. If set to false, jobs with identical names will share a log
3973 filename. Note that when this option is set to false log files will be
3974 opened in append mode and if log files already exist the previous
3975 contents will not be overwritten. Default: true.
3977 .. option:: group_reporting
3979 It may sometimes be interesting to display statistics for groups of jobs as
3980 a whole instead of for each individual job. This is especially true if
3981 :option:`numjobs` is used; looking at individual thread/process output
3982 quickly becomes unwieldy. To see the final report per-group instead of
3983 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
3984 same reporting group, unless if separated by a :option:`stonewall`, or by
3985 using :option:`new_group`.
3987 NOTE: When :option: `group_reporting` is used along with `json` output,
3988 there are certain per-job properties which can be different between jobs
3989 but do not have a natural group-level equivalent. Examples include
3990 `kb_base`, `unit_base`, `sig_figs`, `thread_number`, `pid`, and
3991 `job_start`. For these properties, the values for the first job are
3992 recorded for the group.
3994 .. option:: new_group
3996 Start a new reporting group. See: :option:`group_reporting`. If not given,
3997 all jobs in a file will be part of the same reporting group, unless
3998 separated by a :option:`stonewall`.
4000 .. option:: stats=bool
4002 By default, fio collects and shows final output results for all jobs
4003 that run. If this option is set to 0, then fio will ignore it in
4004 the final stat output.
4006 .. option:: write_bw_log=str
4008 If given, write a bandwidth log for this job. Can be used to store data of
4009 the bandwidth of the jobs in their lifetime.
4011 If no str argument is given, the default filename of
4012 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
4013 will still append the type of log. So if one specifies::
4017 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
4018 of the job (`1..N`, where `N` is the number of jobs). If
4019 :option:`per_job_logs` is false, then the filename will not include the
4022 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
4023 text files into nice graphs. See `Log File Formats`_ for how data is
4024 structured within the file.
4026 .. option:: write_lat_log=str
4028 Same as :option:`write_bw_log`, except this option creates I/O
4029 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
4030 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
4031 latency files instead. See :option:`write_bw_log` for details about
4032 the filename format and `Log File Formats`_ for how data is structured
4035 .. option:: write_hist_log=str
4037 Same as :option:`write_bw_log` but writes an I/O completion latency
4038 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
4039 file will be empty unless :option:`log_hist_msec` has also been set.
4040 See :option:`write_bw_log` for details about the filename format and
4041 `Log File Formats`_ for how data is structured within the file.
4043 .. option:: write_iops_log=str
4045 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
4046 :file:`name_iops.x.log`) instead. Because fio defaults to individual
4047 I/O logging, the value entry in the IOPS log will be 1 unless windowed
4048 logging (see :option:`log_avg_msec`) has been enabled. See
4049 :option:`write_bw_log` for details about the filename format and `Log
4050 File Formats`_ for how data is structured within the file.
4052 .. option:: log_entries=int
4054 By default, fio will log an entry in the iops, latency, or bw log for
4055 every I/O that completes. The initial number of I/O log entries is 1024.
4056 When the log entries are all used, new log entries are dynamically
4057 allocated. This dynamic log entry allocation may negatively impact
4058 time-related statistics such as I/O tail latencies (e.g. 99.9th percentile
4059 completion latency). This option allows specifying a larger initial
4060 number of log entries to avoid run-time allocations of new log entries,
4061 resulting in more precise time-related I/O statistics.
4062 Also see :option:`log_avg_msec`. Defaults to 1024.
4064 .. option:: log_avg_msec=int
4066 By default, fio will log an entry in the iops, latency, or bw log for every
4067 I/O that completes. When writing to the disk log, that can quickly grow to a
4068 very large size. Setting this option makes fio average the each log entry
4069 over the specified period of time, reducing the resolution of the log. See
4070 :option:`log_max_value` as well. Defaults to 0, logging all entries.
4071 Also see `Log File Formats`_.
4073 .. option:: log_hist_msec=int
4075 Same as :option:`log_avg_msec`, but logs entries for completion latency
4076 histograms. Computing latency percentiles from averages of intervals using
4077 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
4078 histogram entries over the specified period of time, reducing log sizes for
4079 high IOPS devices while retaining percentile accuracy. See
4080 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
4081 Defaults to 0, meaning histogram logging is disabled.
4083 .. option:: log_hist_coarseness=int
4085 Integer ranging from 0 to 6, defining the coarseness of the resolution of
4086 the histogram logs enabled with :option:`log_hist_msec`. For each increment
4087 in coarseness, fio outputs half as many bins. Defaults to 0, for which
4088 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
4089 and `Log File Formats`_.
4091 .. option:: log_max_value=bool
4093 If :option:`log_avg_msec` is set, fio logs the average over that window. If
4094 you instead want to log the maximum value, set this option to 1. Defaults to
4095 0, meaning that averaged values are logged.
