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
760 The clock_id passed to the call to `clock_gettime` used to record
761 job_start in the `json` output format. Default is 0, or CLOCK_REALTIME.
767 .. option:: directory=str
769 Prefix filenames with this directory. Used to place files in a different
770 location than :file:`./`. You can specify a number of directories by
771 separating the names with a ':' character. These directories will be
772 assigned equally distributed to job clones created by :option:`numjobs` as
773 long as they are using generated filenames. If specific `filename(s)` are
774 set fio will use the first listed directory, and thereby matching the
775 `filename` semantic (which generates a file for each clone if not
776 specified, but lets all clones use the same file if set).
778 See the :option:`filename` option for information on how to escape "``:``"
779 characters within the directory path itself.
781 Note: To control the directory fio will use for internal state files
782 use :option:`--aux-path`.
784 .. option:: filename=str
786 Fio normally makes up a `filename` based on the job name, thread number, and
787 file number (see :option:`filename_format`). If you want to share files
788 between threads in a job or several
789 jobs with fixed file paths, specify a `filename` for each of them to override
790 the default. If the ioengine is file based, you can specify a number of files
791 by separating the names with a ':' colon. So if you wanted a job to open
792 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
793 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
794 specified, :option:`nrfiles` is ignored. The size of regular files specified
795 by this option will be :option:`size` divided by number of files unless an
796 explicit size is specified by :option:`filesize`.
798 Each colon in the wanted path must be escaped with a ``\``
799 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
800 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
801 :file:`F:\\filename` then you would use ``filename=F\:\filename``.
803 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
804 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
805 Note: Windows and FreeBSD (refer to geom(4)) prevent write access to areas
806 of the disk containing in-use data (e.g. filesystems).
808 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
809 of the two depends on the read/write direction set.
811 .. option:: filename_format=str
813 If sharing multiple files between jobs, it is usually necessary to have fio
814 generate the exact names that you want. By default, fio will name a file
815 based on the default file format specification of
816 :file:`jobname.jobnumber.filenumber`. With this option, that can be
817 customized. Fio will recognize and replace the following keywords in this
821 The name of the worker thread or process.
823 IP of the fio process when using client/server mode.
825 The incremental number of the worker thread or process.
827 The incremental number of the file for that worker thread or
830 To have dependent jobs share a set of files, this option can be set to have
831 fio generate filenames that are shared between the two. For instance, if
832 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
833 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
834 will be used if no other format specifier is given.
836 If you specify a path then the directories will be created up to the
837 main directory for the file. So for example if you specify
838 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
839 created before the file setup part of the job. If you specify
840 :option:`directory` then the path will be relative that directory,
841 otherwise it is treated as the absolute path.
843 .. option:: unique_filename=bool
845 To avoid collisions between networked clients, fio defaults to prefixing any
846 generated filenames (with a directory specified) with the source of the
847 client connecting. To disable this behavior, set this option to 0.
849 .. option:: opendir=str
851 Recursively open any files below directory `str`. This accepts only a
852 single directory and unlike related options, colons appearing in the
853 path must not be escaped.
855 .. option:: lockfile=str
857 Fio defaults to not locking any files before it does I/O to them. If a file
858 or file descriptor is shared, fio can serialize I/O to that file to make the
859 end result consistent. This is usual for emulating real workloads that share
860 files. The lock modes are:
863 No locking. The default.
865 Only one thread or process may do I/O at a time, excluding all
868 Read-write locking on the file. Many readers may
869 access the file at the same time, but writes get exclusive access.
871 .. option:: nrfiles=int
873 Number of files to use for this job. Defaults to 1. The size of files
874 will be :option:`size` divided by this unless explicit size is specified by
875 :option:`filesize`. Files are created for each thread separately, and each
876 file will have a file number within its name by default, as explained in
877 :option:`filename` section.
880 .. option:: openfiles=int
882 Number of files to keep open at the same time. Defaults to the same as
883 :option:`nrfiles`, can be set smaller to limit the number simultaneous
886 .. option:: file_service_type=str
888 Defines how fio decides which file from a job to service next. The following
892 Choose a file at random.
895 Round robin over opened files. This is the default.
898 Finish one file before moving on to the next. Multiple files can
899 still be open depending on :option:`openfiles`.
902 Use a *Zipf* distribution to decide what file to access.
905 Use a *Pareto* distribution to decide what file to access.
908 Use a *Gaussian* (normal) distribution to decide what file to
914 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
915 tell fio how many I/Os to issue before switching to a new file. For example,
916 specifying ``file_service_type=random:8`` would cause fio to issue
917 8 I/Os before selecting a new file at random. For the non-uniform
918 distributions, a floating point postfix can be given to influence how the
919 distribution is skewed. See :option:`random_distribution` for a description
920 of how that would work.
922 .. option:: ioscheduler=str
924 Attempt to switch the device hosting the file to the specified I/O scheduler
927 .. option:: create_serialize=bool
929 If true, serialize the file creation for the jobs. This may be handy to
930 avoid interleaving of data files, which may greatly depend on the filesystem
931 used and even the number of processors in the system. Default: true.
933 .. option:: create_fsync=bool
935 :manpage:`fsync(2)` the data file after creation. This is the default.
937 .. option:: create_on_open=bool
939 If true, don't pre-create files but allow the job's open() to create a file
940 when it's time to do I/O. Default: false -- pre-create all necessary files
943 .. option:: create_only=bool
945 If true, fio will only run the setup phase of the job. If files need to be
946 laid out or updated on disk, only that will be done -- the actual job contents
947 are not executed. Default: false.
949 .. option:: allow_file_create=bool
951 If true, fio is permitted to create files as part of its workload. If this
952 option is false, then fio will error out if
953 the files it needs to use don't already exist. Default: true.
955 .. option:: allow_mounted_write=bool
957 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
958 to what appears to be a mounted device or partition. This should help catch
959 creating inadvertently destructive tests, not realizing that the test will
960 destroy data on the mounted file system. Note that some platforms don't allow
961 writing against a mounted device regardless of this option. Default: false.
963 .. option:: pre_read=bool
965 If this is given, files will be pre-read into memory before starting the
966 given I/O operation. This will also clear the :option:`invalidate` flag,
967 since it is pointless to pre-read and then drop the cache. This will only
968 work for I/O engines that are seek-able, since they allow you to read the
969 same data multiple times. Thus it will not work on non-seekable I/O engines
970 (e.g. network, splice). Default: false.
972 .. option:: unlink=bool
974 Unlink the job files when done. Not the default, as repeated runs of that
975 job would then waste time recreating the file set again and again. Default:
978 .. option:: unlink_each_loop=bool
980 Unlink job files after each iteration or loop. Default: false.
982 .. option:: zonemode=str
987 The :option:`zonerange`, :option:`zonesize`,
988 :option `zonecapacity` and option:`zoneskip`
989 parameters are ignored.
991 I/O happens in a single zone until
992 :option:`zonesize` bytes have been transferred.
993 After that number of bytes has been
994 transferred processing of the next zone
995 starts. :option `zonecapacity` is ignored.
997 Zoned block device mode. I/O happens
998 sequentially in each zone, even if random I/O
999 has been selected. Random I/O happens across
1000 all zones instead of being restricted to a
1001 single zone. The :option:`zoneskip` parameter
1002 is ignored. :option:`zonerange` and
1003 :option:`zonesize` must be identical.
1004 Trim is handled using a zone reset operation.
1005 Trim only considers non-empty sequential write
1006 required and sequential write preferred zones.
1008 .. option:: zonerange=int
1010 Size of a single zone. See also :option:`zonesize` and
1013 .. option:: zonesize=int
1015 For :option:`zonemode` =strided, this is the number of bytes to
1016 transfer before skipping :option:`zoneskip` bytes. If this parameter
1017 is smaller than :option:`zonerange` then only a fraction of each zone
1018 with :option:`zonerange` bytes will be accessed. If this parameter is
1019 larger than :option:`zonerange` then each zone will be accessed
1020 multiple times before skipping to the next zone.
1022 For :option:`zonemode` =zbd, this is the size of a single zone. The
1023 :option:`zonerange` parameter is ignored in this mode.
1026 .. option:: zonecapacity=int
1028 For :option:`zonemode` =zbd, this defines the capacity of a single zone,
1029 which is the accessible area starting from the zone start address.
1030 This parameter only applies when using :option:`zonemode` =zbd in
1031 combination with regular block devices. If not specified it defaults to
1032 the zone size. If the target device is a zoned block device, the zone
1033 capacity is obtained from the device information and this option is
1036 .. option:: zoneskip=int
1038 For :option:`zonemode` =strided, the number of bytes to skip after
1039 :option:`zonesize` bytes of data have been transferred. This parameter
1040 must be zero for :option:`zonemode` =zbd.
1042 .. option:: read_beyond_wp=bool
1044 This parameter applies to :option:`zonemode` =zbd only.
1046 Zoned block devices are block devices that consist of multiple zones.
1047 Each zone has a type, e.g. conventional or sequential. A conventional
1048 zone can be written at any offset that is a multiple of the block
1049 size. Sequential zones must be written sequentially. The position at
1050 which a write must occur is called the write pointer. A zoned block
1051 device can be either drive managed, host managed or host aware. For
1052 host managed devices the host must ensure that writes happen
1053 sequentially. Fio recognizes host managed devices and serializes
1054 writes to sequential zones for these devices.
1056 If a read occurs in a sequential zone beyond the write pointer then
1057 the zoned block device will complete the read without reading any data
1058 from the storage medium. Since such reads lead to unrealistically high
1059 bandwidth and IOPS numbers fio only reads beyond the write pointer if
1060 explicitly told to do so. Default: false.
1062 .. option:: max_open_zones=int
1064 When a zone of a zoned block device is partially written (i.e. not all
1065 sectors of the zone have been written), the zone is in one of three
1066 conditions: 'implicit open', 'explicit open' or 'closed'. Zoned block
1067 devices may have a limit called 'max_open_zones' (same name as the
1068 parameter) on the total number of zones that can simultaneously be in
1069 the 'implicit open' or 'explicit open' conditions. Zoned block devices
1070 may have another limit called 'max_active_zones', on the total number of
1071 zones that can simultaneously be in the three conditions. The
1072 :option:`max_open_zones` parameter limits the number of zones to which
1073 write commands are issued by all fio jobs, that is, limits the number of
1074 zones that will be in the conditions. When the device has the
1075 max_open_zones limit and does not have the max_active_zones limit, the
1076 :option:`max_open_zones` parameter limits the number of zones in the two
1077 open conditions up to the limit. In this case, fio includes zones in the
1078 two open conditions to the write target zones at fio start. When the
1079 device has both the max_open_zones and the max_active_zones limits, the
1080 :option:`max_open_zones` parameter limits the number of zones in the
1081 three conditions up to the limit. In this case, fio includes zones in
1082 the three conditions to the write target zones at fio start.
1084 This parameter is relevant only if the :option:`zonemode` =zbd is used.
1085 The default value is always equal to the max_open_zones limit of the
1086 target zoned block device and a value higher than this limit cannot be
1087 specified by users unless the option :option:`ignore_zone_limits` is
1088 specified. When :option:`ignore_zone_limits` is specified or the target
1089 device does not have the max_open_zones limit, :option:`max_open_zones`
1090 can specify 0 to disable any limit on the number of zones that can be
1091 simultaneously written to by all jobs.
1093 .. option:: job_max_open_zones=int
1095 In the same manner as :option:`max_open_zones`, limit the number of open
1096 zones per fio job, that is, the number of zones that a single job can
1097 simultaneously write to. A value of zero indicates no limit.
1100 .. option:: ignore_zone_limits=bool
1102 If this option is used, fio will ignore the maximum number of open
1103 zones limit of the zoned block device in use, thus allowing the
1104 option :option:`max_open_zones` value to be larger than the device
1105 reported limit. Default: false.
1107 .. option:: zone_reset_threshold=float
1109 A number between zero and one that indicates the ratio of written bytes
1110 in the zones with write pointers in the IO range to the size of the IO
1111 range. When current ratio is above this ratio, zones are reset
1112 periodically as :option:`zone_reset_frequency` specifies. If there are
1113 multiple jobs when using this option, the IO range for all write jobs
1116 .. option:: zone_reset_frequency=float
1118 A number between zero and one that indicates how often a zone reset
1119 should be issued if the zone reset threshold has been exceeded. A zone
1120 reset is submitted after each (1 / zone_reset_frequency) write
1121 requests. This and the previous parameter can be used to simulate
1122 garbage collection activity.
1128 .. option:: direct=bool
1130 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
1131 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
1132 ioengines don't support direct I/O. Default: false.
1134 .. option:: buffered=bool
1136 If value is true, use buffered I/O. This is the opposite of the
1137 :option:`direct` option. Defaults to true.
1139 .. option:: readwrite=str, rw=str
1141 Type of I/O pattern. Accepted values are:
1148 Sequential trims (Linux block devices and SCSI
1149 character devices only).
1155 Random trims (Linux block devices and SCSI
1156 character devices only).
1158 Sequential mixed reads and writes.
1160 Random mixed reads and writes.
1162 Sequential trim+write sequences. Blocks will be trimmed first,
1163 then the same blocks will be written to. So if ``io_size=64K``
1164 is specified, Fio will trim a total of 64K bytes and also
1165 write 64K bytes on the same trimmed blocks. This behaviour
1166 will be consistent with ``number_ios`` or other Fio options
1167 limiting the total bytes or number of I/O's.
1169 Like trimwrite, but uses random offsets rather
1170 than sequential writes.
1172 Fio defaults to read if the option is not specified. For the mixed I/O
1173 types, the default is to split them 50/50. For certain types of I/O the
1174 result may still be skewed a bit, since the speed may be different.
1176 It is possible to specify the number of I/Os to do before getting a new
1177 offset by appending ``:<nr>`` to the end of the string given. For a
1178 random read, it would look like ``rw=randread:8`` for passing in an offset
1179 modifier with a value of 8. If the suffix is used with a sequential I/O
1180 pattern, then the *<nr>* value specified will be **added** to the generated
1181 offset for each I/O turning sequential I/O into sequential I/O with holes.
1182 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1183 the :option:`rw_sequencer` option.
1185 .. option:: rw_sequencer=str
1187 If an offset modifier is given by appending a number to the ``rw=<str>``
1188 line, then this option controls how that number modifies the I/O offset
1189 being generated. Accepted values are:
1192 Generate sequential offset.
1194 Generate the same offset.
1196 ``sequential`` is only useful for random I/O, where fio would normally
1197 generate a new random offset for every I/O. If you append e.g. 8 to
1198 randread, i.e. ``rw=randread:8`` you would get a new random offset for
1199 every 8 I/Os. The result would be a sequence of 8 sequential offsets
1200 with a random starting point. However this behavior may change if a
1201 sequential I/O reaches end of the file. As sequential I/O is already
1202 sequential, setting ``sequential`` for that would not result in any
1203 difference. ``identical`` behaves in a similar fashion, except it sends
1204 the same offset 8 number of times before generating a new offset.
1209 rw_sequencer=sequential
1212 The generated sequence of offsets will look like this:
1213 4k, 8k, 12k, 16k, 20k, 24k, 28k, 32k, 92k, 96k, 100k, 104k, 108k,
1214 112k, 116k, 120k, 48k, 52k ...
1219 rw_sequencer=identical
1222 The generated sequence of offsets will look like this:
1223 4k, 4k, 4k, 4k, 4k, 4k, 4k, 4k, 92k, 92k, 92k, 92k, 92k, 92k, 92k, 92k,
1226 .. option:: unified_rw_reporting=str
1228 Fio normally reports statistics on a per data direction basis, meaning that
1229 reads, writes, and trims are accounted and reported separately. This option
1230 determines whether fio reports the results normally, summed together, or as
1232 Accepted values are:
1235 Normal statistics reporting.
1238 Statistics are summed per data direction and reported together.
1241 Statistics are reported normally, followed by the mixed statistics.
1244 Backward-compatible alias for **none**.
1247 Backward-compatible alias for **mixed**.
1252 .. option:: randrepeat=bool
1254 Seed all random number generators in a predictable way so the pattern
1255 is repeatable across runs. Default: true.
1257 .. option:: allrandrepeat=bool
1259 Alias for :option:`randrepeat`. Default: true.
1261 .. option:: randseed=int
1263 Seed the random number generators based on this seed value, to be able to
1264 control what sequence of output is being generated. If not set, the random
1265 sequence depends on the :option:`randrepeat` setting.
1267 .. option:: fallocate=str
1269 Whether pre-allocation is performed when laying down files.
1270 Accepted values are:
1273 Do not pre-allocate space.
1276 Use a platform's native pre-allocation call but fall back to
1277 **none** behavior if it fails/is not implemented.
1280 Pre-allocate via :manpage:`posix_fallocate(3)`.
1283 Pre-allocate via :manpage:`fallocate(2)` with
1284 FALLOC_FL_KEEP_SIZE set.
1287 Extend file to final size via :manpage:`ftruncate(2)`
1288 instead of allocating.
1291 Backward-compatible alias for **none**.
1294 Backward-compatible alias for **posix**.
1296 May not be available on all supported platforms. **keep** is only available
1297 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1298 because ZFS doesn't support pre-allocation. Default: **native** if any
1299 pre-allocation methods except **truncate** are available, **none** if not.
1301 Note that using **truncate** on Windows will interact surprisingly
1302 with non-sequential write patterns. When writing to a file that has
1303 been extended by setting the end-of-file information, Windows will
1304 backfill the unwritten portion of the file up to that offset with
1305 zeroes before issuing the new write. This means that a single small
1306 write to the end of an extended file will stall until the entire
1307 file has been filled with zeroes.
1309 .. option:: fadvise_hint=str
1311 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1312 advise the kernel on what I/O patterns are likely to be issued.
1313 Accepted values are:
1316 Backwards-compatible hint for "no hint".
1319 Backwards compatible hint for "advise with fio workload type". This
1320 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1321 for a sequential workload.
1324 Advise using **FADV_SEQUENTIAL**.
1327 Advise using **FADV_RANDOM**.
1330 Advise using **FADV_NOREUSE**. This may be a no-op on older Linux
1331 kernels. Since Linux 6.3, it provides a hint to the LRU algorithm.