4097 .. option:: log_offset=bool
4099 If this is set, the iolog options will include the byte offset for the I/O
4100 entry as well as the other data values. Defaults to 0 meaning that
4101 offsets are not present in logs. Also see `Log File Formats`_.
4103 .. option:: log_compression=int
4105 If this is set, fio will compress the I/O logs as it goes, to keep the
4106 memory footprint lower. When a log reaches the specified size, that chunk is
4107 removed and compressed in the background. Given that I/O logs are fairly
4108 highly compressible, this yields a nice memory savings for longer runs. The
4109 downside is that the compression will consume some background CPU cycles, so
4110 it may impact the run. This, however, is also true if the logging ends up
4111 consuming most of the system memory. So pick your poison. The I/O logs are
4112 saved normally at the end of a run, by decompressing the chunks and storing
4113 them in the specified log file. This feature depends on the availability of
4116 .. option:: log_compression_cpus=str
4118 Define the set of CPUs that are allowed to handle online log compression for
4119 the I/O jobs. This can provide better isolation between performance
4120 sensitive jobs, and background compression work. See
4121 :option:`cpus_allowed` for the format used.
4123 .. option:: log_store_compressed=bool
4125 If set, fio will store the log files in a compressed format. They can be
4126 decompressed with fio, using the :option:`--inflate-log` command line
4127 parameter. The files will be stored with a :file:`.fz` suffix.
4129 .. option:: log_unix_epoch=bool
4131 Backwards compatible alias for log_alternate_epoch.
4133 .. option:: log_alternate_epoch=bool
4135 If set, fio will log timestamps based on the epoch used by the clock specified
4136 in the log_alternate_epoch_clock_id option, to the log files produced by
4137 enabling write_type_log for each log type, instead of the default zero-based
4140 .. option:: log_alternate_epoch_clock_id=int
4142 Specifies the clock_id to be used by clock_gettime to obtain the alternate
4143 epoch if log_alternate_epoch is true. Otherwise has no effect. Default
4144 value is 0, or CLOCK_REALTIME.
4146 .. option:: block_error_percentiles=bool
4148 If set, record errors in trim block-sized units from writes and trims and
4149 output a histogram of how many trims it took to get to errors, and what kind
4150 of error was encountered.
4152 .. option:: bwavgtime=int
4154 Average the calculated bandwidth over the given time. Value is specified in
4155 milliseconds. If the job also does bandwidth logging through
4156 :option:`write_bw_log`, then the minimum of this option and
4157 :option:`log_avg_msec` will be used. Default: 500ms.
4159 .. option:: iopsavgtime=int
4161 Average the calculated IOPS over the given time. Value is specified in
4162 milliseconds. If the job also does IOPS logging through
4163 :option:`write_iops_log`, then the minimum of this option and
4164 :option:`log_avg_msec` will be used. Default: 500ms.
4166 .. option:: disk_util=bool
4168 Generate disk utilization statistics, if the platform supports it.
4171 .. option:: disable_lat=bool
4173 Disable measurements of total latency numbers. Useful only for cutting back
4174 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
4175 performance at really high IOPS rates. Note that to really get rid of a
4176 large amount of these calls, this option must be used with
4177 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
4179 .. option:: disable_clat=bool
4181 Disable measurements of completion latency numbers. See
4182 :option:`disable_lat`.
4184 .. option:: disable_slat=bool
4186 Disable measurements of submission latency numbers. See
4187 :option:`disable_lat`.
4189 .. option:: disable_bw_measurement=bool, disable_bw=bool
4191 Disable measurements of throughput/bandwidth numbers. See
4192 :option:`disable_lat`.
4194 .. option:: slat_percentiles=bool
4196 Report submission latency percentiles. Submission latency is not recorded
4197 for synchronous ioengines.
4199 .. option:: clat_percentiles=bool
4201 Report completion latency percentiles.
4203 .. option:: lat_percentiles=bool
4205 Report total latency percentiles. Total latency is the sum of submission
4206 latency and completion latency.
4208 .. option:: percentile_list=float_list
4210 Overwrite the default list of percentiles for latencies and the block error
4211 histogram. Each number is a floating point number in the range (0,100], and
4212 the maximum length of the list is 20. Use ``:`` to separate the numbers. For
4213 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
4214 latency durations below which 99.5% and 99.9% of the observed latencies fell,
4217 .. option:: significant_figures=int
4219 If using :option:`--output-format` of `normal`, set the significant
4220 figures to this value. Higher values will yield more precise IOPS and
4221 throughput units, while lower values will round. Requires a minimum
4222 value of 1 and a maximum value of 10. Defaults to 4.
4228 .. option:: exitall_on_error
4230 When one job finishes in error, terminate the rest. The default is to wait
4231 for each job to finish.
4233 .. option:: continue_on_error=str
4235 Normally fio will exit the job on the first observed failure. If this option
4236 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
4237 EILSEQ) until the runtime is exceeded or the I/O size specified is
4238 completed. If this option is used, there are two more stats that are
4239 appended, the total error count and the first error. The error field given
4240 in the stats is the first error that was hit during the run.
4242 Note: a write error from the device may go unnoticed by fio when using
4243 buffered IO, as the write() (or similar) system call merely dirties the
4244 kernel pages, unless :option:`sync` or :option:`direct` is used. Device IO
4245 errors occur when the dirty data is actually written out to disk. If fully
4246 sync writes aren't desirable, :option:`fsync` or :option:`fdatasync` can be
4247 used as well. This is specific to writes, as reads are always synchronous.
4249 The allowed values are:
4252 Exit on any I/O or verify errors.
4255 Continue on read errors, exit on all others.
4258 Continue on write errors, exit on all others.
4261 Continue on any I/O error, exit on all others.
4264 Continue on verify errors, exit on all others.
4267 Continue on all errors.
4270 Backward-compatible alias for 'none'.
4273 Backward-compatible alias for 'all'.
4275 .. option:: ignore_error=str
4277 Sometimes you want to ignore some errors during test in that case you can
4278 specify error list for each error type, instead of only being able to
4279 ignore the default 'non-fatal error' using :option:`continue_on_error`.
4280 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
4281 given error type is separated with ':'. Error may be symbol ('ENOSPC',
4282 'ENOMEM') or integer. Example::
4284 ignore_error=EAGAIN,ENOSPC:122
4286 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
4287 WRITE. This option works by overriding :option:`continue_on_error` with
4288 the list of errors for each error type if any.
4290 .. option:: error_dump=bool
4292 If set dump every error even if it is non fatal, true by default. If
4293 disabled only fatal error will be dumped.
4295 Running predefined workloads
4296 ----------------------------
4298 Fio includes predefined profiles that mimic the I/O workloads generated by
4301 .. option:: profile=str
4303 The predefined workload to run. Current profiles are:
4306 Threaded I/O bench (tiotest/tiobench) like workload.
4309 Aerospike Certification Tool (ACT) like workload.
4311 To view a profile's additional options use :option:`--cmdhelp` after specifying
4312 the profile. For example::
4314 $ fio --profile=act --cmdhelp
4319 .. option:: device-names=str
4324 .. option:: load=int
4327 ACT load multiplier. Default: 1.
4329 .. option:: test-duration=time
4332 How long the entire test takes to run. When the unit is omitted, the value
4333 is given in seconds. Default: 24h.
4335 .. option:: threads-per-queue=int
4338 Number of read I/O threads per device. Default: 8.
4340 .. option:: read-req-num-512-blocks=int
4343 Number of 512B blocks to read at the time. Default: 3.
4345 .. option:: large-block-op-kbytes=int
4348 Size of large block ops in KiB (writes). Default: 131072.
4353 Set to run ACT prep phase.
4355 Tiobench profile options
4356 ~~~~~~~~~~~~~~~~~~~~~~~~
4358 .. option:: size=str
4363 .. option:: block=int
4366 Block size in bytes. Default: 4096.
4368 .. option:: numruns=int
4378 .. option:: threads=int
4383 Interpreting the output
4384 -----------------------
4387 Example output was based on the following:
4388 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
4389 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
4390 --runtime=2m --rw=rw
4392 Fio spits out a lot of output. While running, fio will display the status of the
4393 jobs created. An example of that would be::
4395 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]
4397 The characters inside the first set of square brackets denote the current status of
4398 each thread. The first character is the first job defined in the job file, and so
4399 forth. The possible values (in typical life cycle order) are:
4401 +------+-----+-----------------------------------------------------------+
4403 +======+=====+===========================================================+
4404 | P | | Thread setup, but not started. |
4405 +------+-----+-----------------------------------------------------------+
4406 | C | | Thread created. |
4407 +------+-----+-----------------------------------------------------------+
4408 | I | | Thread initialized, waiting or generating necessary data. |
4409 +------+-----+-----------------------------------------------------------+
4410 | | p | Thread running pre-reading file(s). |
4411 +------+-----+-----------------------------------------------------------+
4412 | | / | Thread is in ramp period. |
4413 +------+-----+-----------------------------------------------------------+
4414 | | R | Running, doing sequential reads. |
4415 +------+-----+-----------------------------------------------------------+
4416 | | r | Running, doing random reads. |
4417 +------+-----+-----------------------------------------------------------+
4418 | | W | Running, doing sequential writes. |
4419 +------+-----+-----------------------------------------------------------+
4420 | | w | Running, doing random writes. |
4421 +------+-----+-----------------------------------------------------------+
4422 | | M | Running, doing mixed sequential reads/writes. |
4423 +------+-----+-----------------------------------------------------------+
4424 | | m | Running, doing mixed random reads/writes. |
4425 +------+-----+-----------------------------------------------------------+
4426 | | D | Running, doing sequential trims. |
4427 +------+-----+-----------------------------------------------------------+
4428 | | d | Running, doing random trims. |
4429 +------+-----+-----------------------------------------------------------+
4430 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
4431 +------+-----+-----------------------------------------------------------+
4432 | | V | Running, doing verification of written data. |
4433 +------+-----+-----------------------------------------------------------+
4434 | f | | Thread finishing. |
4435 +------+-----+-----------------------------------------------------------+