1332 See the :manpage:`posix_fadvise(2)` man page.
1334 .. option:: write_hint=str
1336 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1337 from a write. Only supported on Linux, as of version 4.13. Accepted
1341 No particular life time associated with this file.
1344 Data written to this file has a short life time.
1347 Data written to this file has a medium life time.
1350 Data written to this file has a long life time.
1353 Data written to this file has a very long life time.
1355 The values are all relative to each other, and no absolute meaning
1356 should be associated with them.
1358 .. option:: offset=int
1360 Start I/O at the provided offset in the file, given as either a fixed size in
1361 bytes, zones or a percentage. If a percentage is given, the generated offset will be
1362 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1363 provided. Data before the given offset will not be touched. This
1364 effectively caps the file size at `real_size - offset`. Can be combined with
1365 :option:`size` to constrain the start and end range of the I/O workload.
1366 A percentage can be specified by a number between 1 and 100 followed by '%',
1367 for example, ``offset=20%`` to specify 20%. In ZBD mode, value can be set as
1368 number of zones using 'z'.
1370 .. option:: offset_align=int
1372 If set to non-zero value, the byte offset generated by a percentage ``offset``
1373 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1374 offset is aligned to the minimum block size.
1376 .. option:: offset_increment=int
1378 If this is provided, then the real offset becomes `offset + offset_increment
1379 * thread_number`, where the thread number is a counter that starts at 0 and
1380 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1381 specified). This option is useful if there are several jobs which are
1382 intended to operate on a file in parallel disjoint segments, with even
1383 spacing between the starting points. Percentages can be used for this option.
1384 If a percentage is given, the generated offset will be aligned to the minimum
1385 ``blocksize`` or to the value of ``offset_align`` if provided. In ZBD mode, value can
1386 also be set as number of zones using 'z'.
1388 .. option:: number_ios=int
1390 Fio will normally perform I/Os until it has exhausted the size of the region
1391 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1392 condition). With this setting, the range/size can be set independently of
1393 the number of I/Os to perform. When fio reaches this number, it will exit
1394 normally and report status. Note that this does not extend the amount of I/O
1395 that will be done, it will only stop fio if this condition is met before
1396 other end-of-job criteria.
1398 .. option:: fsync=int
1400 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1401 the dirty data for every number of blocks given. For example, if you give 32
1402 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1403 using non-buffered I/O, we may not sync the file. The exception is the sg
1404 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1405 means fio does not periodically issue and wait for a sync to complete. Also
1406 see :option:`end_fsync` and :option:`fsync_on_close`.
1408 .. option:: fdatasync=int
1410 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1411 not metadata blocks. In Windows, DragonFlyBSD or OSX there is no
1412 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1413 Defaults to 0, which means fio does not periodically issue and wait for a
1414 data-only sync to complete.
1416 .. option:: write_barrier=int
1418 Make every `N-th` write a barrier write.
1420 .. option:: sync_file_range=str:int
1422 Use :manpage:`sync_file_range(2)` for every `int` number of write
1423 operations. Fio will track range of writes that have happened since the last
1424 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1427 SYNC_FILE_RANGE_WAIT_BEFORE
1429 SYNC_FILE_RANGE_WRITE
1431 SYNC_FILE_RANGE_WAIT_AFTER
1433 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1434 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1435 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1438 .. option:: overwrite=bool
1440 If true, writes to a file will always overwrite existing data. If the file
1441 doesn't already exist, it will be created before the write phase begins. If
1442 the file exists and is large enough for the specified write phase, nothing
1443 will be done. Default: false.
1445 .. option:: end_fsync=bool
1447 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1450 .. option:: fsync_on_close=bool
1452 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1453 from :option:`end_fsync` in that it will happen on every file close, not
1454 just at the end of the job. Default: false.
1456 .. option:: rwmixread=int
1458 Percentage of a mixed workload that should be reads. Default: 50.
1460 .. option:: rwmixwrite=int
1462 Percentage of a mixed workload that should be writes. If both
1463 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1464 add up to 100%, the latter of the two will be used to override the
1465 first. This may interfere with a given rate setting, if fio is asked to
1466 limit reads or writes to a certain rate. If that is the case, then the
1467 distribution may be skewed. Default: 50.
1469 .. option:: random_distribution=str:float[:float][,str:float][,str:float]
1471 By default, fio will use a completely uniform random distribution when asked
1472 to perform random I/O. Sometimes it is useful to skew the distribution in
1473 specific ways, ensuring that some parts of the data is more hot than others.
1474 fio includes the following distribution models:
1477 Uniform random distribution
1486 Normal (Gaussian) distribution
1489 Zoned random distribution
1492 Zone absolute random distribution
1494 When using a **zipf** or **pareto** distribution, an input value is also
1495 needed to define the access pattern. For **zipf**, this is the `Zipf
1496 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1497 program, :command:`fio-genzipf`, that can be used visualize what the given input
1498 values will yield in terms of hit rates. If you wanted to use **zipf** with
1499 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1500 option. If a non-uniform model is used, fio will disable use of the random
1501 map. For the **normal** distribution, a normal (Gaussian) deviation is
1502 supplied as a value between 0 and 100.
1504 The second, optional float is allowed for **pareto**, **zipf** and **normal** distributions.
1505 It allows one to set base of distribution in non-default place, giving more control
1506 over most probable outcome. This value is in range [0-1] which maps linearly to
1507 range of possible random values.
1508 Defaults are: random for **pareto** and **zipf**, and 0.5 for **normal**.
1509 If you wanted to use **zipf** with a `theta` of 1.2 centered on 1/4 of allowed value range,
1510 you would use ``random_distribution=zipf:1.2:0.25``.
1512 For a **zoned** distribution, fio supports specifying percentages of I/O
1513 access that should fall within what range of the file or device. For
1514 example, given a criteria of:
1516 * 60% of accesses should be to the first 10%
1517 * 30% of accesses should be to the next 20%
1518 * 8% of accesses should be to the next 30%
1519 * 2% of accesses should be to the next 40%
1521 we can define that through zoning of the random accesses. For the above
1522 example, the user would do::
1524 random_distribution=zoned:60/10:30/20:8/30:2/40
1526 A **zoned_abs** distribution works exactly like the **zoned**, except
1527 that it takes absolute sizes. For example, let's say you wanted to
1528 define access according to the following criteria:
1530 * 60% of accesses should be to the first 20G
1531 * 30% of accesses should be to the next 100G
1532 * 10% of accesses should be to the next 500G
1534 we can define an absolute zoning distribution with:
1536 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1538 For both **zoned** and **zoned_abs**, fio supports defining up to
1541 Similarly to how :option:`bssplit` works for setting ranges and
1542 percentages of block sizes. Like :option:`bssplit`, it's possible to
1543 specify separate zones for reads, writes, and trims. If just one set
1544 is given, it'll apply to all of them. This goes for both **zoned**
1545 **zoned_abs** distributions.
1547 .. option:: percentage_random=int[,int][,int]
1549 For a random workload, set how big a percentage should be random. This
1550 defaults to 100%, in which case the workload is fully random. It can be set
1551 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1552 sequential. Any setting in between will result in a random mix of sequential
1553 and random I/O, at the given percentages. Comma-separated values may be
1554 specified for reads, writes, and trims as described in :option:`blocksize`.
1556 .. option:: norandommap
1558 Normally fio will cover every block of the file when doing random I/O. If
1559 this option is given, fio will just get a new random offset without looking
1560 at past I/O history. This means that some blocks may not be read or written,
1561 and that some blocks may be read/written more than once. If this option is
1562 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1563 only intact blocks are verified, i.e., partially-overwritten blocks are
1564 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1565 the same block to be overwritten, which can cause verification errors. Either
1566 do not use norandommap in this case, or also use the lfsr random generator.
1568 .. option:: softrandommap=bool
1570 See :option:`norandommap`. If fio runs with the random block map enabled and
1571 it fails to allocate the map, if this option is set it will continue without
1572 a random block map. As coverage will not be as complete as with random maps,
1573 this option is disabled by default.
1575 .. option:: random_generator=str
1577 Fio supports the following engines for generating I/O offsets for random I/O:
1580 Strong 2^88 cycle random number generator.
1582 Linear feedback shift register generator.
1584 Strong 64-bit 2^258 cycle random number generator.
1586 **tausworthe** is a strong random number generator, but it requires tracking
1587 on the side if we want to ensure that blocks are only read or written
1588 once. **lfsr** guarantees that we never generate the same offset twice, and
1589 it's also less computationally expensive. It's not a true random generator,
1590 however, though for I/O purposes it's typically good enough. **lfsr** only
1591 works with single block sizes, not with workloads that use multiple block
1592 sizes. If used with such a workload, fio may read or write some blocks
1593 multiple times. The default value is **tausworthe**, unless the required
1594 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1595 selected automatically.
1601 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1603 The block size in bytes used for I/O units. Default: 4096. A single value
1604 applies to reads, writes, and trims. Comma-separated values may be
1605 specified for reads, writes, and trims. A value not terminated in a comma
1606 applies to subsequent types.
1611 means 256k for reads, writes and trims.
1614 means 8k for reads, 32k for writes and trims.
1617 means 8k for reads, 32k for writes, and default for trims.
1620 means default for reads, 8k for writes and trims.
1623 means default for reads, 8k for writes, and default for trims.
1625 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1627 A range of block sizes in bytes for I/O units. The issued I/O unit will
1628 always be a multiple of the minimum size, unless
1629 :option:`blocksize_unaligned` is set.
1631 Comma-separated ranges may be specified for reads, writes, and trims as
1632 described in :option:`blocksize`.
1634 Example: ``bsrange=1k-4k,2k-8k`` also the ':' delimiter ``bsrange=1k:4k,2k:8k``.
1636 .. option:: bssplit=str[,str][,str]
1638 Sometimes you want even finer grained control of the block sizes
1639 issued, not just an even split between them. This option allows you to
1640 weight various block sizes, so that you are able to define a specific
1641 amount of block sizes issued. The format for this option is::
1643 bssplit=blocksize/percentage:blocksize/percentage
1645 for as many block sizes as needed. So if you want to define a workload
1646 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1649 bssplit=4k/10:64k/50:32k/40
1651 Ordering does not matter. If the percentage is left blank, fio will
1652 fill in the remaining values evenly. So a bssplit option like this one::
1654 bssplit=4k/50:1k/:32k/
1656 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1657 add up to 100, if bssplit is given a range that adds up to more, it
1660 Comma-separated values may be specified for reads, writes, and trims as
1661 described in :option:`blocksize`.
1663 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1664 having 90% 4k writes and 10% 8k writes, you would specify::
1666 bssplit=2k/50:4k/50,4k/90:8k/10
1668 Fio supports defining up to 64 different weights for each data
1671 .. option:: blocksize_unaligned, bs_unaligned
1673 If set, fio will issue I/O units with any size within
1674 :option:`blocksize_range`, not just multiples of the minimum size. This
1675 typically won't work with direct I/O, as that normally requires sector
1678 .. option:: bs_is_seq_rand=bool
1680 If this option is set, fio will use the normal read,write blocksize settings
1681 as sequential,random blocksize settings instead. Any random read or write
1682 will use the WRITE blocksize settings, and any sequential read or write will
1683 use the READ blocksize settings.
1685 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1687 Boundary to which fio will align random I/O units. Default:
1688 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1689 I/O, though it usually depends on the hardware block size. This option is
1690 mutually exclusive with using a random map for files, so it will turn off
1691 that option. Comma-separated values may be specified for reads, writes, and
1692 trims as described in :option:`blocksize`.
1698 .. option:: zero_buffers
1700 Initialize buffers with all zeros. Default: fill buffers with random data.
1702 .. option:: refill_buffers
1704 If this option is given, fio will refill the I/O buffers on every
1705 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1706 naturally. Defaults to being unset i.e., the buffer is only filled at
1707 init time and the data in it is reused when possible but if any of
1708 :option:`verify`, :option:`buffer_compress_percentage` or
1709 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1710 automatically enabled.
1712 .. option:: scramble_buffers=bool
1714 If :option:`refill_buffers` is too costly and the target is using data
1715 deduplication, then setting this option will slightly modify the I/O buffer
1716 contents to defeat normal de-dupe attempts. This is not enough to defeat
1717 more clever block compression attempts, but it will stop naive dedupe of
1718 blocks. Default: true.
1720 .. option:: buffer_compress_percentage=int
1722 If this is set, then fio will attempt to provide I/O buffer content
1723 (on WRITEs) that compresses to the specified level. Fio does this by
1724 providing a mix of random data followed by fixed pattern data. The
1725 fixed pattern is either zeros, or the pattern specified by
1726 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1727 might skew the compression ratio slightly. Setting
1728 `buffer_compress_percentage` to a value other than 100 will also
1729 enable :option:`refill_buffers` in order to reduce the likelihood that
1730 adjacent blocks are so similar that they over compress when seen
1731 together. See :option:`buffer_compress_chunk` for how to set a finer or
1732 coarser granularity for the random/fixed data region. Defaults to unset
1733 i.e., buffer data will not adhere to any compression level.
1735 .. option:: buffer_compress_chunk=int
1737 This setting allows fio to manage how big the random/fixed data region
1738 is when using :option:`buffer_compress_percentage`. When
1739 `buffer_compress_chunk` is set to some non-zero value smaller than the
1740 block size, fio can repeat the random/fixed region throughout the I/O
1741 buffer at the specified interval (which particularly useful when
1742 bigger block sizes are used for a job). When set to 0, fio will use a
1743 chunk size that matches the block size resulting in a single
1744 random/fixed region within the I/O buffer. Defaults to 512. When the
1745 unit is omitted, the value is interpreted in bytes.
1747 .. option:: buffer_pattern=str
1749 If set, fio will fill the I/O buffers with this pattern or with the contents
1750 of a file. If not set, the contents of I/O buffers are defined by the other
1751 options related to buffer contents. The setting can be any pattern of bytes,
1752 and can be prefixed with 0x for hex values. It may also be a string, where
1753 the string must then be wrapped with ``""``. Or it may also be a filename,
1754 where the filename must be wrapped with ``''`` in which case the file is
1755 opened and read. Note that not all the file contents will be read if that
1756 would cause the buffers to overflow. So, for example::
1758 buffer_pattern='filename'
1762 buffer_pattern="abcd"
1770 buffer_pattern=0xdeadface
1772 Also you can combine everything together in any order::
1774 buffer_pattern=0xdeadface"abcd"-12'filename'
1776 .. option:: dedupe_percentage=int
1778 If set, fio will generate this percentage of identical buffers when
1779 writing. These buffers will be naturally dedupable. The contents of the
1780 buffers depend on what other buffer compression settings have been set. It's
1781 possible to have the individual buffers either fully compressible, or not at
1782 all -- this option only controls the distribution of unique buffers. Setting
1783 this option will also enable :option:`refill_buffers` to prevent every buffer
1786 .. option:: dedupe_mode=str
1788 If ``dedupe_percentage=<int>`` is given, then this option controls how fio
1789 generates the dedupe buffers.
1792 Generate dedupe buffers by repeating previous writes
1794 Generate dedupe buffers from working set
1796 ``repeat`` is the default option for fio. Dedupe buffers are generated
1797 by repeating previous unique write.
1799 ``working_set`` is a more realistic workload.
1800 With ``working_set``, ``dedupe_working_set_percentage=<int>`` should be provided.
1801 Given that, fio will use the initial unique write buffers as its working set.
1802 Upon deciding to dedupe, fio will randomly choose a buffer from the working set.
1803 Note that by using ``working_set`` the dedupe percentage will converge
1804 to the desired over time while ``repeat`` maintains the desired percentage
1807 .. option:: dedupe_working_set_percentage=int
1809 If ``dedupe_mode=<str>`` is set to ``working_set``, then this controls
1810 the percentage of size of the file or device used as the buffers
1811 fio will choose to generate the dedupe buffers from
1813 Note that size needs to be explicitly provided and only 1 file per
1816 .. option:: dedupe_global=bool
1818 This controls whether the deduplication buffers will be shared amongst
1819 all jobs that have this option set. The buffers are spread evenly between
1822 .. option:: invalidate=bool
1824 Invalidate the buffer/page cache parts of the files to be used prior to
1825 starting I/O if the platform and file type support it. Defaults to true.
1826 This will be ignored if :option:`pre_read` is also specified for the
1829 .. option:: sync=str
1831 Whether, and what type, of synchronous I/O to use for writes. The allowed
1835 Do not use synchronous IO, the default.
1841 Use synchronous file IO. For the majority of I/O engines,
1842 this means using O_SYNC.
1848 Use synchronous data IO. For the majority of I/O engines,
1849 this means using O_DSYNC.
1852 .. option:: iomem=str, mem=str
1854 Fio can use various types of memory as the I/O unit buffer. The allowed
1858 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1862 Use shared memory as the buffers. Allocated through
1863 :manpage:`shmget(2)`.
1866 Same as shm, but use huge pages as backing.
1869 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1870 be file backed if a filename is given after the option. The format
1871 is `mem=mmap:/path/to/file`.
1874 Use a memory mapped huge file as the buffer backing. Append filename
1875 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1878 Same as mmap, but use a MMAP_SHARED mapping.
1881 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1882 The :option:`ioengine` must be `rdma`.
1884 The area allocated is a function of the maximum allowed bs size for the job,
1885 multiplied by the I/O depth given. Note that for **shmhuge** and
1886 **mmaphuge** to work, the system must have free huge pages allocated. This
1887 can normally be checked and set by reading/writing
1888 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1889 is 2 or 4MiB in size depending on the platform. So to calculate the
1890 number of huge pages you need for a given job file, add up the I/O
1891 depth of all jobs (normally one unless :option:`iodepth` is used) and
1892 multiply by the maximum bs set. Then divide that number by the huge
1893 page size. You can see the size of the huge pages in
1894 :file:`/proc/meminfo`. If no huge pages are allocated by having a
1895 non-zero number in `nr_hugepages`, using **mmaphuge** or **shmhuge**
1896 will fail. Also see :option:`hugepage-size`.