4436 | E | | Thread exited, not reaped by main thread yet. |
4437 +------+-----+-----------------------------------------------------------+
4438 | _ | | Thread reaped. |
4439 +------+-----+-----------------------------------------------------------+
4440 | X | | Thread reaped, exited with an error. |
4441 +------+-----+-----------------------------------------------------------+
4442 | K | | Thread reaped, exited due to signal. |
4443 +------+-----+-----------------------------------------------------------+
4446 Example output was based on the following:
4447 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
4448 --time_based --rate=2512k --bs=256K --numjobs=10 \
4449 --name=readers --rw=read --name=writers --rw=write
4451 Fio will condense the thread string as not to take up more space on the command
4452 line than needed. For instance, if you have 10 readers and 10 writers running,
4453 the output would look like this::
4455 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]
4457 Note that the status string is displayed in order, so it's possible to tell which of
4458 the jobs are currently doing what. In the example above this means that jobs 1--10
4459 are readers and 11--20 are writers.
4461 The other values are fairly self explanatory -- number of threads currently
4462 running and doing I/O, the number of currently open files (f=), the estimated
4463 completion percentage, the rate of I/O since last check (read speed listed first,
4464 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
4465 and time to completion for the current running group. It's impossible to estimate
4466 runtime of the following groups (if any).
4469 Example output was based on the following:
4470 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
4471 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
4472 --bs=7K --name=Client1 --rw=write
4474 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
4475 each thread, group of threads, and disks in that order. For each overall thread (or
4476 group) the output looks like::
4478 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
4479 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
4480 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
4481 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
4482 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
4483 clat percentiles (usec):
4484 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
4485 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
4486 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
4487 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
4489 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
4490 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
4491 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
4492 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
4493 lat (msec) : 100=0.65%
4494 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
4495 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
4496 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4497 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4498 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
4499 latency : target=0, window=0, percentile=100.00%, depth=8
4501 The job name (or first job's name when using :option:`group_reporting`) is printed,
4502 along with the group id, count of jobs being aggregated, last error id seen (which
4503 is 0 when there are no errors), pid/tid of that thread and the time the job/group
4504 completed. Below are the I/O statistics for each data direction performed (showing
4505 writes in the example above). In the order listed, they denote:
4508 The string before the colon shows the I/O direction the statistics
4509 are for. **IOPS** is the average I/Os performed per second. **BW**
4510 is the average bandwidth rate shown as: value in power of 2 format
4511 (value in power of 10 format). The last two values show: (**total
4512 I/O performed** in power of 2 format / **runtime** of that thread).
4515 Submission latency (**min** being the minimum, **max** being the
4516 maximum, **avg** being the average, **stdev** being the standard
4517 deviation). This is the time from when fio initialized the I/O
4518 to submission. For synchronous ioengines this includes the time
4519 up until just before the ioengine's queue function is called.
4520 For asynchronous ioengines this includes the time up through the
4521 completion of the ioengine's queue function (and commit function
4522 if it is defined). For sync I/O this row is not displayed as the
4523 slat is negligible. This value can be in nanoseconds,
4524 microseconds or milliseconds --- fio will choose the most
4525 appropriate base and print that (in the example above
4526 nanoseconds was the best scale). Note: in :option:`--minimal`
4527 mode latencies are always expressed in microseconds.
4530 Completion latency. Same names as slat, this denotes the time from
4531 submission to completion of the I/O pieces. For sync I/O, this
4532 represents the time from when the I/O was submitted to the
4533 operating system to when it was completed. For asynchronous
4534 ioengines this is the time from when the ioengine's queue (and
4535 commit if available) functions were completed to when the I/O's
4536 completion was reaped by fio.
4539 Total latency. Same names as slat and clat, this denotes the time from
4540 when fio created the I/O unit to completion of the I/O operation.
4541 It is the sum of submission and completion latency.
4544 Bandwidth statistics based on measurements from discrete
4545 intervals. Fio continuously monitors bytes transferred and I/O
4546 operations completed. By default fio calculates bandwidth in
4547 each half-second interval (see :option:`bwavgtime`) and reports
4548 descriptive statistics for the measurements here. Same names as
4549 the xlat stats, but also includes the number of samples taken
4550 (**samples**) and an approximate percentage of total aggregate
4551 bandwidth this thread received in its group (**per**). This
4552 last value is only really useful if the threads in this group
4553 are on the same disk, since they are then competing for disk
4557 IOPS statistics based on measurements from discrete intervals.
4558 For details see the description for bw above. See
4559 :option:`iopsavgtime` to control the duration of the intervals.
4560 Same values reported here as for bw except for percentage.
4562 **lat (nsec/usec/msec)**
4563 The distribution of I/O completion latencies. This is the time from when
4564 I/O leaves fio and when it gets completed. Unlike the separate
4565 read/write/trim sections above, the data here and in the remaining
4566 sections apply to all I/Os for the reporting group. 250=0.04% means that
4567 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
4568 of the I/Os required 250 to 499us for completion.
4571 CPU usage. User and system time, along with the number of context
4572 switches this thread went through, usage of system and user time, and
4573 finally the number of major and minor page faults. The CPU utilization
4574 numbers are averages for the jobs in that reporting group, while the
4575 context and fault counters are summed.
4578 The distribution of I/O depths over the job lifetime. The numbers are
4579 divided into powers of 2 and each entry covers depths from that value
4580 up to those that are lower than the next entry -- e.g., 16= covers
4581 depths from 16 to 31. Note that the range covered by a depth
4582 distribution entry can be different to the range covered by the
4583 equivalent submit/complete distribution entry.
4586 How many pieces of I/O were submitting in a single submit call. Each
4587 entry denotes that amount and below, until the previous entry -- e.g.,
4588 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
4589 call. Note that the range covered by a submit distribution entry can
4590 be different to the range covered by the equivalent depth distribution
4594 Like the above submit number, but for completions instead.
4597 The number of read/write/trim requests issued, and how many of them were
4601 These values are for :option:`latency_target` and related options. When
4602 these options are engaged, this section describes the I/O depth required
4603 to meet the specified latency target.
4606 Example output was based on the following:
4607 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
4608 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
4609 --rate=11M --name=write --rw=write --bs=2k --rate=700k
4611 After each client has been listed, the group statistics are printed. They
4612 will look like this::
4614 Run status group 0 (all jobs):
4615 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
4616 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
4618 For each data direction it prints:
4621 Aggregate bandwidth of threads in this group followed by the
4622 minimum and maximum bandwidth of all the threads in this group.
4623 Values outside of brackets are power-of-2 format and those
4624 within are the equivalent value in a power-of-10 format.
4626 Aggregate I/O performed of all threads in this group. The
4627 format is the same as bw.
4629 The smallest and longest runtimes of the threads in this group.
4631 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
4633 Disk stats (read/write):
4634 sda: ios=16398/16511, sectors=32321/65472, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
4636 Each value is printed for both reads and writes, with reads first. The
4640 Number of I/Os performed by all groups.
4642 Amount of data transferred in units of 512 bytes for all groups.
4644 Number of merges performed by the I/O scheduler.
4646 Number of ticks we kept the disk busy.
4648 Total time spent in the disk queue.
4650 The disk utilization. A value of 100% means we kept the disk
4651 busy constantly, 50% would be a disk idling half of the time.
4653 It is also possible to get fio to dump the current output while it is running,
4654 without terminating the job. To do that, send fio the **USR1** signal. You can
4655 also get regularly timed dumps by using the :option:`--status-interval`
4656 parameter, or by creating a file in :file:`/tmp` named
4657 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
4658 current output status.
4664 For scripted usage where you typically want to generate tables or graphs of the
4665 results, fio can output the results in a semicolon separated format. The format
4666 is one long line of values, such as::
4668 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%
4669 A description of this job goes here.
4671 The job description (if provided) follows on a second line for terse v2.
4672 It appears on the same line for other terse versions.
4674 To enable terse output, use the :option:`--minimal` or
4675 :option:`--output-format`\=terse command line options. The
4676 first value is the version of the terse output format. If the output has to be
4677 changed for some reason, this number will be incremented by 1 to signify that
4680 Split up, the format is as follows (comments in brackets denote when a
4681 field was introduced or whether it's specific to some terse version):
4685 terse version, fio version [v3], jobname, groupid, error
4689 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4690 Submission latency: min, max, mean, stdev (usec)
4691 Completion latency: min, max, mean, stdev (usec)
4692 Completion latency percentiles: 20 fields (see below)
4693 Total latency: min, max, mean, stdev (usec)
4694 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4695 IOPS [v5]: min, max, mean, stdev, number of samples
4701 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4702 Submission latency: min, max, mean, stdev (usec)
4703 Completion latency: min, max, mean, stdev (usec)
4704 Completion latency percentiles: 20 fields (see below)
4705 Total latency: min, max, mean, stdev (usec)
4706 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4707 IOPS [v5]: min, max, mean, stdev, number of samples
4709 TRIM status [all but version 3]:
4711 Fields are similar to READ/WRITE status.
4715 user, system, context switches, major faults, minor faults
4719 <=1, 2, 4, 8, 16, 32, >=64
4721 I/O latencies microseconds::
4723 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
4725 I/O latencies milliseconds::
4727 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
4729 Disk utilization [v3]::
4731 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
4732 time spent in queue, disk utilization percentage
4734 Additional Info (dependent on continue_on_error, default off)::
4736 total # errors, first error code
4738 Additional Info (dependent on description being set)::
4742 Completion latency percentiles can be a grouping of up to 20 sets, so for the
4743 terse output fio writes all of them. Each field will look like this::
4747 which is the Xth percentile, and the `usec` latency associated with it.