1898 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1899 should point there. So if it's mounted in :file:`/huge`, you would use
1900 `mem=mmaphuge:/huge/somefile`.
1902 .. option:: iomem_align=int, mem_align=int
1904 This indicates the memory alignment of the I/O memory buffers. Note that
1905 the given alignment is applied to the first I/O unit buffer, if using
1906 :option:`iodepth` the alignment of the following buffers are given by the
1907 :option:`bs` used. In other words, if using a :option:`bs` that is a
1908 multiple of the page sized in the system, all buffers will be aligned to
1909 this value. If using a :option:`bs` that is not page aligned, the alignment
1910 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1913 .. option:: hugepage-size=int
1915 Defines the size of a huge page. Must at least be equal to the system
1916 setting, see :file:`/proc/meminfo` and
1917 :file:`/sys/kernel/mm/hugepages/`. Defaults to 2 or 4MiB depending on
1918 the platform. Should probably always be a multiple of megabytes, so
1919 using ``hugepage-size=Xm`` is the preferred way to set this to avoid
1920 setting a non-pow-2 bad value.
1922 .. option:: lockmem=int
1924 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1925 simulate a smaller amount of memory. The amount specified is per worker.
1931 .. option:: size=int
1933 The total size of file I/O for each thread of this job. Fio will run until
1934 this many bytes has been transferred, unless runtime is altered by other means
1935 such as (1) :option:`runtime`, (2) :option:`io_size` (3) :option:`number_ios`,
1936 (4) gaps/holes while doing I/O's such as ``rw=read:16K``, or (5) sequential
1937 I/O reaching end of the file which is possible when :option:`percentage_random`
1939 Fio will divide this size between the available files determined by options
1940 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1941 specified by the job. If the result of division happens to be 0, the size is
1942 set to the physical size of the given files or devices if they exist.
1943 If this option is not specified, fio will use the full size of the given
1944 files or devices. If the files do not exist, size must be given. It is also
1945 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1946 given, fio will use 20% of the full size of the given files or devices.
1947 In ZBD mode, value can also be set as number of zones using 'z'.
1948 Can be combined with :option:`offset` to constrain the start and end range
1949 that I/O will be done within.
1951 .. option:: io_size=int, io_limit=int
1953 Normally fio operates within the region set by :option:`size`, which means
1954 that the :option:`size` option sets both the region and size of I/O to be
1955 performed. Sometimes that is not what you want. With this option, it is
1956 possible to define just the amount of I/O that fio should do. For instance,
1957 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1958 will perform I/O within the first 20GiB but exit when 5GiB have been
1959 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1960 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1961 the 0..20GiB region.
1963 .. option:: filesize=irange(int)
1965 Individual file sizes. May be a range, in which case fio will select sizes for
1966 files at random within the given range. If not given, each created file is the
1967 same size. This option overrides :option:`size` in terms of file size, i.e. if
1968 :option:`filesize` is specified then :option:`size` becomes merely the default
1969 for :option:`io_size` and has no effect at all if :option:`io_size` is set
1972 .. option:: file_append=bool
1974 Perform I/O after the end of the file. Normally fio will operate within the
1975 size of a file. If this option is set, then fio will append to the file
1976 instead. This has identical behavior to setting :option:`offset` to the size
1977 of a file. This option is ignored on non-regular files.
1979 .. option:: fill_device=bool, fill_fs=bool
1981 Sets size to something really large and waits for ENOSPC (no space left on
1982 device) or EDQUOT (disk quota exceeded)
1983 as the terminating condition. Only makes sense with sequential
1984 write. For a read workload, the mount point will be filled first then I/O
1985 started on the result. This option doesn't make sense if operating on a raw
1986 device node, since the size of that is already known by the file system.
1987 Additionally, writing beyond end-of-device will not return ENOSPC there.
1993 .. option:: ioengine=str
1995 Defines how the job issues I/O to the file. The following types are defined:
1998 Basic :manpage:`read(2)` or :manpage:`write(2)`
1999 I/O. :manpage:`lseek(2)` is used to position the I/O location.
2000 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
2003 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
2004 all supported operating systems except for Windows.
2007 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
2008 queuing by coalescing adjacent I/Os into a single submission.
2011 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
2014 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
2017 Fast Linux native asynchronous I/O. Supports async IO
2018 for both direct and buffered IO.
2019 This engine defines engine specific options.
2022 Fast Linux native asynchronous I/O for pass through commands.
2023 This engine defines engine specific options.
2026 Linux native asynchronous I/O. Note that Linux may only support
2027 queued behavior with non-buffered I/O (set ``direct=1`` or
2029 This engine defines engine specific options.
2032 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
2033 :manpage:`aio_write(3)`.
2036 Solaris native asynchronous I/O.
2039 Windows native asynchronous I/O. Default on Windows.
2042 File is memory mapped with :manpage:`mmap(2)` and data copied
2043 to/from using :manpage:`memcpy(3)`.
2046 :manpage:`splice(2)` is used to transfer the data and
2047 :manpage:`vmsplice(2)` to transfer data from user space to the
2051 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
2052 ioctl, or if the target is an sg character device we use
2053 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
2054 I/O. Requires :option:`filename` option to specify either block or
2055 character devices. This engine supports trim operations.
2056 The sg engine includes engine specific options.
2059 Read, write, trim and ZBC/ZAC operations to a zoned
2060 block device using libzbc library. The target can be
2061 either an SG character device or a block device file.
2064 Doesn't transfer any data, just pretends to. This is mainly used to
2065 exercise fio itself and for debugging/testing purposes.
2068 Transfer over the network to given ``host:port``. Depending on the
2069 :option:`protocol` used, the :option:`hostname`, :option:`port`,
2070 :option:`listen` and :option:`filename` options are used to specify
2071 what sort of connection to make, while the :option:`protocol` option
2072 determines which protocol will be used. This engine defines engine
2076 Like **net**, but uses :manpage:`splice(2)` and
2077 :manpage:`vmsplice(2)` to map data and send/receive.
2078 This engine defines engine specific options.
2081 Doesn't transfer any data, but burns CPU cycles according to the
2082 :option:`cpuload`, :option:`cpuchunks` and :option:`cpumode` options.
2083 Setting :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
2084 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
2085 to get desired CPU usage, as the cpuload only loads a
2086 single CPU at the desired rate. A job never finishes unless there is
2087 at least one non-cpuio job.
2088 Setting :option:`cpumode`\=qsort replace the default noop instructions loop
2089 by a qsort algorithm to consume more energy.
2092 The RDMA I/O engine supports both RDMA memory semantics
2093 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
2094 InfiniBand, RoCE and iWARP protocols. This engine defines engine
2098 I/O engine that does regular fallocate to simulate data transfer as
2102 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
2105 does fallocate(,mode = 0).
2108 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
2111 I/O engine that sends :manpage:`ftruncate(2)` operations in response
2112 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
2113 size to the current block offset. :option:`blocksize` is ignored.
2116 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
2117 defragment activity in request to DDIR_WRITE event.
2120 I/O engine supporting direct access to Ceph Reliable Autonomic
2121 Distributed Object Store (RADOS) via librados. This ioengine
2122 defines engine specific options.
2125 I/O engine supporting direct access to Ceph Rados Block Devices
2126 (RBD) via librbd without the need to use the kernel rbd driver. This
2127 ioengine defines engine specific options.
2130 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
2131 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
2133 This engine only supports direct IO of iodepth=1; you need to scale this
2134 via numjobs. blocksize defines the size of the objects to be created.
2136 TRIM is translated to object deletion.
2139 Using GlusterFS libgfapi sync interface to direct access to
2140 GlusterFS volumes without having to go through FUSE. This ioengine
2141 defines engine specific options.
2144 Using GlusterFS libgfapi async interface to direct access to
2145 GlusterFS volumes without having to go through FUSE. This ioengine
2146 defines engine specific options.
2149 Read and write through Hadoop (HDFS). The :option:`filename` option
2150 is used to specify host,port of the hdfs name-node to connect. This
2151 engine interprets offsets a little differently. In HDFS, files once
2152 created cannot be modified so random writes are not possible. To
2153 imitate this the libhdfs engine expects a bunch of small files to be
2154 created over HDFS and will randomly pick a file from them
2155 based on the offset generated by fio backend (see the example
2156 job file to create such files, use ``rw=write`` option). Please
2157 note, it may be necessary to set environment variables to work
2158 with HDFS/libhdfs properly. Each job uses its own connection to
2162 Read, write and erase an MTD character device (e.g.,
2163 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
2164 underlying device type, the I/O may have to go in a certain pattern,
2165 e.g., on NAND, writing sequentially to erase blocks and discarding
2166 before overwriting. The `trimwrite` mode works well for this
2170 Read and write using device DAX to a persistent memory device (e.g.,
2171 /dev/dax0.0) through the PMDK libpmem library.
2174 Prefix to specify loading an external I/O engine object file. Append
2175 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
2176 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
2177 absolute or relative. See :file:`engines/skeleton_external.c` for
2178 details of writing an external I/O engine.
2181 Simply create the files and do no I/O to them. You still need to
2182 set `filesize` so that all the accounting still occurs, but no
2183 actual I/O will be done other than creating the file.
2186 Simply do stat() and do no I/O to the file. You need to set 'filesize'
2187 and 'nrfiles', so that files will be created.
2188 This engine is to measure file lookup and meta data access.
2191 Simply delete the files by unlink() and do no I/O to them. You need to set 'filesize'
2192 and 'nrfiles', so that the files will be created.
2193 This engine is to measure file delete.
2196 Read and write using mmap I/O to a file on a filesystem
2197 mounted with DAX on a persistent memory device through the PMDK
2201 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2202 This engine is very basic and issues calls to IME whenever an IO is
2206 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2207 This engine uses iovecs and will try to stack as much IOs as possible
2208 (if the IOs are "contiguous" and the IO depth is not exceeded)
2209 before issuing a call to IME.
2212 Asynchronous read and write using DDN's Infinite Memory Engine (IME).
2213 This engine will try to stack as much IOs as possible by creating
2214 requests for IME. FIO will then decide when to commit these requests.
2217 Read and write iscsi lun with libiscsi.
2220 Read and write a Network Block Device (NBD).
2223 I/O engine supporting libcufile synchronous access to nvidia-fs and a
2224 GPUDirect Storage-supported filesystem. This engine performs
2225 I/O without transferring buffers between user-space and the kernel,
2226 unless :option:`verify` is set or :option:`cuda_io` is `posix`.
2227 :option:`iomem` must not be `cudamalloc`. This ioengine defines
2228 engine specific options.
2231 I/O engine supporting asynchronous read and write operations to the
2232 DAOS File System (DFS) via libdfs.
2235 I/O engine supporting asynchronous read and write operations to
2236 NFS filesystems from userspace via libnfs. This is useful for
2237 achieving higher concurrency and thus throughput than is possible
2241 Execute 3rd party tools. Could be used to perform monitoring during jobs runtime.
2244 I/O engine using the xNVMe C API, for NVMe devices. The xnvme engine provides
2245 flexibility to access GNU/Linux Kernel NVMe driver via libaio, IOCTLs, io_uring,
2246 the SPDK NVMe driver, or your own custom NVMe driver. The xnvme engine includes
2247 engine specific options. (See https://xnvme.io).
2250 Use the libblkio library
2251 (https://gitlab.com/libblkio/libblkio). The specific
2252 *driver* to use must be set using
2253 :option:`libblkio_driver`. If
2254 :option:`mem`/:option:`iomem` is not specified, memory
2255 allocation is delegated to libblkio (and so is
2256 guaranteed to work with the selected *driver*). One
2257 libblkio instance is used per process, so all jobs
2258 setting option :option:`thread` will share a single
2259 instance (with one queue per thread) and must specify
2260 compatible options. Note that some drivers don't allow
2261 several instances to access the same device or file
2262 simultaneously, but allow it for threads.
2264 I/O engine specific parameters
2265 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2267 In addition, there are some parameters which are only valid when a specific
2268 :option:`ioengine` is in use. These are used identically to normal parameters,
2269 with the caveat that when used on the command line, they must come after the
2270 :option:`ioengine` that defines them is selected.
2272 .. option:: cmdprio_percentage=int[,int] : [io_uring] [libaio]
2274 Set the percentage of I/O that will be issued with the highest priority.
2275 Default: 0. A single value applies to reads and writes. Comma-separated
2276 values may be specified for reads and writes. For this option to be
2277 effective, NCQ priority must be supported and enabled, and the :option:`direct`
2278 option must be set. fio must also be run as the root user. Unlike
2279 slat/clat/lat stats, which can be tracked and reported independently, per
2280 priority stats only track and report a single type of latency. By default,
2281 completion latency (clat) will be reported, if :option:`lat_percentiles` is
2282 set, total latency (lat) will be reported.
2284 .. option:: cmdprio_class=int[,int] : [io_uring] [libaio]
2286 Set the I/O priority class to use for I/Os that must be issued with
2287 a priority when :option:`cmdprio_percentage` or
2288 :option:`cmdprio_bssplit` is set. If not specified when
2289 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2290 this defaults to the highest priority class. A single value applies
2291 to reads and writes. Comma-separated values may be specified for
2292 reads and writes. See :manpage:`ionice(1)`. See also the
2293 :option:`prioclass` option.
2295 .. option:: cmdprio_hint=int[,int] : [io_uring] [libaio]
2297 Set the I/O priority hint to use for I/Os that must be issued with
2298 a priority when :option:`cmdprio_percentage` or
2299 :option:`cmdprio_bssplit` is set. If not specified when
2300 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2301 this defaults to 0 (no hint). A single value applies to reads and
2302 writes. Comma-separated values may be specified for reads and writes.
2303 See also the :option:`priohint` option.
2305 .. option:: cmdprio=int[,int] : [io_uring] [libaio]
2307 Set the I/O priority value to use for I/Os that must be issued with
2308 a priority when :option:`cmdprio_percentage` or
2309 :option:`cmdprio_bssplit` is set. If not specified when
2310 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2312 Linux limits us to a positive value between 0 and 7, with 0 being the
2313 highest. A single value applies to reads and writes. Comma-separated
2314 values may be specified for reads and writes. See :manpage:`ionice(1)`.
2315 Refer to an appropriate manpage for other operating systems since
2316 meaning of priority may differ. See also the :option:`prio` option.
2318 .. option:: cmdprio_bssplit=str[,str] : [io_uring] [libaio]
2320 To get a finer control over I/O priority, this option allows
2321 specifying the percentage of IOs that must have a priority set
2322 depending on the block size of the IO. This option is useful only
2323 when used together with the :option:`bssplit` option, that is,
2324 multiple different block sizes are used for reads and writes.
2326 The first accepted format for this option is the same as the format of
2327 the :option:`bssplit` option:
2329 cmdprio_bssplit=blocksize/percentage:blocksize/percentage
2331 In this case, each entry will use the priority class, priority hint
2332 and priority level defined by the options :option:`cmdprio_class`,
2333 :option:`cmdprio` and :option:`cmdprio_hint` respectively.
2335 The second accepted format for this option is:
2337 cmdprio_bssplit=blocksize/percentage/class/level:blocksize/percentage/class/level
2339 In this case, the priority class and priority level is defined inside
2340 each entry. In comparison with the first accepted format, the second
2341 accepted format does not restrict all entries to have the same priority
2342 class and priority level.
2344 The third accepted format for this option is:
2346 cmdprio_bssplit=blocksize/percentage/class/level/hint:...
2348 This is an extension of the second accepted format that allows one to
2349 also specify a priority hint.
2351 For all formats, only the read and write data directions are supported,
2352 values for trim IOs are ignored. This option is mutually exclusive with
2353 the :option:`cmdprio_percentage` option.
2355 .. option:: fixedbufs : [io_uring] [io_uring_cmd]
2357 If fio is asked to do direct IO, then Linux will map pages for each
2358 IO call, and release them when IO is done. If this option is set, the
2359 pages are pre-mapped before IO is started. This eliminates the need to
2360 map and release for each IO. This is more efficient, and reduces the
2363 .. option:: nonvectored=int : [io_uring] [io_uring_cmd]
2365 With this option, fio will use non-vectored read/write commands, where
2366 address must contain the address directly. Default is -1.
2368 .. option:: force_async=int : [io_uring] [io_uring_cmd]
2370 Normal operation for io_uring is to try and issue an sqe as
2371 non-blocking first, and if that fails, execute it in an async manner.
2372 With this option set to N, then every N request fio will ask sqe to
2373 be issued in an async manner. Default is 0.
2375 .. option:: registerfiles : [io_uring] [io_uring_cmd]
2377 With this option, fio registers the set of files being used with the
2378 kernel. This avoids the overhead of managing file counts in the kernel,
2379 making the submission and completion part more lightweight. Required
2380 for the below :option:`sqthread_poll` option.
2382 .. option:: sqthread_poll : [io_uring] [io_uring_cmd] [xnvme]
2384 Normally fio will submit IO by issuing a system call to notify the
2385 kernel of available items in the SQ ring. If this option is set, the
2386 act of submitting IO will be done by a polling thread in the kernel.
2387 This frees up cycles for fio, at the cost of using more CPU in the
2388 system. As submission is just the time it takes to fill in the sqe
2389 entries and any syscall required to wake up the idle kernel thread,
2390 fio will not report submission latencies.
2392 .. option:: sqthread_poll_cpu=int : [io_uring] [io_uring_cmd]
2394 When :option:`sqthread_poll` is set, this option provides a way to
2395 define which CPU should be used for the polling thread.
2397 .. option:: cmd_type=str : [io_uring_cmd]
2399 Specifies the type of uring passthrough command to be used. Supported
2400 value is nvme. Default is nvme.
2404 [io_uring] [io_uring_cmd] [xnvme]
2406 If this option is set, fio will attempt to use polled IO completions.
2407 Normal IO completions generate interrupts to signal the completion of
2408 IO, polled completions do not. Hence they are require active reaping
2409 by the application. The benefits are more efficient IO for high IOPS
2410 scenarios, and lower latencies for low queue depth IO.
2414 Use poll queues. This is incompatible with
2415 :option:`libblkio_wait_mode=eventfd <libblkio_wait_mode>` and
2416 :option:`libblkio_force_enable_completion_eventfd`.
2420 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2425 If this option is set, fio will attempt to use polled IO completions.
2426 This will have a similar effect as (io_uring)hipri. Only SCSI READ and
2427 WRITE commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor
2428 VERIFY). Older versions of the Linux sg driver that do not support
2429 hipri will simply ignore this flag and do normal IO. The Linux SCSI
2430 Low Level Driver (LLD) that "owns" the device also needs to support
2431 hipri (also known as iopoll and mq_poll). The MegaRAID driver is an
2432 example of a SCSI LLD. Default: clear (0) which does normal
2433 (interrupted based) IO.
2435 .. option:: userspace_reap : [libaio]
2437 Normally, with the libaio engine in use, fio will use the
2438 :manpage:`io_getevents(2)` system call to reap newly returned events. With
2439 this flag turned on, the AIO ring will be read directly from user-space to
2440 reap events. The reaping mode is only enabled when polling for a minimum of
2441 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
2443 .. option:: hipri_percentage : [pvsync2]
2445 When hipri is set this determines the probability of a pvsync2 I/O being high
2446 priority. The default is 100%.
2448 .. option:: nowait=bool : [pvsync2] [libaio] [io_uring] [io_uring_cmd]
2450 By default if a request cannot be executed immediately (e.g. resource starvation,
2451 waiting on locks) it is queued and the initiating process will be blocked until
2452 the required resource becomes free.
2454 This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and
2455 the call will return instantly with EAGAIN or a partial result rather than waiting.
2457 It is useful to also use ignore_error=EAGAIN when using this option.
2459 Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2.
2460 They return EOPNOTSUP instead of EAGAIN.
2462 For cached I/O, using this option usually means a request operates only with
2463 cached data. Currently the RWF_NOWAIT flag does not supported for cached write.
2465 For direct I/O, requests will only succeed if cache invalidation isn't required,
2466 file blocks are fully allocated and the disk request could be issued immediately.
2468 .. option:: fdp=bool : [io_uring_cmd] [xnvme]
2470 Enable Flexible Data Placement mode for write commands.
2472 .. option:: fdp_pli_select=str : [io_uring_cmd] [xnvme]
2474 Defines how fio decides which placement ID to use next. The following
2478 Choose a placement ID at random (uniform).
2481 Round robin over available placement IDs. This is the
2484 The available placement ID index/indices is defined by the option
2487 .. option:: fdp_pli=str : [io_uring_cmd] [xnvme]
2489 Select which Placement ID Index/Indicies this job is allowed to use for
2490 writes. By default, the job will cycle through all available Placement
2491 IDs, so use this to isolate these identifiers to specific jobs. If you
2492 want fio to use placement identifier only at indices 0, 2 and 5 specify
2495 .. option:: md_per_io_size=int : [io_uring_cmd] [xnvme]
2497 Size in bytes for separate metadata buffer per IO. Default: 0.
2499 .. option:: pi_act=int : [io_uring_cmd] [xnvme]
2501 Action to take when nvme namespace is formatted with protection
2502 information. If this is set to 1 and namespace is formatted with
2503 metadata size equal to protection information size, fio won't use
2504 separate metadata buffer or extended logical block. If this is set to
2505 1 and namespace is formatted with metadata size greater than protection
2506 information size, fio will not generate or verify the protection
2507 information portion of metadata for write or read case respectively.
2508 If this is set to 0, fio generates protection information for
2509 write case and verifies for read case. Default: 1.
2511 For 16 bit CRC generation fio will use isa-l if available otherwise
2512 it will use the default slower generator.
2513 (see: https://github.com/intel/isa-l)
2515 .. option:: pi_chk=str[,str][,str] : [io_uring_cmd] [xnvme]
2517 Controls the protection information check. This can take one or more
2518 of these values. Default: none.
2521 Enables protection information checking of guard field.
2523 Enables protection information checking of logical block
2524 reference tag field.
2526 Enables protection information checking of application tag field.
2528 .. option:: apptag=int : [io_uring_cmd] [xnvme]
2530 Specifies logical block application tag value, if namespace is
2531 formatted to use end to end protection information. Default: 0x1234.
2533 .. option:: apptag_mask=int : [io_uring_cmd] [xnvme]
2535 Specifies logical block application tag mask value, if namespace is
2536 formatted to use end to end protection information. Default: 0xffff.
2538 .. option:: num_range=int : [io_uring_cmd]
2540 For trim command this will be the number of ranges to trim per I/O
2541 request. The number of logical blocks per range is determined by the
2542 :option:`bs` option which should be a multiple of logical block size.
2543 This cannot be used with read or write. Note that setting this
2544 option > 1, :option:`log_offset` will not be able to log all the
2545 offsets. Default: 1.
2547 .. option:: cpuload=int : [cpuio]
2549 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2550 option when using cpuio I/O engine.
2552 .. option:: cpuchunks=int : [cpuio]
2554 Split the load into cycles of the given time. In microseconds.
2556 .. option:: cpumode=str : [cpuio]
2558 Specify how to stress the CPU. It can take these two values:
2561 This is the default where the CPU executes noop instructions.
2563 Replace the default noop instructions loop with a qsort algorithm to
2564 consume more energy.
2566 .. option:: exit_on_io_done=bool : [cpuio]
2568 Detect when I/O threads are done, then exit.
2570 .. option:: namenode=str : [libhdfs]
2572 The hostname or IP address of a HDFS cluster namenode to contact.
2574 .. option:: port=int
2578 The listening port of the HFDS cluster namenode.
2582 The TCP or UDP port to bind to or connect to. If this is used with
2583 :option:`numjobs` to spawn multiple instances of the same job type, then
2584 this will be the starting port number since fio will use a range of
2589 The port to use for RDMA-CM communication. This should be the same value
2590 on the client and the server side.
2592 .. option:: hostname=str : [netsplice] [net] [rdma]
2594 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
2595 is a TCP listener or UDP reader, the hostname is not used and must be omitted
2596 unless it is a valid UDP multicast address.
2598 .. option:: serverip=str : [librpma_*]
2600 The IP address to be used for RDMA-CM based I/O.
2602 .. option:: direct_write_to_pmem=bool : [librpma_*]
2604 Set to 1 only when Direct Write to PMem from the remote host is possible.
2605 Otherwise, set to 0.
2607 .. option:: busy_wait_polling=bool : [librpma_*_server]
2609 Set to 0 to wait for completion instead of busy-wait polling completion.
2612 .. option:: interface=str : [netsplice] [net]
2614 The IP address of the network interface used to send or receive UDP
2617 .. option:: ttl=int : [netsplice] [net]
2619 Time-to-live value for outgoing UDP multicast packets. Default: 1.
2621 .. option:: nodelay=bool : [netsplice] [net]
2623 Set TCP_NODELAY on TCP connections.
2625 .. option:: protocol=str, proto=str : [netsplice] [net]
2627 The network protocol to use. Accepted values are:
2630 Transmission control protocol.
2632 Transmission control protocol V6.
2634 User datagram protocol.
2636 User datagram protocol V6.
2642 When the protocol is TCP, UDP or VSOCK, the port must also be given, as well as the
2643 hostname if the job is a TCP or VSOCK listener or UDP reader. For unix sockets, the
2644 normal :option:`filename` option should be used and the port is invalid.
2645 When the protocol is VSOCK, the :option:`hostname` is the CID of the remote VM.
2647 .. option:: listen : [netsplice] [net]
2649 For TCP network connections, tell fio to listen for incoming connections
2650 rather than initiating an outgoing connection. The :option:`hostname` must
2651 be omitted if this option is used.
2653 .. option:: pingpong : [netsplice] [net]
2655 Normally a network writer will just continue writing data, and a network
2656 reader will just consume packages. If ``pingpong=1`` is set, a writer will
2657 send its normal payload to the reader, then wait for the reader to send the
2658 same payload back. This allows fio to measure network latencies. The
2659 submission and completion latencies then measure local time spent sending or
2660 receiving, and the completion latency measures how long it took for the
2661 other end to receive and send back. For UDP multicast traffic
2662 ``pingpong=1`` should only be set for a single reader when multiple readers
2663 are listening to the same address.
2665 .. option:: window_size : [netsplice] [net]
2667 Set the desired socket buffer size for the connection.
2669 .. option:: mss : [netsplice] [net]
2671 Set the TCP maximum segment size (TCP_MAXSEG).
2673 .. option:: donorname=str : [e4defrag]
2675 File will be used as a block donor (swap extents between files).
2677 .. option:: inplace=int : [e4defrag]
2679 Configure donor file blocks allocation strategy:
2682 Default. Preallocate donor's file on init.
2684 Allocate space immediately inside defragment event, and free right
2687 .. option:: clustername=str : [rbd,rados]
2689 Specifies the name of the Ceph cluster.
2691 .. option:: rbdname=str : [rbd]
2693 Specifies the name of the RBD.
2695 .. option:: clientname=str : [rbd,rados]
2697 Specifies the username (without the 'client.' prefix) used to access the
2698 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2699 the full *type.id* string. If no type. prefix is given, fio will add
2700 'client.' by default.
2702 .. option:: conf=str : [rados]
2704 Specifies the configuration path of ceph cluster, so conf file does not
2705 have to be /etc/ceph/ceph.conf.
2707 .. option:: busy_poll=bool : [rbd,rados]
2709 Poll store instead of waiting for completion. Usually this provides better
2710 throughput at cost of higher(up to 100%) CPU utilization.
2712 .. option:: touch_objects=bool : [rados]
2714 During initialization, touch (create if do not exist) all objects (files).
2715 Touching all objects affects ceph caches and likely impacts test results.
2718 .. option:: pool=str :
2722 Specifies the name of the Ceph pool containing RBD or RADOS data.
2726 Specify the label or UUID of the DAOS pool to connect to.
2728 .. option:: cont=str : [dfs]
2730 Specify the label or UUID of the DAOS container to open.
2732 .. option:: chunk_size=int
2736 Specify a different chunk size (in bytes) for the dfs file.
2737 Use DAOS container's chunk size by default.
2741 The size of the chunk to use for each file.
2743 .. option:: object_class=str : [dfs]
2745 Specify a different object class for the dfs file.
2746 Use DAOS container's object class by default.
2748 .. option:: skip_bad=bool : [mtd]
2750 Skip operations against known bad blocks.
2752 .. option:: hdfsdirectory : [libhdfs]
2754 libhdfs will create chunk in this HDFS directory.
2756 .. option:: verb=str : [rdma]
2758 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2759 values are write, read, send and recv. These correspond to the equivalent
2760 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2761 specified on the client side of the connection. See the examples folder.
2763 .. option:: bindname=str : [rdma]
2765 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2766 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2767 will be passed into the rdma_bind_addr() function and on the client site it
2768 will be used in the rdma_resolve_add() function. This can be useful when
2769 multiple paths exist between the client and the server or in certain loopback
2772 .. option:: stat_type=str : [filestat]
2774 Specify stat system call type to measure lookup/getattr performance.
2775 Default is **stat** for :manpage:`stat(2)`.
2777 .. option:: readfua=bool : [sg]
2779 With readfua option set to 1, read operations include
2780 the force unit access (fua) flag. Default is 0.
2782 .. option:: writefua=bool : [sg]
2784 With writefua option set to 1, write operations include
2785 the force unit access (fua) flag. Default is 0.
2787 .. option:: sg_write_mode=str : [sg]
2789 Specify the type of write commands to issue. This option can take ten values:
2792 This is the default where write opcodes are issued as usual.
2793 **write_and_verify**
2794 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2795 directs the device to carry out a medium verification with no data
2796 comparison. The writefua option is ignored with this selection.
2798 This option is deprecated. Use write_and_verify instead.
2800 Issue WRITE SAME commands. This transfers a single block to the device
2801 and writes this same block of data to a contiguous sequence of LBAs
2802 beginning at the specified offset. fio's block size parameter specifies
2803 the amount of data written with each command. However, the amount of data
2804 actually transferred to the device is equal to the device's block
2805 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2806 write 16 sectors with each command. fio will still generate 8k of data
2807 for each command but only the first 512 bytes will be used and
2808 transferred to the device. The writefua option is ignored with this
2811 This option is deprecated. Use write_same instead.
2813 Issue WRITE SAME(16) commands as above but with the No Data Output
2814 Buffer (NDOB) bit set. No data will be transferred to the device with
2815 this bit set. Data written will be a pre-determined pattern such as
2818 Issue WRITE STREAM(16) commands. Use the **stream_id** option to specify
2819 the stream identifier.
2820 **verify_bytchk_00**
2821 Issue VERIFY commands with BYTCHK set to 00. This directs the
2822 device to carry out a medium verification with no data comparison.
2823 **verify_bytchk_01**
2824 Issue VERIFY commands with BYTCHK set to 01. This directs the device to
2825 compare the data on the device with the data transferred to the device.
2826 **verify_bytchk_11**
2827 Issue VERIFY commands with BYTCHK set to 11. This transfers a
2828 single block to the device and compares the contents of this block with the
2829 data on the device beginning at the specified offset. fio's block size
2830 parameter specifies the total amount of data compared with this command.
2831 However, only one block (sector) worth of data is transferred to the device.
2832 This is similar to the WRITE SAME command except that data is compared instead
2835 .. option:: stream_id=int : [sg]
2837 Set the stream identifier for WRITE STREAM commands. If this is set to 0 (which is not
2838 a valid stream identifier) fio will open a stream and then close it when done. Default
2841 .. option:: http_host=str : [http]
2843 Hostname to connect to. For S3, this could be the bucket hostname.
2844 Default is **localhost**
2846 .. option:: http_user=str : [http]
2848 Username for HTTP authentication.
2850 .. option:: http_pass=str : [http]
2852 Password for HTTP authentication.
2854 .. option:: https=str : [http]
2856 Enable HTTPS instead of http. *on* enables HTTPS; *insecure*
2857 will enable HTTPS, but disable SSL peer verification (use with
2858 caution!). Default is **off**
2860 .. option:: http_mode=str : [http]
2862 Which HTTP access mode to use: *webdav*, *swift*, or *s3*.
2863 Default is **webdav**
2865 .. option:: http_s3_region=str : [http]
2867 The S3 region/zone string.
2868 Default is **us-east-1**
2870 .. option:: http_s3_key=str : [http]
2874 .. option:: http_s3_keyid=str : [http]
2876 The S3 key/access id.
2878 .. option:: http_s3_sse_customer_key=str : [http]
2880 The encryption customer key in SSE server side.
2882 .. option:: http_s3_sse_customer_algorithm=str : [http]
2884 The encryption customer algorithm in SSE server side.
2885 Default is **AES256**
2887 .. option:: http_s3_storage_class=str : [http]
2889 Which storage class to access. User-customizable settings.
2890 Default is **STANDARD**
2892 .. option:: http_swift_auth_token=str : [http]
2894 The Swift auth token. See the example configuration file on how
2897 .. option:: http_verbose=int : [http]
2899 Enable verbose requests from libcurl. Useful for debugging. 1
2900 turns on verbose logging from libcurl, 2 additionally enables
2901 HTTP IO tracing. Default is **0**
2903 .. option:: uri=str : [nbd]
2905 Specify the NBD URI of the server to test. The string
2906 is a standard NBD URI
2907 (see https://github.com/NetworkBlockDevice/nbd/tree/master/doc).
2908 Example URIs: nbd://localhost:10809
2909 nbd+unix:///?socket=/tmp/socket
2910 nbds://tlshost/exportname
2912 .. option:: gpu_dev_ids=str : [libcufile]
2914 Specify the GPU IDs to use with CUDA. This is a colon-separated list of
2915 int. GPUs are assigned to workers roundrobin. Default is 0.
2917 .. option:: cuda_io=str : [libcufile]
2919 Specify the type of I/O to use with CUDA. Default is **cufile**.
2922 Use libcufile and nvidia-fs. This option performs I/O directly
2923 between a GPUDirect Storage filesystem and GPU buffers,
2924 avoiding use of a bounce buffer. If :option:`verify` is set,
2925 cudaMemcpy is used to copy verificaton data between RAM and GPU.
2926 Verification data is copied from RAM to GPU before a write
2927 and from GPU to RAM after a read. :option:`direct` must be 1.
2929 Use POSIX to perform I/O with a RAM buffer, and use cudaMemcpy
2930 to transfer data between RAM and the GPUs. Data is copied from
2931 GPU to RAM before a write and copied from RAM to GPU after a
2932 read. :option:`verify` does not affect use of cudaMemcpy.
2934 .. option:: nfs_url=str : [nfs]
2936 URL in libnfs format, eg nfs://<server|ipv4|ipv6>/path[?arg=val[&arg=val]*]
2937 Refer to the libnfs README for more details.
2939 .. option:: program=str : [exec]
2941 Specify the program to execute.
2943 .. option:: arguments=str : [exec]
2945 Specify arguments to pass to program.
2946 Some special variables can be expanded to pass fio's job details to the program.
2949 Replaced by the duration of the job in seconds.
2951 Replaced by the name of the job.
2953 .. option:: grace_time=int : [exec]
2955 Specify the time between the SIGTERM and SIGKILL signals. Default is 1 second.
2957 .. option:: std_redirect=bool : [exec]
2959 If set, stdout and stderr streams are redirected to files named from the job name. Default is true.
2961 .. option:: xnvme_async=str : [xnvme]
2963 Select the xnvme async command interface. This can take these values.
2966 This is default and use to emulate asynchronous I/O by using a
2967 single thread to create a queue pair on top of a synchronous
2968 I/O interface using the NVMe driver IOCTL.
2970 Emulate an asynchronous I/O interface with a pool of userspace
2971 threads on top of a synchronous I/O interface using the NVMe
2972 driver IOCTL. By default four threads are used.
2974 Linux native asynchronous I/O interface which supports both
2975 direct and buffered I/O.
2977 Fast Linux native asynchronous I/O interface for NVMe pass
2978 through commands. This only works with NVMe character device
2981 Use Linux aio for Asynchronous I/O.
2983 Use the posix asynchronous I/O interface to perform one or
2984 more I/O operations asynchronously.
2986 Use the user-space VFIO-based backend, implemented using
2987 libvfn instead of SPDK.