4749 For `Disk utilization`, all disks used by fio are shown. So for each disk there
4750 will be a disk utilization section.
4752 Below is a single line containing short names for each of the fields in the
4753 minimal output v3, separated by semicolons::
4755 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth_kb;read_iops;read_runtime_ms;read_slat_min_us;read_slat_max_us;read_slat_mean_us;read_slat_dev_us;read_clat_min_us;read_clat_max_us;read_clat_mean_us;read_clat_dev_us;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_us;read_lat_max_us;read_lat_mean_us;read_lat_dev_us;read_bw_min_kb;read_bw_max_kb;read_bw_agg_pct;read_bw_mean_kb;read_bw_dev_kb;write_kb;write_bandwidth_kb;write_iops;write_runtime_ms;write_slat_min_us;write_slat_max_us;write_slat_mean_us;write_slat_dev_us;write_clat_min_us;write_clat_max_us;write_clat_mean_us;write_clat_dev_us;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_us;write_lat_max_us;write_lat_mean_us;write_lat_dev_us;write_bw_min_kb;write_bw_max_kb;write_bw_agg_pct;write_bw_mean_kb;write_bw_dev_kb;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
4757 In client/server mode terse output differs from what appears when jobs are run
4758 locally. Disk utilization data is omitted from the standard terse output and
4759 for v3 and later appears on its own separate line at the end of each terse
4766 The `json` output format is intended to be both human readable and convenient
4767 for automated parsing. For the most part its sections mirror those of the
4768 `normal` output. The `runtime` value is reported in msec and the `bw` value is
4769 reported in 1024 bytes per second units.
4775 The `json+` output format is identical to the `json` output format except that it
4776 adds a full dump of the completion latency bins. Each `bins` object contains a
4777 set of (key, value) pairs where keys are latency durations and values count how
4778 many I/Os had completion latencies of the corresponding duration. For example,
4781 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
4783 This data indicates that one I/O required 87,552ns to complete, two I/Os required
4784 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
4786 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
4787 json+ output and generates CSV-formatted latency data suitable for plotting.
4789 The latency durations actually represent the midpoints of latency intervals.
4790 For details refer to :file:`stat.h`.
4796 There are two trace file format that you can encounter. The older (v1) format is
4797 unsupported since version 1.20-rc3 (March 2008). It will still be described
4798 below in case that you get an old trace and want to understand it.
4800 In any case the trace is a simple text file with a single action per line.
4803 Trace file format v1
4804 ~~~~~~~~~~~~~~~~~~~~
4806 Each line represents a single I/O action in the following format::
4810 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
4812 This format is not supported in fio versions >= 1.20-rc3.
4815 Trace file format v2
4816 ~~~~~~~~~~~~~~~~~~~~
4818 The second version of the trace file format was added in fio version 1.17. It
4819 allows one to access more than one file per trace and has a bigger set of possible
4822 The first line of the trace file has to be::
4826 Following this can be lines in two different formats, which are described below.
4828 The file management format::
4832 The `filename` is given as an absolute path. The `action` can be one of these:
4835 Add the given `filename` to the trace.
4837 Open the file with the given `filename`. The `filename` has to have
4838 been added with the **add** action before.
4840 Close the file with the given `filename`. The file has to have been
4844 The file I/O action format::
4846 filename action offset length
4848 The `filename` is given as an absolute path, and has to have been added and
4849 opened before it can be used with this format. The `offset` and `length` are
4850 given in bytes. The `action` can be one of these:
4853 Wait for `offset` microseconds. Everything below 100 is discarded.
4854 The time is relative to the previous `wait` statement. Note that
4855 action `wait` is not allowed as of version 3, as the same behavior
4856 can be achieved using timestamps.
4858 Read `length` bytes beginning from `offset`.
4860 Write `length` bytes beginning from `offset`.
4862 :manpage:`fsync(2)` the file.
4864 :manpage:`fdatasync(2)` the file.
4866 Trim the given file from the given `offset` for `length` bytes.
4869 Trace file format v3
4870 ~~~~~~~~~~~~~~~~~~~~
4872 The third version of the trace file format was added in fio version 3.31. It
4873 forces each action to have a timestamp associated with it.
4875 The first line of the trace file has to be::
4879 Following this can be lines in two different formats, which are described below.
4881 The file management format::
4883 timestamp filename action
4885 The file I/O action format::
4887 timestamp filename action offset length
4889 The `timestamp` is relative to the beginning of the run (ie starts at 0). The
4890 `filename`, `action`, `offset` and `length` are identical to version 2, except
4891 that version 3 does not allow the `wait` action.
4894 I/O Replay - Merging Traces
4895 ---------------------------
4897 Colocation is a common practice used to get the most out of a machine.
4898 Knowing which workloads play nicely with each other and which ones don't is
4899 a much harder task. While fio can replay workloads concurrently via multiple
4900 jobs, it leaves some variability up to the scheduler making results harder to
4901 reproduce. Merging is a way to make the order of events consistent.
4903 Merging is integrated into I/O replay and done when a
4904 :option:`merge_blktrace_file` is specified. The list of files passed to
4905 :option:`read_iolog` go through the merge process and output a single file
4906 stored to the specified file. The output file is passed on as if it were the
4907 only file passed to :option:`read_iolog`. An example would look like::
4909 $ fio --read_iolog="<file1>:<file2>" --merge_blktrace_file="<output_file>"
4911 Creating only the merged file can be done by passing the command line argument
4912 :option:`--merge-blktrace-only`.
4914 Scaling traces can be done to see the relative impact of any particular trace
4915 being slowed down or sped up. :option:`merge_blktrace_scalars` takes in a colon
4916 separated list of percentage scalars. It is index paired with the files passed
4917 to :option:`read_iolog`.
4919 With scaling, it may be desirable to match the running time of all traces.
4920 This can be done with :option:`merge_blktrace_iters`. It is index paired with
4921 :option:`read_iolog` just like :option:`merge_blktrace_scalars`.
4923 In an example, given two traces, A and B, each 60s long. If we want to see
4924 the impact of trace A issuing IOs twice as fast and repeat trace A over the
4925 runtime of trace B, the following can be done::
4927 $ fio --read_iolog="<trace_a>:"<trace_b>" --merge_blktrace_file"<output_file>" --merge_blktrace_scalars="50:100" --merge_blktrace_iters="2:1"
4929 This runs trace A at 2x the speed twice for approximately the same runtime as
4930 a single run of trace B.
4933 CPU idleness profiling
4934 ----------------------
4936 In some cases, we want to understand CPU overhead in a test. For example, we
4937 test patches for the specific goodness of whether they reduce CPU usage.
4938 Fio implements a balloon approach to create a thread per CPU that runs at idle
4939 priority, meaning that it only runs when nobody else needs the cpu.
4940 By measuring the amount of work completed by the thread, idleness of each CPU
4941 can be derived accordingly.
4943 An unit work is defined as touching a full page of unsigned characters. Mean and
4944 standard deviation of time to complete an unit work is reported in "unit work"
4945 section. Options can be chosen to report detailed percpu idleness or overall
4946 system idleness by aggregating percpu stats.
4949 Verification and triggers
4950 -------------------------
4952 Fio is usually run in one of two ways, when data verification is done. The first
4953 is a normal write job of some sort with verify enabled. When the write phase has
4954 completed, fio switches to reads and verifies everything it wrote. The second
4955 model is running just the write phase, and then later on running the same job
4956 (but with reads instead of writes) to repeat the same I/O patterns and verify
4957 the contents. Both of these methods depend on the write phase being completed,
4958 as fio otherwise has no idea how much data was written.
4960 With verification triggers, fio supports dumping the current write state to
4961 local files. Then a subsequent read verify workload can load this state and know
4962 exactly where to stop. This is useful for testing cases where power is cut to a
4963 server in a managed fashion, for instance.
4965 A verification trigger consists of two things:
4967 1) Storing the write state of each job.
4968 2) Executing a trigger command.
4970 The write state is relatively small, on the order of hundreds of bytes to single
4971 kilobytes. It contains information on the number of completions done, the last X
4974 A trigger is invoked either through creation ('touch') of a specified file in
4975 the system, or through a timeout setting. If fio is run with
4976 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
4977 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
4978 will fire off the trigger (thus saving state, and executing the trigger
4981 For client/server runs, there's both a local and remote trigger. If fio is
4982 running as a server backend, it will send the job states back to the client for
4983 safe storage, then execute the remote trigger, if specified. If a local trigger
4984 is specified, the server will still send back the write state, but the client
4985 will then execute the trigger.
4987 Verification trigger example
4988 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4990 Let's say we want to run a powercut test on the remote Linux machine 'server'.
4991 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
4992 some point during the run, and we'll run this test from the safety or our local
4993 machine, 'localbox'. On the server, we'll start the fio backend normally::
4995 server# fio --server
4997 and on the client, we'll fire off the workload::
4999 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
5001 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
5003 echo b > /proc/sysrq-trigger
5005 on the server once it has received the trigger and sent us the write state. This
5006 will work, but it's not **really** cutting power to the server, it's merely
5007 abruptly rebooting it. If we have a remote way of cutting power to the server
5008 through IPMI or similar, we could do that through a local trigger command
5009 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
5010 ipmi-reboot. On localbox, we could then have run fio with a local trigger
5013 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
5015 For this case, fio would wait for the server to send us the write state, then
5016 execute ``ipmi-reboot server`` when that happened.