2989 Do not transfer any data; just pretend to. This is mainly used
2990 for introspective performance evaluation.
2992 .. option:: xnvme_sync=str : [xnvme]
2994 Select the xnvme synchronous command interface. This can take these values.
2997 This is default and uses Linux NVMe Driver ioctl() for
3000 This supports regular as well as vectored pread() and pwrite()
3003 This is the same as psync except that it also supports zone
3004 management commands using Linux block layer IOCTLs.
3006 .. option:: xnvme_admin=str : [xnvme]
3008 Select the xnvme admin command interface. This can take these values.
3011 This is default and uses linux NVMe Driver ioctl() for admin
3014 Use Linux Block Layer ioctl() and sysfs for admin commands.
3016 .. option:: xnvme_dev_nsid=int : [xnvme]
3018 xnvme namespace identifier for userspace NVMe driver, SPDK or vfio.
3020 .. option:: xnvme_dev_subnqn=str : [xnvme]
3022 Sets the subsystem NQN for fabrics. This is for xNVMe to utilize a
3023 fabrics target with multiple systems.
3025 .. option:: xnvme_mem=str : [xnvme]
3027 Select the xnvme memory backend. This can take these values.
3030 This is the default posix memory backend for linux NVMe driver.
3032 Use hugepages, instead of existing posix memory backend. The
3033 memory backend uses hugetlbfs. This require users to allocate
3034 hugepages, mount hugetlbfs and set an environment variable for
3037 Uses SPDK's memory allocator.
3039 Uses libvfn's memory allocator. This also specifies the use
3040 of libvfn backend instead of SPDK.
3042 .. option:: xnvme_iovec=int : [xnvme]
3044 If this option is set. xnvme will use vectored read/write commands.
3046 .. option:: libblkio_driver=str : [libblkio]
3048 The libblkio *driver* to use. Different drivers access devices through
3049 different underlying interfaces. Available drivers depend on the
3050 libblkio version in use and are listed at
3051 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3053 .. option:: libblkio_path=str : [libblkio]
3055 Sets the value of the driver-specific "path" property before connecting
3056 the libblkio instance, which identifies the target device or file on
3057 which to perform I/O. Its exact semantics are driver-dependent and not
3058 all drivers may support it; see
3059 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3061 .. option:: libblkio_pre_connect_props=str : [libblkio]
3063 A colon-separated list of additional libblkio properties to be set after
3064 creating but before connecting the libblkio instance. Each property must
3065 have the format ``<name>=<value>``. Colons can be escaped as ``\:``.
3066 These are set after the engine sets any other properties, so those can
3067 be overridden. Available properties depend on the libblkio version in use
3069 https://libblkio.gitlab.io/libblkio/blkio.html#properties
3071 .. option:: libblkio_num_entries=int : [libblkio]
3073 Sets the value of the driver-specific "num-entries" property before
3074 starting the libblkio instance. Its exact semantics are driver-dependent
3075 and not all drivers may support it; see
3076 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3078 .. option:: libblkio_queue_size=int : [libblkio]
3080 Sets the value of the driver-specific "queue-size" property before
3081 starting the libblkio instance. Its exact semantics are driver-dependent
3082 and not all drivers may support it; see
3083 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3085 .. option:: libblkio_pre_start_props=str : [libblkio]
3087 A colon-separated list of additional libblkio properties to be set after
3088 connecting but before starting the libblkio instance. Each property must
3089 have the format ``<name>=<value>``. Colons can be escaped as ``\:``.
3090 These are set after the engine sets any other properties, so those can
3091 be overridden. Available properties depend on the libblkio version in use
3093 https://libblkio.gitlab.io/libblkio/blkio.html#properties
3095 .. option:: libblkio_vectored : [libblkio]
3097 Submit vectored read and write requests.
3099 .. option:: libblkio_write_zeroes_on_trim : [libblkio]
3101 Submit trims as "write zeroes" requests instead of discard requests.
3103 .. option:: libblkio_wait_mode=str : [libblkio]
3105 How to wait for completions:
3108 Use a blocking call to ``blkioq_do_io()``.
3110 Use a blocking call to ``read()`` on the completion eventfd.
3112 Use a busy loop with a non-blocking call to ``blkioq_do_io()``.
3114 .. option:: libblkio_force_enable_completion_eventfd : [libblkio]
3116 Enable the queue's completion eventfd even when unused. This may impact
3117 performance. The default is to enable it only if
3118 :option:`libblkio_wait_mode=eventfd <libblkio_wait_mode>`.
3120 .. option:: no_completion_thread : [windowsaio]
3122 Avoid using a separate thread for completion polling.
3127 .. option:: iodepth=int
3129 Number of I/O units to keep in flight against the file. Note that
3130 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
3131 for small degrees when :option:`verify_async` is in use). Even async
3132 engines may impose OS restrictions causing the desired depth not to be
3133 achieved. This may happen on Linux when using libaio and not setting
3134 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
3135 eye on the I/O depth distribution in the fio output to verify that the
3136 achieved depth is as expected. Default: 1.
3138 .. option:: iodepth_batch_submit=int, iodepth_batch=int
3140 This defines how many pieces of I/O to submit at once. It defaults to 1
3141 which means that we submit each I/O as soon as it is available, but can be
3142 raised to submit bigger batches of I/O at the time. If it is set to 0 the
3143 :option:`iodepth` value will be used.
3145 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
3147 This defines how many pieces of I/O to retrieve at once. It defaults to 1
3148 which means that we'll ask for a minimum of 1 I/O in the retrieval process
3149 from the kernel. The I/O retrieval will go on until we hit the limit set by
3150 :option:`iodepth_low`. If this variable is set to 0, then fio will always
3151 check for completed events before queuing more I/O. This helps reduce I/O
3152 latency, at the cost of more retrieval system calls.
3154 .. option:: iodepth_batch_complete_max=int
3156 This defines maximum pieces of I/O to retrieve at once. This variable should
3157 be used along with :option:`iodepth_batch_complete_min`\=int variable,
3158 specifying the range of min and max amount of I/O which should be
3159 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
3164 iodepth_batch_complete_min=1
3165 iodepth_batch_complete_max=<iodepth>
3167 which means that we will retrieve at least 1 I/O and up to the whole
3168 submitted queue depth. If none of I/O has been completed yet, we will wait.
3172 iodepth_batch_complete_min=0
3173 iodepth_batch_complete_max=<iodepth>
3175 which means that we can retrieve up to the whole submitted queue depth, but
3176 if none of I/O has been completed yet, we will NOT wait and immediately exit
3177 the system call. In this example we simply do polling.
3179 .. option:: iodepth_low=int
3181 The low water mark indicating when to start filling the queue
3182 again. Defaults to the same as :option:`iodepth`, meaning that fio will
3183 attempt to keep the queue full at all times. If :option:`iodepth` is set to
3184 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
3185 16 requests, it will let the depth drain down to 4 before starting to fill
3188 .. option:: serialize_overlap=bool
3190 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
3191 When two or more I/Os are submitted simultaneously, there is no guarantee that
3192 the I/Os will be processed or completed in the submitted order. Further, if
3193 two or more of those I/Os are writes, any overlapping region between them can
3194 become indeterminate/undefined on certain storage. These issues can cause
3195 verification to fail erratically when at least one of the racing I/Os is
3196 changing data and the overlapping region has a non-zero size. Setting
3197 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
3198 serializing in-flight I/Os that have a non-zero overlap. Note that setting
3199 this option can reduce both performance and the :option:`iodepth` achieved.
3201 This option only applies to I/Os issued for a single job except when it is
3202 enabled along with :option:`io_submit_mode`\=offload. In offload mode, fio
3203 will check for overlap among all I/Os submitted by offload jobs with :option:`serialize_overlap`
3208 .. option:: io_submit_mode=str
3210 This option controls how fio submits the I/O to the I/O engine. The default
3211 is `inline`, which means that the fio job threads submit and reap I/O
3212 directly. If set to `offload`, the job threads will offload I/O submission
3213 to a dedicated pool of I/O threads. This requires some coordination and thus
3214 has a bit of extra overhead, especially for lower queue depth I/O where it
3215 can increase latencies. The benefit is that fio can manage submission rates
3216 independently of the device completion rates. This avoids skewed latency
3217 reporting if I/O gets backed up on the device side (the coordinated omission
3218 problem). Note that this option cannot reliably be used with async IO
3225 .. option:: thinkcycles=int
3227 Stall the job for the specified number of cycles after an I/O has completed before
3228 issuing the next. May be used to simulate processing being done by an application.
3229 This is not taken into account for the time to be waited on for :option:`thinktime`.
3230 Might not have any effect on some platforms, this can be checked by trying a setting
3231 a high enough amount of thinkcycles.
3233 .. option:: thinktime=time
3235 Stall the job for the specified period of time after an I/O has completed before issuing the
3236 next. May be used to simulate processing being done by an application.
3237 When the unit is omitted, the value is interpreted in microseconds. See
3238 :option:`thinktime_blocks`, :option:`thinktime_iotime` and :option:`thinktime_spin`.
3240 .. option:: thinktime_spin=time
3242 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
3243 something with the data received, before falling back to sleeping for the
3244 rest of the period specified by :option:`thinktime`. When the unit is
3245 omitted, the value is interpreted in microseconds.
3247 .. option:: thinktime_blocks=int
3249 Only valid if :option:`thinktime` is set - control how many blocks to issue,
3250 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
3251 fio wait :option:`thinktime` usecs after every block. This effectively makes any
3252 queue depth setting redundant, since no more than 1 I/O will be queued
3253 before we have to complete it and do our :option:`thinktime`. In other words, this
3254 setting effectively caps the queue depth if the latter is larger.
3256 .. option:: thinktime_blocks_type=str
3258 Only valid if :option:`thinktime` is set - control how :option:`thinktime_blocks`
3259 triggers. The default is `complete`, which triggers thinktime when fio completes
3260 :option:`thinktime_blocks` blocks. If this is set to `issue`, then the trigger happens
3263 .. option:: thinktime_iotime=time
3265 Only valid if :option:`thinktime` is set - control :option:`thinktime`
3266 interval by time. The :option:`thinktime` stall is repeated after IOs
3267 are executed for :option:`thinktime_iotime`. For example,
3268 ``--thinktime_iotime=9s --thinktime=1s`` repeat 10-second cycle with IOs
3269 for 9 seconds and stall for 1 second. When the unit is omitted,
3270 :option:`thinktime_iotime` is interpreted as a number of seconds. If
3271 this option is used together with :option:`thinktime_blocks`, the
3272 :option:`thinktime` stall is repeated after :option:`thinktime_iotime`
3273 or after :option:`thinktime_blocks` IOs, whichever happens first.
3275 .. option:: rate=int[,int][,int]
3277 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
3278 suffix rules apply. Comma-separated values may be specified for reads,
3279 writes, and trims as described in :option:`blocksize`.
3281 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
3282 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
3283 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
3284 latter will only limit reads.
3286 .. option:: rate_min=int[,int][,int]
3288 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
3289 to meet this requirement will cause the job to exit. Comma-separated values
3290 may be specified for reads, writes, and trims as described in
3291 :option:`blocksize`.
3293 .. option:: rate_iops=int[,int][,int]
3295 Cap the bandwidth to this number of IOPS. Basically the same as
3296 :option:`rate`, just specified independently of bandwidth. If the job is
3297 given a block size range instead of a fixed value, the smallest block size
3298 is used as the metric. Comma-separated values may be specified for reads,
3299 writes, and trims as described in :option:`blocksize`.
3301 .. option:: rate_iops_min=int[,int][,int]
3303 If fio doesn't meet this rate of I/O, it will cause the job to exit.
3304 Comma-separated values may be specified for reads, writes, and trims as
3305 described in :option:`blocksize`.
3307 .. option:: rate_process=str
3309 This option controls how fio manages rated I/O submissions. The default is
3310 `linear`, which submits I/O in a linear fashion with fixed delays between
3311 I/Os that gets adjusted based on I/O completion rates. If this is set to
3312 `poisson`, fio will submit I/O based on a more real world random request
3313 flow, known as the Poisson process
3314 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
3315 10^6 / IOPS for the given workload.
3317 .. option:: rate_ignore_thinktime=bool
3319 By default, fio will attempt to catch up to the specified rate setting,
3320 if any kind of thinktime setting was used. If this option is set, then
3321 fio will ignore the thinktime and continue doing IO at the specified
3322 rate, instead of entering a catch-up mode after thinktime is done.
3324 .. option:: rate_cycle=int
3326 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
3327 of milliseconds. Defaults to 1000.
3333 .. option:: latency_target=time
3335 If set, fio will attempt to find the max performance point that the given
3336 workload will run at while maintaining a latency below this target. When
3337 the unit is omitted, the value is interpreted in microseconds. See
3338 :option:`latency_window` and :option:`latency_percentile`.
3340 .. option:: latency_window=time
3342 Used with :option:`latency_target` to specify the sample window that the job
3343 is run at varying queue depths to test the performance. When the unit is
3344 omitted, the value is interpreted in microseconds.
3346 .. option:: latency_percentile=float
3348 The percentage of I/Os that must fall within the criteria specified by
3349 :option:`latency_target` and :option:`latency_window`. If not set, this
3350 defaults to 100.0, meaning that all I/Os must be equal or below to the value
3351 set by :option:`latency_target`.
3353 .. option:: latency_run=bool
3355 Used with :option:`latency_target`. If false (default), fio will find
3356 the highest queue depth that meets :option:`latency_target` and exit. If
3357 true, fio will continue running and try to meet :option:`latency_target`
3358 by adjusting queue depth.
3360 .. option:: max_latency=time[,time][,time]
3362 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
3363 maximum latency. When the unit is omitted, the value is interpreted in
3364 microseconds. Comma-separated values may be specified for reads, writes,
3365 and trims as described in :option:`blocksize`.
3371 .. option:: write_iolog=str
3373 Write the issued I/O patterns to the specified file. See
3374 :option:`read_iolog`. Specify a separate file for each job, otherwise the
3375 iologs will be interspersed and the file may be corrupt. This file will
3376 be opened in append mode.
3378 .. option:: read_iolog=str
3380 Open an iolog with the specified filename and replay the I/O patterns it
3381 contains. This can be used to store a workload and replay it sometime
3382 later. The iolog given may also be a blktrace binary file, which allows fio
3383 to replay a workload captured by :command:`blktrace`. See
3384 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
3385 replay, the file needs to be turned into a blkparse binary data file first
3386 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
3387 You can specify a number of files by separating the names with a ':'
3388 character. See the :option:`filename` option for information on how to
3389 escape ':' characters within the file names. These files will
3390 be sequentially assigned to job clones created by :option:`numjobs`.
3391 '-' is a reserved name, meaning read from stdin, notably if
3392 :option:`filename` is set to '-' which means stdin as well, then
3393 this flag can't be set to '-'.
3395 .. option:: read_iolog_chunked=bool
3397 Determines how iolog is read. If false(default) entire :option:`read_iolog`
3398 will be read at once. If selected true, input from iolog will be read
3399 gradually. Useful when iolog is very large, or it is generated.
3401 .. option:: merge_blktrace_file=str
3403 When specified, rather than replaying the logs passed to :option:`read_iolog`,
3404 the logs go through a merge phase which aggregates them into a single
3405 blktrace. The resulting file is then passed on as the :option:`read_iolog`
3406 parameter. The intention here is to make the order of events consistent.
3407 This limits the influence of the scheduler compared to replaying multiple
3408 blktraces via concurrent jobs.
3410 .. option:: merge_blktrace_scalars=float_list
3412 This is a percentage based option that is index paired with the list of
3413 files passed to :option:`read_iolog`. When merging is performed, scale
3414 the time of each event by the corresponding amount. For example,
3415 ``--merge_blktrace_scalars="50:100"`` runs the first trace in halftime
3416 and the second trace in realtime. This knob is separately tunable from
3417 :option:`replay_time_scale` which scales the trace during runtime and
3418 does not change the output of the merge unlike this option.
3420 .. option:: merge_blktrace_iters=float_list
3422 This is a whole number option that is index paired with the list of files
3423 passed to :option:`read_iolog`. When merging is performed, run each trace
3424 for the specified number of iterations. For example,
3425 ``--merge_blktrace_iters="2:1"`` runs the first trace for two iterations
3426 and the second trace for one iteration.
3428 .. option:: replay_no_stall=bool
3430 When replaying I/O with :option:`read_iolog` the default behavior is to
3431 attempt to respect the timestamps within the log and replay them with the
3432 appropriate delay between IOPS. By setting this variable fio will not
3433 respect the timestamps and attempt to replay them as fast as possible while
3434 still respecting ordering. The result is the same I/O pattern to a given
3435 device, but different timings.
3437 .. option:: replay_time_scale=int
3439 When replaying I/O with :option:`read_iolog`, fio will honor the
3440 original timing in the trace. With this option, it's possible to scale
3441 the time. It's a percentage option, if set to 50 it means run at 50%
3442 the original IO rate in the trace. If set to 200, run at twice the
3443 original IO rate. Defaults to 100.
3445 .. option:: replay_redirect=str
3447 While replaying I/O patterns using :option:`read_iolog` the default behavior
3448 is to replay the IOPS onto the major/minor device that each IOP was recorded
3449 from. This is sometimes undesirable because on a different machine those
3450 major/minor numbers can map to a different device. Changing hardware on the
3451 same system can also result in a different major/minor mapping.
3452 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
3453 device regardless of the device it was recorded
3454 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
3455 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
3456 multiple devices will be replayed onto a single device, if the trace
3457 contains multiple devices. If you want multiple devices to be replayed
3458 concurrently to multiple redirected devices you must blkparse your trace
3459 into separate traces and replay them with independent fio invocations.
3460 Unfortunately this also breaks the strict time ordering between multiple
3463 .. option:: replay_align=int
3465 Force alignment of the byte offsets in a trace to this value. The value
3466 must be a power of 2.
3468 .. option:: replay_scale=int
3470 Scale byte offsets down by this factor when replaying traces. Should most
3471 likely use :option:`replay_align` as well.
3473 .. option:: replay_skip=str
3475 Sometimes it's useful to skip certain IO types in a replay trace.
3476 This could be, for instance, eliminating the writes in the trace.
3477 Or not replaying the trims/discards, if you are redirecting to
3478 a device that doesn't support them. This option takes a comma
3479 separated list of read, write, trim, sync.
3482 Threads, processes and job synchronization
3483 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3487 Fio defaults to creating jobs by using fork, however if this option is
3488 given, fio will create jobs by using POSIX Threads' function
3489 :manpage:`pthread_create(3)` to create threads instead.
3491 .. option:: wait_for=str
3493 If set, the current job won't be started until all workers of the specified
3494 waitee job are done.
3496 ``wait_for`` operates on the job name basis, so there are a few
3497 limitations. First, the waitee must be defined prior to the waiter job
3498 (meaning no forward references). Second, if a job is being referenced as a
3499 waitee, it must have a unique name (no duplicate waitees).
3501 .. option:: nice=int
3503 Run the job with the given nice value. See man :manpage:`nice(2)`.
3505 On Windows, values less than -15 set the process class to "High"; -1 through
3506 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
3509 .. option:: prio=int
3511 Set the I/O priority value of this job. Linux limits us to a positive value
3512 between 0 and 7, with 0 being the highest. See man
3513 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
3514 systems since meaning of priority may differ. For per-command priority
3515 setting, see I/O engine specific :option:`cmdprio_percentage` and
3516 :option:`cmdprio` options.
3518 .. option:: prioclass=int
3520 Set the I/O priority class. See man :manpage:`ionice(1)`. For per-command
3521 priority setting, see I/O engine specific :option:`cmdprio_percentage`
3522 and :option:`cmdprio_class` options.
3524 .. option:: priohint=int
3526 Set the I/O priority hint. This is only applicable to platforms that
3527 support I/O priority classes and to devices with features controlled
3528 through priority hints, e.g. block devices supporting command duration
3529 limits, or CDL. CDL is a way to indicate the desired maximum latency
3530 of I/Os so that the device can optimize its internal command scheduling
3531 according to the latency limits indicated by the user.
3533 For per-I/O priority hint setting, see the I/O engine specific
3534 :option:`cmdprio_hint` option.
3536 .. option:: cpus_allowed=str
3538 Controls the same options as :option:`cpumask`, but accepts a textual
3539 specification of the permitted CPUs instead and CPUs are indexed from 0. So
3540 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
3541 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
3542 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
3544 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
3545 processor group will be used and affinity settings are inherited from the
3546 system. An fio build configured to target Windows 7 makes options that set
3547 CPUs processor group aware and values will set both the processor group
3548 and a CPU from within that group. For example, on a system where processor
3549 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
3550 values between 0 and 39 will bind CPUs from processor group 0 and
3551 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
3552 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
3553 single ``cpus_allowed`` option must be from the same processor group. For
3554 Windows fio builds not built for Windows 7, CPUs will only be selected from
3555 (and be relative to) whatever processor group fio happens to be running in
3556 and CPUs from other processor groups cannot be used.
3558 .. option:: cpus_allowed_policy=str
3560 Set the policy of how fio distributes the CPUs specified by
3561 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
3564 All jobs will share the CPU set specified.
3566 Each job will get a unique CPU from the CPU set.
3568 **shared** is the default behavior, if the option isn't specified. If
3569 **split** is specified, then fio will assign one cpu per job. If not
3570 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
3573 .. option:: cpumask=int
3575 Set the CPU affinity of this job. The parameter given is a bit mask of
3576 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
3577 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
3578 :manpage:`sched_setaffinity(2)`. This may not work on all supported
3579 operating systems or kernel versions. This option doesn't work well for a
3580 higher CPU count than what you can store in an integer mask, so it can only
3581 control cpus 1-32. For boxes with larger CPU counts, use
3582 :option:`cpus_allowed`.
3584 .. option:: numa_cpu_nodes=str
3586 Set this job running on specified NUMA nodes' CPUs. The arguments allow
3587 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
3588 NUMA options support, fio must be built on a system with libnuma-dev(el)
3591 .. option:: numa_mem_policy=str
3593 Set this job's memory policy and corresponding NUMA nodes. Format of the
3598 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
3599 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
3600 policies, no node needs to be specified. For ``prefer``, only one node is
3601 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
3602 follows: a comma delimited list of numbers, A-B ranges, or `all`.
3604 .. option:: cgroup=str
3606 Add job to this control group. If it doesn't exist, it will be created. The
3607 system must have a mounted cgroup blkio mount point for this to work. If
3608 your system doesn't have it mounted, you can do so with::
3610 # mount -t cgroup -o blkio none /cgroup
3612 .. option:: cgroup_weight=int
3614 Set the weight of the cgroup to this value. See the documentation that comes
3615 with the kernel, allowed values are in the range of 100..1000.
3617 .. option:: cgroup_nodelete=bool
3619 Normally fio will delete the cgroups it has created after the job
3620 completion. To override this behavior and to leave cgroups around after the
3621 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
3622 to inspect various cgroup files after job completion. Default: false.
3624 .. option:: flow_id=int
3626 The ID of the flow. If not specified, it defaults to being a global
3627 flow. See :option:`flow`.
3629 .. option:: flow=int
3631 Weight in token-based flow control. If this value is used, then fio
3632 regulates the activity between two or more jobs sharing the same
3633 flow_id. Fio attempts to keep each job activity proportional to other
3634 jobs' activities in the same flow_id group, with respect to requested
3635 weight per job. That is, if one job has `flow=3', another job has
3636 `flow=2' and another with `flow=1`, then there will be a roughly 3:2:1
3637 ratio in how much one runs vs the others.
3639 .. option:: flow_sleep=int
3641 The period of time, in microseconds, to wait after the flow counter
3642 has exceeded its proportion before retrying operations.
3644 .. option:: stonewall, wait_for_previous
3646 Wait for preceding jobs in the job file to exit, before starting this
3647 one. Can be used to insert serialization points in the job file. A stone
3648 wall also implies starting a new reporting group, see
3649 :option:`group_reporting`.
3653 By default, fio will continue running all other jobs when one job finishes.
3654 Sometimes this is not the desired action. Setting ``exitall`` will instead
3655 make fio terminate all jobs in the same group, as soon as one job of that
3658 .. option:: exit_what=str
3660 By default, fio will continue running all other jobs when one job finishes.
3661 Sometimes this is not the desired action. Setting ``exitall`` will
3662 instead make fio terminate all jobs in the same group. The option
3663 ``exit_what`` allows one to control which jobs get terminated when ``exitall``
3664 is enabled. The default is ``group`` and does not change the behaviour of
3665 ``exitall``. The setting ``all`` terminates all jobs. The setting ``stonewall``
3666 terminates all currently running jobs across all groups and continues execution
3667 with the next stonewalled group.
3669 .. option:: exec_prerun=str
3671 Before running this job, issue the command specified through
3672 :manpage:`system(3)`. Output is redirected in a file called
3673 :file:`jobname.prerun.txt`.
3675 .. option:: exec_postrun=str
3677 After the job completes, issue the command specified though
3678 :manpage:`system(3)`. Output is redirected in a file called
3679 :file:`jobname.postrun.txt`.
3683 Instead of running as the invoking user, set the user ID to this value
3684 before the thread/process does any work.
3688 Set group ID, see :option:`uid`.
3694 .. option:: verify_only
3696 Do not perform specified workload, only verify data still matches previous
3697 invocation of this workload. This option allows one to check data multiple
3698 times at a later date without overwriting it. This option makes sense only
3699 for workloads that write data, and does not support workloads with the
3700 :option:`time_based` option set.
3702 .. option:: do_verify=bool
3704 Run the verify phase after a write phase. Only valid if :option:`verify` is
3707 .. option:: verify=str
3709 If writing to a file, fio can verify the file contents after each iteration
3710 of the job. Each verification method also implies verification of special
3711 header, which is written to the beginning of each block. This header also
3712 includes meta information, like offset of the block, block number, timestamp
3713 when block was written, etc. :option:`verify` can be combined with
3714 :option:`verify_pattern` option. The allowed values are:
3717 Use an md5 sum of the data area and store it in the header of
3721 Use an experimental crc64 sum of the data area and store it in the
3722 header of each block.
3725 Use a crc32c sum of the data area and store it in the header of
3726 each block. This will automatically use hardware acceleration
3727 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
3728 fall back to software crc32c if none is found. Generally the
3729 fastest checksum fio supports when hardware accelerated.
3735 Use a crc32 sum of the data area and store it in the header of each
3739 Use a crc16 sum of the data area and store it in the header of each
3743 Use a crc7 sum of the data area and store it in the header of each
3747 Use xxhash as the checksum function. Generally the fastest software
3748 checksum that fio supports.
3751 Use sha512 as the checksum function.
3754 Use sha256 as the checksum function.
3757 Use optimized sha1 as the checksum function.
3760 Use optimized sha3-224 as the checksum function.
3763 Use optimized sha3-256 as the checksum function.
3766 Use optimized sha3-384 as the checksum function.
3769 Use optimized sha3-512 as the checksum function.
3772 This option is deprecated, since now meta information is included in
3773 generic verification header and meta verification happens by
3774 default. For detailed information see the description of the
3775 :option:`verify` setting. This option is kept because of
3776 compatibility's sake with old configurations. Do not use it.
3779 Verify a strict pattern. Normally fio includes a header with some
3780 basic information and checksumming, but if this option is set, only
3781 the specific pattern set with :option:`verify_pattern` is verified.
3784 Only pretend to verify. Useful for testing internals with
3785 :option:`ioengine`\=null, not for much else.
3787 This option can be used for repeated burn-in tests of a system to make sure
3788 that the written data is also correctly read back. If the data direction
3789 given is a read or random read, fio will assume that it should verify a
3790 previously written file. If the data direction includes any form of write,
3791 the verify will be of the newly written data.
3793 To avoid false verification errors, do not use the norandommap option when
3794 verifying data with async I/O engines and I/O depths > 1. Or use the
3795 norandommap and the lfsr random generator together to avoid writing to the
3796 same offset with multiple outstanding I/Os.
3798 .. option:: verify_offset=int
3800 Swap the verification header with data somewhere else in the block before
3801 writing. It is swapped back before verifying.
3803 .. option:: verify_interval=int
3805 Write the verification header at a finer granularity than the
3806 :option:`blocksize`. It will be written for chunks the size of
3807 ``verify_interval``. :option:`blocksize` should divide this evenly.
3809 .. option:: verify_pattern=str
3811 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
3812 filling with totally random bytes, but sometimes it's interesting to fill
3813 with a known pattern for I/O verification purposes. Depending on the width
3814 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
3815 be either a decimal or a hex number). The ``verify_pattern`` if larger than
3816 a 32-bit quantity has to be a hex number that starts with either "0x" or
3817 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
3818 format, which means that for each block offset will be written and then
3819 verified back, e.g.::
3823 Or use combination of everything::
3825 verify_pattern=0xff%o"abcd"-12
3827 .. option:: verify_fatal=bool
3829 Normally fio will keep checking the entire contents before quitting on a
3830 block verification failure. If this option is set, fio will exit the job on
3831 the first observed failure. Default: false.
3833 .. option:: verify_dump=bool
3835 If set, dump the contents of both the original data block and the data block
3836 we read off disk to files. This allows later analysis to inspect just what
3837 kind of data corruption occurred. Off by default.
3839 .. option:: verify_async=int
3841 Fio will normally verify I/O inline from the submitting thread. This option
3842 takes an integer describing how many async offload threads to create for I/O
3843 verification instead, causing fio to offload the duty of verifying I/O
3844 contents to one or more separate threads. If using this offload option, even
3845 sync I/O engines can benefit from using an :option:`iodepth` setting higher
3846 than 1, as it allows them to have I/O in flight while verifies are running.
3847 Defaults to 0 async threads, i.e. verification is not asynchronous.
3849 .. option:: verify_async_cpus=str
3851 Tell fio to set the given CPU affinity on the async I/O verification
3852 threads. See :option:`cpus_allowed` for the format used.
3854 .. option:: verify_backlog=int
3856 Fio will normally verify the written contents of a job that utilizes verify
3857 once that job has completed. In other words, everything is written then
3858 everything is read back and verified. You may want to verify continually
3859 instead for a variety of reasons. Fio stores the meta data associated with
3860 an I/O block in memory, so for large verify workloads, quite a bit of memory
3861 would be used up holding this meta data. If this option is enabled, fio will
3862 write only N blocks before verifying these blocks.
3864 .. option:: verify_backlog_batch=int
3866 Control how many blocks fio will verify if :option:`verify_backlog` is
3867 set. If not set, will default to the value of :option:`verify_backlog`
3868 (meaning the entire queue is read back and verified). If
3869 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
3870 blocks will be verified, if ``verify_backlog_batch`` is larger than
3871 :option:`verify_backlog`, some blocks will be verified more than once.
3873 .. option:: verify_state_save=bool
3875 When a job exits during the write phase of a verify workload, save its
3876 current state. This allows fio to replay up until that point, if the verify
3877 state is loaded for the verify read phase. The format of the filename is,
3880 <type>-<jobname>-<jobindex>-verify.state.
3882 <type> is "local" for a local run, "sock" for a client/server socket
3883 connection, and "ip" (192.168.0.1, for instance) for a networked
3884 client/server connection. Defaults to true.
3886 .. option:: verify_state_load=bool
3888 If a verify termination trigger was used, fio stores the current write state
3889 of each thread. This can be used at verification time so that fio knows how
3890 far it should verify. Without this information, fio will run a full
3891 verification pass, according to the settings in the job file used. Default
3894 .. option:: experimental_verify=bool
3896 Enable experimental verification. Standard verify records I/O metadata
3897 for later use during the verification phase. Experimental verify
3898 instead resets the file after the write phase and then replays I/Os for
3899 the verification phase.
3901 .. option:: trim_percentage=int
3903 Number of verify blocks to discard/trim.
3905 .. option:: trim_verify_zero=bool
3907 Verify that trim/discarded blocks are returned as zeros.
3909 .. option:: trim_backlog=int
3911 Trim after this number of blocks are written.
3913 .. option:: trim_backlog_batch=int
3915 Trim this number of I/O blocks.
3920 .. option:: steadystate=str:float, ss=str:float
3922 Define the criterion and limit for assessing steady state performance. The
3923 first parameter designates the criterion whereas the second parameter sets
3924 the threshold. When the criterion falls below the threshold for the
3925 specified duration, the job will stop. For example, `iops_slope:0.1%` will
3926 direct fio to terminate the job when the least squares regression slope
3927 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
3928 this will apply to all jobs in the group. Below is the list of available
3929 steady state assessment criteria. All assessments are carried out using only
3930 data from the rolling collection window. Threshold limits can be expressed
3931 as a fixed value or as a percentage of the mean in the collection window.
3933 When using this feature, most jobs should include the :option:`time_based`
3934 and :option:`runtime` options or the :option:`loops` option so that fio does not
3935 stop running after it has covered the full size of the specified file(s) or device(s).
3938 Collect IOPS data. Stop the job if all individual IOPS measurements
3939 are within the specified limit of the mean IOPS (e.g., ``iops:2``
3940 means that all individual IOPS values must be within 2 of the mean,
3941 whereas ``iops:0.2%`` means that all individual IOPS values must be
3942 within 0.2% of the mean IOPS to terminate the job).
3945 Collect IOPS data and calculate the least squares regression
3946 slope. Stop the job if the slope falls below the specified limit.
3949 Collect bandwidth data. Stop the job if all individual bandwidth
3950 measurements are within the specified limit of the mean bandwidth.
3953 Collect bandwidth data and calculate the least squares regression
3954 slope. Stop the job if the slope falls below the specified limit.
3956 .. option:: steadystate_duration=time, ss_dur=time
3958 A rolling window of this duration will be used to judge whether steady
3959 state has been reached. Data will be collected every
3960 :option:`ss_interval`. The default is 0 which disables steady state
3961 detection. When the unit is omitted, the value is interpreted in
3964 .. option:: steadystate_ramp_time=time, ss_ramp=time
3966 Allow the job to run for the specified duration before beginning data
3967 collection for checking the steady state job termination criterion. The
3968 default is 0. When the unit is omitted, the value is interpreted in seconds.
3970 .. option:: steadystate_check_interval=time, ss_interval=time
3972 The values during the rolling window will be collected with a period of
3973 this value. If :option:`ss_interval` is 30s and :option:`ss_dur` is
3974 300s, 10 measurements will be taken. Default is 1s but that might not
3975 converge, especially for slower devices, so set this accordingly. When
3976 the unit is omitted, the value is interpreted in seconds.
3979 Measurements and reporting
3980 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3982 .. option:: per_job_logs=bool
3984 If set to true, fio generates bw/clat/iops logs with per job unique
3985 filenames. If set to false, jobs with identical names will share a log
3986 filename. Note that when this option is set to false log files will be
3987 opened in append mode and if log files already exist the previous
3988 contents will not be overwritten. Default: true.
3990 .. option:: group_reporting
3992 It may sometimes be interesting to display statistics for groups of jobs as
3993 a whole instead of for each individual job. This is especially true if
3994 :option:`numjobs` is used; looking at individual thread/process output
3995 quickly becomes unwieldy. To see the final report per-group instead of
3996 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
3997 same reporting group, unless if separated by a :option:`stonewall`, or by
3998 using :option:`new_group`.
4000 NOTE: When :option:`group_reporting` is used along with `json` output,
4001 there are certain per-job properties which can be different between jobs
4002 but do not have a natural group-level equivalent. Examples include
4003 `kb_base`, `unit_base`, `sig_figs`, `thread_number`, `pid`, and
4004 `job_start`. For these properties, the values for the first job are
4005 recorded for the group.
4007 .. option:: new_group
4009 Start a new reporting group. See: :option:`group_reporting`. If not given,
4010 all jobs in a file will be part of the same reporting group, unless
4011 separated by a :option:`stonewall`.
4013 .. option:: stats=bool
4015 By default, fio collects and shows final output results for all jobs
4016 that run. If this option is set to 0, then fio will ignore it in
4017 the final stat output.
4019 .. option:: write_bw_log=str
4021 If given, write a bandwidth log for this job. Can be used to store data of
4022 the bandwidth of the jobs in their lifetime.