5018 Loading verify state
5019 ~~~~~~~~~~~~~~~~~~~~
5021 To load stored write state, a read verification job file must contain the
5022 :option:`verify_state_load` option. If that is set, fio will load the previously
5023 stored state. For a local fio run this is done by loading the files directly,
5024 and on a client/server run, the server backend will ask the client to send the
5025 files over and load them from there.
5031 Fio supports a variety of log file formats, for logging latencies, bandwidth,
5032 and IOPS. The logs share a common format, which looks like this:
5034 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
5035 *offset* (`bytes`), *command priority*
5037 *Time* for the log entry is always in milliseconds. The *value* logged depends
5038 on the type of log, it will be one of the following:
5041 Value is latency in nsecs
5047 *Data direction* is one of the following:
5056 The entry's *block size* is always in bytes. The *offset* is the position in bytes
5057 from the start of the file for that particular I/O. The logging of the offset can be
5058 toggled with :option:`log_offset`.
5060 *Command priority* is 0 for normal priority and 1 for high priority. This is controlled
5061 by the ioengine specific :option:`cmdprio_percentage`.
5063 Fio defaults to logging every individual I/O but when windowed logging is set
5064 through :option:`log_avg_msec`, either the average (by default) or the maximum
5065 (:option:`log_max_value` is set) *value* seen over the specified period of time
5066 is recorded. Each *data direction* seen within the window period will aggregate
5067 its values in a separate row. Further, when using windowed logging the *block
5068 size* and *offset* entries will always contain 0.
5074 Normally fio is invoked as a stand-alone application on the machine where the
5075 I/O workload should be generated. However, the backend and frontend of fio can
5076 be run separately i.e., the fio server can generate an I/O workload on the "Device
5077 Under Test" while being controlled by a client on another machine.
5079 Start the server on the machine which has access to the storage DUT::
5083 where `args` defines what fio listens to. The arguments are of the form
5084 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
5085 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
5086 *hostname* is either a hostname or IP address, and *port* is the port to listen
5087 to (only valid for TCP/IP, not a local socket). Some examples:
5091 Start a fio server, listening on all interfaces on the default port (8765).
5093 2) ``fio --server=ip:hostname,4444``
5095 Start a fio server, listening on IP belonging to hostname and on port 4444.
5097 3) ``fio --server=ip6:::1,4444``
5099 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
5101 4) ``fio --server=,4444``
5103 Start a fio server, listening on all interfaces on port 4444.
5105 5) ``fio --server=1.2.3.4``
5107 Start a fio server, listening on IP 1.2.3.4 on the default port.
5109 6) ``fio --server=sock:/tmp/fio.sock``
5111 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
5113 Once a server is running, a "client" can connect to the fio server with::
5115 fio <local-args> --client=<server> <remote-args> <job file(s)>
5117 where `local-args` are arguments for the client where it is running, `server`
5118 is the connect string, and `remote-args` and `job file(s)` are sent to the
5119 server. The `server` string follows the same format as it does on the server
5120 side, to allow IP/hostname/socket and port strings.
5122 Note that all job options must be defined in job files when running fio as a
5123 client. Any job options specified in `remote-args` will be ignored.
5125 Fio can connect to multiple servers this way::
5127 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
5129 If the job file is located on the fio server, then you can tell the server to
5130 load a local file as well. This is done by using :option:`--remote-config` ::
5132 fio --client=server --remote-config /path/to/file.fio
5134 Then fio will open this local (to the server) job file instead of being passed
5135 one from the client.
5137 If you have many servers (example: 100 VMs/containers), you can input a pathname
5138 of a file containing host IPs/names as the parameter value for the
5139 :option:`--client` option. For example, here is an example :file:`host.list`
5140 file containing 2 hostnames::
5142 host1.your.dns.domain
5143 host2.your.dns.domain
5145 The fio command would then be::
5147 fio --client=host.list <job file(s)>
5149 In this mode, you cannot input server-specific parameters or job files -- all
5150 servers receive the same job file.
5152 In order to let ``fio --client`` runs use a shared filesystem from multiple
5153 hosts, ``fio --client`` now prepends the IP address of the server to the
5154 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
5155 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
5156 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
5157 192.168.10.121, then fio will create two files::
5159 /mnt/nfs/fio/192.168.10.120.fileio.tmp
5160 /mnt/nfs/fio/192.168.10.121.fileio.tmp
5162 Terse output in client/server mode will differ slightly from what is produced
5163 when fio is run in stand-alone mode. See the terse output section for details.