4024 If no str argument is given, the default filename of
4025 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
4026 will still append the type of log. So if one specifies::
4030 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
4031 of the job (`1..N`, where `N` is the number of jobs). If
4032 :option:`per_job_logs` is false, then the filename will not include the
4035 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
4036 text files into nice graphs. See `Log File Formats`_ for how data is
4037 structured within the file.
4039 .. option:: write_lat_log=str
4041 Same as :option:`write_bw_log`, except this option creates I/O
4042 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
4043 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
4044 latency files instead. See :option:`write_bw_log` for details about
4045 the filename format and `Log File Formats`_ for how data is structured
4048 .. option:: write_hist_log=str
4050 Same as :option:`write_bw_log` but writes an I/O completion latency
4051 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
4052 file will be empty unless :option:`log_hist_msec` has also been set.
4053 See :option:`write_bw_log` for details about the filename format and
4054 `Log File Formats`_ for how data is structured within the file.
4056 .. option:: write_iops_log=str
4058 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
4059 :file:`name_iops.x.log`) instead. Because fio defaults to individual
4060 I/O logging, the value entry in the IOPS log will be 1 unless windowed
4061 logging (see :option:`log_avg_msec`) has been enabled. See
4062 :option:`write_bw_log` for details about the filename format and `Log
4063 File Formats`_ for how data is structured within the file.
4065 .. option:: log_entries=int
4067 By default, fio will log an entry in the iops, latency, or bw log for
4068 every I/O that completes. The initial number of I/O log entries is 1024.
4069 When the log entries are all used, new log entries are dynamically
4070 allocated. This dynamic log entry allocation may negatively impact
4071 time-related statistics such as I/O tail latencies (e.g. 99.9th percentile
4072 completion latency). This option allows specifying a larger initial
4073 number of log entries to avoid run-time allocations of new log entries,
4074 resulting in more precise time-related I/O statistics.
4075 Also see :option:`log_avg_msec`. Defaults to 1024.
4077 .. option:: log_avg_msec=int
4079 By default, fio will log an entry in the iops, latency, or bw log for
4080 every I/O that completes. When writing to the disk log, that can
4081 quickly grow to a very large size. Setting this option directs fio to
4082 instead record an average over the specified duration for each log
4083 entry, reducing the resolution of the log. When the job completes, fio
4084 will flush any accumulated latency log data, so the final log interval
4085 may not match the value specified by this option and there may even be
4086 duplicate timestamps. See :option:`log_window_value` as well. Defaults
4087 to 0, logging entries for each I/O. Also see `Log File Formats`_.
4089 .. option:: log_hist_msec=int
4091 Same as :option:`log_avg_msec`, but logs entries for completion latency
4092 histograms. Computing latency percentiles from averages of intervals using
4093 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
4094 histogram entries over the specified period of time, reducing log sizes for
4095 high IOPS devices while retaining percentile accuracy. See
4096 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
4097 Defaults to 0, meaning histogram logging is disabled.
4099 .. option:: log_hist_coarseness=int
4101 Integer ranging from 0 to 6, defining the coarseness of the resolution of
4102 the histogram logs enabled with :option:`log_hist_msec`. For each increment
4103 in coarseness, fio outputs half as many bins. Defaults to 0, for which
4104 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
4105 and `Log File Formats`_.
4107 .. option:: log_window_value=str, log_max_value=str
4109 If :option:`log_avg_msec` is set, fio by default logs the average over that
4110 window. This option determines whether fio logs the average, maximum or
4111 both the values over the window. This only affects the latency logging,
4112 as both average and maximum values for iops or bw log will be same.
4113 Accepted values are:
4116 Log average value over the window. The default.
4119 Log maximum value in the window.
4122 Log both average and maximum value over the window.
4125 Backward-compatible alias for **avg**.
4128 Backward-compatible alias for **max**.
4130 .. option:: log_offset=bool
4132 If this is set, the iolog options will include the byte offset for the I/O
4133 entry as well as the other data values. Defaults to 0 meaning that
4134 offsets are not present in logs. Also see `Log File Formats`_.
4136 .. option:: log_compression=int
4138 If this is set, fio will compress the I/O logs as it goes, to keep the
4139 memory footprint lower. When a log reaches the specified size, that chunk is
4140 removed and compressed in the background. Given that I/O logs are fairly
4141 highly compressible, this yields a nice memory savings for longer runs. The
4142 downside is that the compression will consume some background CPU cycles, so
4143 it may impact the run. This, however, is also true if the logging ends up
4144 consuming most of the system memory. So pick your poison. The I/O logs are
4145 saved normally at the end of a run, by decompressing the chunks and storing
4146 them in the specified log file. This feature depends on the availability of
4149 .. option:: log_compression_cpus=str
4151 Define the set of CPUs that are allowed to handle online log compression for
4152 the I/O jobs. This can provide better isolation between performance
4153 sensitive jobs, and background compression work. See
4154 :option:`cpus_allowed` for the format used.
4156 .. option:: log_store_compressed=bool
4158 If set, fio will store the log files in a compressed format. They can be
4159 decompressed with fio, using the :option:`--inflate-log` command line
4160 parameter. The files will be stored with a :file:`.fz` suffix.
4162 .. option:: log_unix_epoch=bool
4164 Backwards compatible alias for log_alternate_epoch.
4166 .. option:: log_alternate_epoch=bool
4168 If set, fio will log timestamps based on the epoch used by the clock specified
4169 in the log_alternate_epoch_clock_id option, to the log files produced by
4170 enabling write_type_log for each log type, instead of the default zero-based
4173 .. option:: log_alternate_epoch_clock_id=int
4175 Specifies the clock_id to be used by clock_gettime to obtain the alternate
4176 epoch if log_alternate_epoch is true. Otherwise has no effect. Default
4177 value is 0, or CLOCK_REALTIME.
4179 .. option:: block_error_percentiles=bool
4181 If set, record errors in trim block-sized units from writes and trims and
4182 output a histogram of how many trims it took to get to errors, and what kind
4183 of error was encountered.
4185 .. option:: bwavgtime=int
4187 Average the calculated bandwidth over the given time. Value is specified in
4188 milliseconds. If the job also does bandwidth logging through
4189 :option:`write_bw_log`, then the minimum of this option and
4190 :option:`log_avg_msec` will be used. Default: 500ms.
4192 .. option:: iopsavgtime=int
4194 Average the calculated IOPS over the given time. Value is specified in
4195 milliseconds. If the job also does IOPS logging through
4196 :option:`write_iops_log`, then the minimum of this option and
4197 :option:`log_avg_msec` will be used. Default: 500ms.
4199 .. option:: disk_util=bool
4201 Generate disk utilization statistics, if the platform supports it.
4204 .. option:: disable_lat=bool
4206 Disable measurements of total latency numbers. Useful only for cutting back
4207 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
4208 performance at really high IOPS rates. Note that to really get rid of a
4209 large amount of these calls, this option must be used with
4210 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
4212 .. option:: disable_clat=bool
4214 Disable measurements of completion latency numbers. See
4215 :option:`disable_lat`.
4217 .. option:: disable_slat=bool
4219 Disable measurements of submission latency numbers. See
4220 :option:`disable_lat`.
4222 .. option:: disable_bw_measurement=bool, disable_bw=bool
4224 Disable measurements of throughput/bandwidth numbers. See
4225 :option:`disable_lat`.
4227 .. option:: slat_percentiles=bool
4229 Report submission latency percentiles. Submission latency is not recorded
4230 for synchronous ioengines.
4232 .. option:: clat_percentiles=bool
4234 Report completion latency percentiles.
4236 .. option:: lat_percentiles=bool
4238 Report total latency percentiles. Total latency is the sum of submission
4239 latency and completion latency.
4241 .. option:: percentile_list=float_list
4243 Overwrite the default list of percentiles for latencies and the block error
4244 histogram. Each number is a floating point number in the range (0,100], and
4245 the maximum length of the list is 20. Use ``:`` to separate the numbers. For
4246 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
4247 latency durations below which 99.5% and 99.9% of the observed latencies fell,
4250 .. option:: significant_figures=int
4252 If using :option:`--output-format` of `normal`, set the significant
4253 figures to this value. Higher values will yield more precise IOPS and
4254 throughput units, while lower values will round. Requires a minimum
4255 value of 1 and a maximum value of 10. Defaults to 4.
4261 .. option:: exitall_on_error
4263 When one job finishes in error, terminate the rest. The default is to wait
4264 for each job to finish.
4266 .. option:: continue_on_error=str
4268 Normally fio will exit the job on the first observed failure. If this option
4269 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
4270 EILSEQ) until the runtime is exceeded or the I/O size specified is
4271 completed. If this option is used, there are two more stats that are
4272 appended, the total error count and the first error. The error field given
4273 in the stats is the first error that was hit during the run.
4275 Note: a write error from the device may go unnoticed by fio when using
4276 buffered IO, as the write() (or similar) system call merely dirties the
4277 kernel pages, unless :option:`sync` or :option:`direct` is used. Device IO
4278 errors occur when the dirty data is actually written out to disk. If fully
4279 sync writes aren't desirable, :option:`fsync` or :option:`fdatasync` can be
4280 used as well. This is specific to writes, as reads are always synchronous.
4282 The allowed values are:
4285 Exit on any I/O or verify errors.
4288 Continue on read errors, exit on all others.
4291 Continue on write errors, exit on all others.
4294 Continue on any I/O error, exit on all others.
4297 Continue on verify errors, exit on all others.
4300 Continue on all errors.
4303 Backward-compatible alias for 'none'.
4306 Backward-compatible alias for 'all'.
4308 .. option:: ignore_error=str
4310 Sometimes you want to ignore some errors during test in that case you can
4311 specify error list for each error type, instead of only being able to
4312 ignore the default 'non-fatal error' using :option:`continue_on_error`.
4313 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
4314 given error type is separated with ':'. Error may be symbol ('ENOSPC',
4315 'ENOMEM') or integer. Example::
4317 ignore_error=EAGAIN,ENOSPC:122
4319 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
4320 WRITE. This option works by overriding :option:`continue_on_error` with
4321 the list of errors for each error type if any.
4323 .. option:: error_dump=bool
4325 If set dump every error even if it is non fatal, true by default. If
4326 disabled only fatal error will be dumped.
4328 Running predefined workloads
4329 ----------------------------
4331 Fio includes predefined profiles that mimic the I/O workloads generated by
4334 .. option:: profile=str
4336 The predefined workload to run. Current profiles are:
4339 Threaded I/O bench (tiotest/tiobench) like workload.
4342 Aerospike Certification Tool (ACT) like workload.
4344 To view a profile's additional options use :option:`--cmdhelp` after specifying
4345 the profile. For example::
4347 $ fio --profile=act --cmdhelp
4352 .. option:: device-names=str
4357 .. option:: load=int
4360 ACT load multiplier. Default: 1.
4362 .. option:: test-duration=time
4365 How long the entire test takes to run. When the unit is omitted, the value
4366 is given in seconds. Default: 24h.
4368 .. option:: threads-per-queue=int
4371 Number of read I/O threads per device. Default: 8.
4373 .. option:: read-req-num-512-blocks=int
4376 Number of 512B blocks to read at the time. Default: 3.
4378 .. option:: large-block-op-kbytes=int
4381 Size of large block ops in KiB (writes). Default: 131072.
4386 Set to run ACT prep phase.
4388 Tiobench profile options
4389 ~~~~~~~~~~~~~~~~~~~~~~~~
4391 .. option:: size=str
4396 .. option:: block=int
4399 Block size in bytes. Default: 4096.
4401 .. option:: numruns=int
4411 .. option:: threads=int
4416 Interpreting the output
4417 -----------------------
4420 Example output was based on the following:
4421 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
4422 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
4423 --runtime=2m --rw=rw
4425 Fio spits out a lot of output. While running, fio will display the status of the
4426 jobs created. An example of that would be::
4428 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]
4430 The characters inside the first set of square brackets denote the current status of
4431 each thread. The first character is the first job defined in the job file, and so
4432 forth. The possible values (in typical life cycle order) are:
4434 +------+-----+-----------------------------------------------------------+
4436 +======+=====+===========================================================+
4437 | P | | Thread setup, but not started. |
4438 +------+-----+-----------------------------------------------------------+
4439 | C | | Thread created. |
4440 +------+-----+-----------------------------------------------------------+
4441 | I | | Thread initialized, waiting or generating necessary data. |
4442 +------+-----+-----------------------------------------------------------+
4443 | | p | Thread running pre-reading file(s). |
4444 +------+-----+-----------------------------------------------------------+
4445 | | / | Thread is in ramp period. |
4446 +------+-----+-----------------------------------------------------------+
4447 | | R | Running, doing sequential reads. |
4448 +------+-----+-----------------------------------------------------------+
4449 | | r | Running, doing random reads. |
4450 +------+-----+-----------------------------------------------------------+
4451 | | W | Running, doing sequential writes. |
4452 +------+-----+-----------------------------------------------------------+
4453 | | w | Running, doing random writes. |
4454 +------+-----+-----------------------------------------------------------+
4455 | | M | Running, doing mixed sequential reads/writes. |
4456 +------+-----+-----------------------------------------------------------+
4457 | | m | Running, doing mixed random reads/writes. |
4458 +------+-----+-----------------------------------------------------------+
4459 | | D | Running, doing sequential trims. |
4460 +------+-----+-----------------------------------------------------------+
4461 | | d | Running, doing random trims. |
4462 +------+-----+-----------------------------------------------------------+
4463 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
4464 +------+-----+-----------------------------------------------------------+
4465 | | V | Running, doing verification of written data. |
4466 +------+-----+-----------------------------------------------------------+
4467 | f | | Thread finishing. |
4468 +------+-----+-----------------------------------------------------------+
4469 | E | | Thread exited, not reaped by main thread yet. |
4470 +------+-----+-----------------------------------------------------------+
4471 | _ | | Thread reaped. |
4472 +------+-----+-----------------------------------------------------------+
4473 | X | | Thread reaped, exited with an error. |
4474 +------+-----+-----------------------------------------------------------+
4475 | K | | Thread reaped, exited due to signal. |
4476 +------+-----+-----------------------------------------------------------+
4479 Example output was based on the following:
4480 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
4481 --time_based --rate=2512k --bs=256K --numjobs=10 \
4482 --name=readers --rw=read --name=writers --rw=write
4484 Fio will condense the thread string as not to take up more space on the command
4485 line than needed. For instance, if you have 10 readers and 10 writers running,
4486 the output would look like this::
4488 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]
4490 Note that the status string is displayed in order, so it's possible to tell which of
4491 the jobs are currently doing what. In the example above this means that jobs 1--10
4492 are readers and 11--20 are writers.
4494 The other values are fairly self explanatory -- number of threads currently
4495 running and doing I/O, the number of currently open files (f=), the estimated
4496 completion percentage, the rate of I/O since last check (read speed listed first,
4497 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
4498 and time to completion for the current running group. It's impossible to estimate
4499 runtime of the following groups (if any).
4502 Example output was based on the following:
4503 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
4504 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
4505 --bs=7K --name=Client1 --rw=write
4507 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
4508 each thread, group of threads, and disks in that order. For each overall thread (or
4509 group) the output looks like::
4511 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
4512 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
4513 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
4514 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
4515 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
4516 clat percentiles (usec):
4517 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
4518 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
4519 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
4520 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
4522 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
4523 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
4524 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
4525 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
4526 lat (msec) : 100=0.65%
4527 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
4528 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
4529 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4530 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4531 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
4532 latency : target=0, window=0, percentile=100.00%, depth=8
4534 The job name (or first job's name when using :option:`group_reporting`) is printed,
4535 along with the group id, count of jobs being aggregated, last error id seen (which
4536 is 0 when there are no errors), pid/tid of that thread and the time the job/group
4537 completed. Below are the I/O statistics for each data direction performed (showing
4538 writes in the example above). In the order listed, they denote:
4541 The string before the colon shows the I/O direction the statistics
4542 are for. **IOPS** is the average I/Os performed per second. **BW**
4543 is the average bandwidth rate shown as: value in power of 2 format
4544 (value in power of 10 format). The last two values show: (**total
4545 I/O performed** in power of 2 format / **runtime** of that thread).
4548 Submission latency (**min** being the minimum, **max** being the
4549 maximum, **avg** being the average, **stdev** being the standard
4550 deviation). This is the time from when fio initialized the I/O
4551 to submission. For synchronous ioengines this includes the time
4552 up until just before the ioengine's queue function is called.
4553 For asynchronous ioengines this includes the time up through the
4554 completion of the ioengine's queue function (and commit function
4555 if it is defined). For sync I/O this row is not displayed as the
4556 slat is negligible. This value can be in nanoseconds,
4557 microseconds or milliseconds --- fio will choose the most
4558 appropriate base and print that (in the example above
4559 nanoseconds was the best scale). Note: in :option:`--minimal`
4560 mode latencies are always expressed in microseconds.
4563 Completion latency. Same names as slat, this denotes the time from
4564 submission to completion of the I/O pieces. For sync I/O, this
4565 represents the time from when the I/O was submitted to the
4566 operating system to when it was completed. For asynchronous
4567 ioengines this is the time from when the ioengine's queue (and
4568 commit if available) functions were completed to when the I/O's
4569 completion was reaped by fio.
4572 Total latency. Same names as slat and clat, this denotes the time from
4573 when fio created the I/O unit to completion of the I/O operation.
4574 It is the sum of submission and completion latency.
4577 Bandwidth statistics based on measurements from discrete
4578 intervals. Fio continuously monitors bytes transferred and I/O
4579 operations completed. By default fio calculates bandwidth in
4580 each half-second interval (see :option:`bwavgtime`) and reports
4581 descriptive statistics for the measurements here. Same names as
4582 the xlat stats, but also includes the number of samples taken
4583 (**samples**) and an approximate percentage of total aggregate
4584 bandwidth this thread received in its group (**per**). This
4585 last value is only really useful if the threads in this group
4586 are on the same disk, since they are then competing for disk
4590 IOPS statistics based on measurements from discrete intervals.
4591 For details see the description for bw above. See
4592 :option:`iopsavgtime` to control the duration of the intervals.
4593 Same values reported here as for bw except for percentage.
4595 **lat (nsec/usec/msec)**
4596 The distribution of I/O completion latencies. This is the time from when
4597 I/O leaves fio and when it gets completed. Unlike the separate
4598 read/write/trim sections above, the data here and in the remaining
4599 sections apply to all I/Os for the reporting group. 250=0.04% means that
4600 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
4601 of the I/Os required 250 to 499us for completion.
4604 CPU usage. User and system time, along with the number of context
4605 switches this thread went through, usage of system and user time, and
4606 finally the number of major and minor page faults. The CPU utilization
4607 numbers are averages for the jobs in that reporting group, while the
4608 context and fault counters are summed.
4611 The distribution of I/O depths over the job lifetime. The numbers are
4612 divided into powers of 2 and each entry covers depths from that value
4613 up to those that are lower than the next entry -- e.g., 16= covers
4614 depths from 16 to 31. Note that the range covered by a depth
4615 distribution entry can be different to the range covered by the
4616 equivalent submit/complete distribution entry.
4619 How many pieces of I/O were submitting in a single submit call. Each
4620 entry denotes that amount and below, until the previous entry -- e.g.,
4621 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
4622 call. Note that the range covered by a submit distribution entry can
4623 be different to the range covered by the equivalent depth distribution
4627 Like the above submit number, but for completions instead.
4630 The number of read/write/trim requests issued, and how many of them were
4634 These values are for :option:`latency_target` and related options. When
4635 these options are engaged, this section describes the I/O depth required
4636 to meet the specified latency target.
4639 Example output was based on the following:
4640 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
4641 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
4642 --rate=11M --name=write --rw=write --bs=2k --rate=700k
4644 After each client has been listed, the group statistics are printed. They
4645 will look like this::
4647 Run status group 0 (all jobs):
4648 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
4649 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
4651 For each data direction it prints:
4654 Aggregate bandwidth of threads in this group followed by the
4655 minimum and maximum bandwidth of all the threads in this group.
4656 Values outside of brackets are power-of-2 format and those
4657 within are the equivalent value in a power-of-10 format.
4659 Aggregate I/O performed of all threads in this group. The
4660 format is the same as bw.
4662 The smallest and longest runtimes of the threads in this group.
4664 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
4666 Disk stats (read/write):
4667 sda: ios=16398/16511, sectors=32321/65472, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
4669 Each value is printed for both reads and writes, with reads first. The
4673 Number of I/Os performed by all groups.
4675 Amount of data transferred in units of 512 bytes for all groups.
4677 Number of merges performed by the I/O scheduler.
4679 Number of ticks we kept the disk busy.
4681 Total time spent in the disk queue.
4683 The disk utilization. A value of 100% means we kept the disk
4684 busy constantly, 50% would be a disk idling half of the time.
4686 It is also possible to get fio to dump the current output while it is running,
4687 without terminating the job. To do that, send fio the **USR1** signal. You can
4688 also get regularly timed dumps by using the :option:`--status-interval`
4689 parameter, or by creating a file in :file:`/tmp` named
4690 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
4691 current output status.
4697 For scripted usage where you typically want to generate tables or graphs of the
4698 results, fio can output the results in a semicolon separated format. The format
4699 is one long line of values, such as::
4701 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%
4702 A description of this job goes here.
4704 The job description (if provided) follows on a second line for terse v2.
4705 It appears on the same line for other terse versions.
4707 To enable terse output, use the :option:`--minimal` or
4708 :option:`--output-format`\=terse command line options. The
4709 first value is the version of the terse output format. If the output has to be
4710 changed for some reason, this number will be incremented by 1 to signify that
4713 Split up, the format is as follows (comments in brackets denote when a
4714 field was introduced or whether it's specific to some terse version):
4718 terse version, fio version [v3], jobname, groupid, error
4722 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4723 Submission latency: min, max, mean, stdev (usec)
4724 Completion latency: min, max, mean, stdev (usec)
4725 Completion latency percentiles: 20 fields (see below)
4726 Total latency: min, max, mean, stdev (usec)
4727 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4728 IOPS [v5]: min, max, mean, stdev, number of samples
4734 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4735 Submission latency: min, max, mean, stdev (usec)
4736 Completion latency: min, max, mean, stdev (usec)
4737 Completion latency percentiles: 20 fields (see below)
4738 Total latency: min, max, mean, stdev (usec)
4739 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4740 IOPS [v5]: min, max, mean, stdev, number of samples
4742 TRIM status [all but version 3]:
4744 Fields are similar to READ/WRITE status.
4748 user, system, context switches, major faults, minor faults
4752 <=1, 2, 4, 8, 16, 32, >=64
4754 I/O latencies microseconds::
4756 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
4758 I/O latencies milliseconds::
4760 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
4762 Disk utilization [v3]::
4764 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
4765 time spent in queue, disk utilization percentage
4767 Additional Info (dependent on continue_on_error, default off)::
4769 total # errors, first error code
4771 Additional Info (dependent on description being set)::
4775 Completion latency percentiles can be a grouping of up to 20 sets, so for the
4776 terse output fio writes all of them. Each field will look like this::
4780 which is the Xth percentile, and the `usec` latency associated with it.
4782 For `Disk utilization`, all disks used by fio are shown. So for each disk there
4783 will be a disk utilization section.
4785 Below is a single line containing short names for each of the fields in the
4786 minimal output v3, separated by semicolons::
4788 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
4790 In client/server mode terse output differs from what appears when jobs are run
4791 locally. Disk utilization data is omitted from the standard terse output and
4792 for v3 and later appears on its own separate line at the end of each terse
4799 The `json` output format is intended to be both human readable and convenient
4800 for automated parsing. For the most part its sections mirror those of the
4801 `normal` output. The `runtime` value is reported in msec and the `bw` value is
4802 reported in 1024 bytes per second units.
4808 The `json+` output format is identical to the `json` output format except that it
4809 adds a full dump of the completion latency bins. Each `bins` object contains a
4810 set of (key, value) pairs where keys are latency durations and values count how
4811 many I/Os had completion latencies of the corresponding duration. For example,
4814 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
4816 This data indicates that one I/O required 87,552ns to complete, two I/Os required
4817 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
4819 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
4820 json+ output and generates CSV-formatted latency data suitable for plotting.
4822 The latency durations actually represent the midpoints of latency intervals.
4823 For details refer to :file:`stat.h`.
4829 There are two trace file format that you can encounter. The older (v1) format is
4830 unsupported since version 1.20-rc3 (March 2008). It will still be described
4831 below in case that you get an old trace and want to understand it.
4833 In any case the trace is a simple text file with a single action per line.
4836 Trace file format v1
4837 ~~~~~~~~~~~~~~~~~~~~
4839 Each line represents a single I/O action in the following format::
4843 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
4845 This format is not supported in fio versions >= 1.20-rc3.
4848 Trace file format v2
4849 ~~~~~~~~~~~~~~~~~~~~
4851 The second version of the trace file format was added in fio version 1.17. It
4852 allows one to access more than one file per trace and has a bigger set of possible
4855 The first line of the trace file has to be::
4859 Following this can be lines in two different formats, which are described below.
4861 The file management format::
4865 The `filename` is given as an absolute path. The `action` can be one of these:
4868 Add the given `filename` to the trace.
4870 Open the file with the given `filename`. The `filename` has to have
4871 been added with the **add** action before.
4873 Close the file with the given `filename`. The file has to have been
4877 The file I/O action format::
4879 filename action offset length
4881 The `filename` is given as an absolute path, and has to have been added and
4882 opened before it can be used with this format. The `offset` and `length` are
4883 given in bytes. The `action` can be one of these:
4886 Wait for `offset` microseconds. Everything below 100 is discarded.
4887 The time is relative to the previous `wait` statement. Note that
4888 action `wait` is not allowed as of version 3, as the same behavior
4889 can be achieved using timestamps.
4891 Read `length` bytes beginning from `offset`.
4893 Write `length` bytes beginning from `offset`.
4895 :manpage:`fsync(2)` the file.
4897 :manpage:`fdatasync(2)` the file.
4899 Trim the given file from the given `offset` for `length` bytes.
4902 Trace file format v3
4903 ~~~~~~~~~~~~~~~~~~~~
4905 The third version of the trace file format was added in fio version 3.31. It
4906 forces each action to have a timestamp associated with it.
4908 The first line of the trace file has to be::
4912 Following this can be lines in two different formats, which are described below.
4914 The file management format::
4916 timestamp filename action
4918 The file I/O action format::
4920 timestamp filename action offset length
4922 The `timestamp` is relative to the beginning of the run (ie starts at 0). The
4923 `filename`, `action`, `offset` and `length` are identical to version 2, except
4924 that version 3 does not allow the `wait` action.
4927 I/O Replay - Merging Traces
4928 ---------------------------
4930 Colocation is a common practice used to get the most out of a machine.
4931 Knowing which workloads play nicely with each other and which ones don't is
4932 a much harder task. While fio can replay workloads concurrently via multiple
4933 jobs, it leaves some variability up to the scheduler making results harder to
4934 reproduce. Merging is a way to make the order of events consistent.
4936 Merging is integrated into I/O replay and done when a
4937 :option:`merge_blktrace_file` is specified. The list of files passed to
4938 :option:`read_iolog` go through the merge process and output a single file
4939 stored to the specified file. The output file is passed on as if it were the
4940 only file passed to :option:`read_iolog`. An example would look like::
4942 $ fio --read_iolog="<file1>:<file2>" --merge_blktrace_file="<output_file>"
4944 Creating only the merged file can be done by passing the command line argument
4945 :option:`--merge-blktrace-only`.
4947 Scaling traces can be done to see the relative impact of any particular trace
4948 being slowed down or sped up. :option:`merge_blktrace_scalars` takes in a colon
4949 separated list of percentage scalars. It is index paired with the files passed
4950 to :option:`read_iolog`.
4952 With scaling, it may be desirable to match the running time of all traces.
4953 This can be done with :option:`merge_blktrace_iters`. It is index paired with
4954 :option:`read_iolog` just like :option:`merge_blktrace_scalars`.
4956 In an example, given two traces, A and B, each 60s long. If we want to see
4957 the impact of trace A issuing IOs twice as fast and repeat trace A over the
4958 runtime of trace B, the following can be done::
4960 $ fio --read_iolog="<trace_a>:"<trace_b>" --merge_blktrace_file"<output_file>" --merge_blktrace_scalars="50:100" --merge_blktrace_iters="2:1"
4962 This runs trace A at 2x the speed twice for approximately the same runtime as
4963 a single run of trace B.
4966 CPU idleness profiling
4967 ----------------------
4969 In some cases, we want to understand CPU overhead in a test. For example, we
4970 test patches for the specific goodness of whether they reduce CPU usage.
4971 Fio implements a balloon approach to create a thread per CPU that runs at idle
4972 priority, meaning that it only runs when nobody else needs the cpu.
4973 By measuring the amount of work completed by the thread, idleness of each CPU
4974 can be derived accordingly.
4976 An unit work is defined as touching a full page of unsigned characters. Mean and
4977 standard deviation of time to complete an unit work is reported in "unit work"
4978 section. Options can be chosen to report detailed percpu idleness or overall
4979 system idleness by aggregating percpu stats.
4982 Verification and triggers
4983 -------------------------
4985 Fio is usually run in one of two ways, when data verification is done. The first
4986 is a normal write job of some sort with verify enabled. When the write phase has
4987 completed, fio switches to reads and verifies everything it wrote. The second
4988 model is running just the write phase, and then later on running the same job
4989 (but with reads instead of writes) to repeat the same I/O patterns and verify
4990 the contents. Both of these methods depend on the write phase being completed,
4991 as fio otherwise has no idea how much data was written.
4993 With verification triggers, fio supports dumping the current write state to
4994 local files. Then a subsequent read verify workload can load this state and know
4995 exactly where to stop. This is useful for testing cases where power is cut to a
4996 server in a managed fashion, for instance.
4998 A verification trigger consists of two things:
5000 1) Storing the write state of each job.
5001 2) Executing a trigger command.
5003 The write state is relatively small, on the order of hundreds of bytes to single
5004 kilobytes. It contains information on the number of completions done, the last X
5007 A trigger is invoked either through creation ('touch') of a specified file in
5008 the system, or through a timeout setting. If fio is run with
5009 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
5010 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
5011 will fire off the trigger (thus saving state, and executing the trigger
5014 For client/server runs, there's both a local and remote trigger. If fio is
5015 running as a server backend, it will send the job states back to the client for
5016 safe storage, then execute the remote trigger, if specified. If a local trigger
5017 is specified, the server will still send back the write state, but the client
5018 will then execute the trigger.
5020 Verification trigger example
5021 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
5023 Let's say we want to run a powercut test on the remote Linux machine 'server'.
5024 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
5025 some point during the run, and we'll run this test from the safety or our local
5026 machine, 'localbox'. On the server, we'll start the fio backend normally::
5028 server# fio --server
5030 and on the client, we'll fire off the workload::
5032 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
5034 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
5036 echo b > /proc/sysrq-trigger
5038 on the server once it has received the trigger and sent us the write state. This
5039 will work, but it's not **really** cutting power to the server, it's merely
5040 abruptly rebooting it. If we have a remote way of cutting power to the server
5041 through IPMI or similar, we could do that through a local trigger command
5042 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
5043 ipmi-reboot. On localbox, we could then have run fio with a local trigger
5046 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
5048 For this case, fio would wait for the server to send us the write state, then
5049 execute ``ipmi-reboot server`` when that happened.
5051 Loading verify state
5052 ~~~~~~~~~~~~~~~~~~~~
5054 To load stored write state, a read verification job file must contain the
5055 :option:`verify_state_load` option. If that is set, fio will load the previously
5056 stored state. For a local fio run this is done by loading the files directly,
5057 and on a client/server run, the server backend will ask the client to send the
5058 files over and load them from there.
5064 Fio supports a variety of log file formats, for logging latencies, bandwidth,
5065 and IOPS. The logs share a common format, which looks like this:
5067 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
5068 *offset* (`bytes`), *command priority*
5070 *Time* for the log entry is always in milliseconds. The *value* logged depends
5071 on the type of log, it will be one of the following:
5074 Value is latency in nsecs
5080 *Data direction* is one of the following:
5089 The entry's *block size* is always in bytes. The *offset* is the position in bytes
5090 from the start of the file for that particular I/O. The logging of the offset can be
5091 toggled with :option:`log_offset`.
5093 *Command priority* is 0 for normal priority and 1 for high priority. This is controlled
5094 by the ioengine specific :option:`cmdprio_percentage`.
5096 Fio defaults to logging every individual I/O but when windowed logging is set
5097 through :option:`log_avg_msec`, either the average (by default), the maximum
5098 (:option:`log_window_value` is set to max) *value* seen over the specified period
5099 of time, or both the average *value* and maximum *value1* (:option:`log_window_value`
5100 is set to both) is recorded. The log file format when both the values are reported
5103 *time* (`msec`), *value*, *value1*, *data direction*, *block size* (`bytes`),
5104 *offset* (`bytes`), *command priority*
5107 Each *data direction* seen within the window period will aggregate its values in a
5108 separate row. Further, when using windowed logging the *block size* and *offset*
5109 entries will always contain 0.
5115 Normally fio is invoked as a stand-alone application on the machine where the
5116 I/O workload should be generated. However, the backend and frontend of fio can
5117 be run separately i.e., the fio server can generate an I/O workload on the "Device
5118 Under Test" while being controlled by a client on another machine.
5120 Start the server on the machine which has access to the storage DUT::
5124 where `args` defines what fio listens to. The arguments are of the form
5125 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
5126 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
5127 *hostname* is either a hostname or IP address, and *port* is the port to listen
5128 to (only valid for TCP/IP, not a local socket). Some examples:
5132 Start a fio server, listening on all interfaces on the default port (8765).
5134 2) ``fio --server=ip:hostname,4444``
5136 Start a fio server, listening on IP belonging to hostname and on port 4444.
5138 3) ``fio --server=ip6:::1,4444``
5140 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
5142 4) ``fio --server=,4444``
5144 Start a fio server, listening on all interfaces on port 4444.
5146 5) ``fio --server=1.2.3.4``
5148 Start a fio server, listening on IP 1.2.3.4 on the default port.
5150 6) ``fio --server=sock:/tmp/fio.sock``
5152 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
5154 Once a server is running, a "client" can connect to the fio server with::
5156 fio <local-args> --client=<server> <remote-args> <job file(s)>
5158 where `local-args` are arguments for the client where it is running, `server`
5159 is the connect string, and `remote-args` and `job file(s)` are sent to the
5160 server. The `server` string follows the same format as it does on the server
5161 side, to allow IP/hostname/socket and port strings.
5163 Note that all job options must be defined in job files when running fio as a
5164 client. Any job options specified in `remote-args` will be ignored.
5166 Fio can connect to multiple servers this way::
5168 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
5170 If the job file is located on the fio server, then you can tell the server to
5171 load a local file as well. This is done by using :option:`--remote-config` ::
5173 fio --client=server --remote-config /path/to/file.fio
5175 Then fio will open this local (to the server) job file instead of being passed
5176 one from the client.
5178 If you have many servers (example: 100 VMs/containers), you can input a pathname
5179 of a file containing host IPs/names as the parameter value for the
5180 :option:`--client` option. For example, here is an example :file:`host.list`
5181 file containing 2 hostnames::
5183 host1.your.dns.domain
5184 host2.your.dns.domain
5186 The fio command would then be::
5188 fio --client=host.list <job file(s)>
5190 In this mode, you cannot input server-specific parameters or job files -- all
5191 servers receive the same job file.
5193 In order to let ``fio --client`` runs use a shared filesystem from multiple
5194 hosts, ``fio --client`` now prepends the IP address of the server to the
5195 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
5196 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
5197 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
5198 192.168.10.121, then fio will create two files::
5200 /mnt/nfs/fio/192.168.10.120.fileio.tmp
5201 /mnt/nfs/fio/192.168.10.121.fileio.tmp
5203 Terse output in client/server mode will differ slightly from what is produced
5204 when fio is run in stand-alone mode. See the terse output section for details.