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 Show available debug options.
99 .. option:: --parse-only
101 Parse options only, don't start any I/O.
103 .. option:: --output=filename
105 Write output to file `filename`.
107 .. option:: --output-format=format
109 Set the reporting `format` to `normal`, `terse`, `json`, or `json+`. Multiple
110 formats can be selected, separated by a comma. `terse` is a CSV based
111 format. `json+` is like `json`, except it adds a full dump of the latency
114 .. option:: --bandwidth-log
116 Generate aggregate bandwidth logs.
118 .. option:: --minimal
120 Print statistics in a terse, semicolon-delimited format.
122 .. option:: --append-terse
124 Print statistics in selected mode AND terse, semicolon-delimited format.
125 **Deprecated**, use :option:`--output-format` instead to select multiple
128 .. option:: --terse-version=version
130 Set terse `version` output format (default 3, or 2 or 4 or 5).
132 .. option:: --version
134 Print version information and exit.
138 Print a summary of the command line options and exit.
140 .. option:: --cpuclock-test
142 Perform test and validation of internal CPU clock.
144 .. option:: --crctest=[test]
146 Test the speed of the built-in checksumming functions. If no argument is
147 given, all of them are tested. Alternatively, a comma separated list can
148 be passed, in which case the given ones are tested.
150 .. option:: --cmdhelp=command
152 Print help information for `command`. May be ``all`` for all commands.
154 .. option:: --enghelp=[ioengine[,command]]
156 List all commands defined by `ioengine`, or print help for `command`
157 defined by `ioengine`. If no `ioengine` is given, list all
160 .. option:: --showcmd=jobfile
162 Convert `jobfile` to a set of command-line options.
164 .. option:: --readonly
166 Turn on safety read-only checks, preventing writes and trims. The
167 ``--readonly`` option is an extra safety guard to prevent users from
168 accidentally starting a write or trim workload when that is not desired.
169 Fio will only modify the device under test if
170 `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite` is given. This
171 safety net can be used as an extra precaution.
173 .. option:: --eta=when
175 Specifies when real-time ETA estimate should be printed. `when` may be
176 `always`, `never` or `auto`. `auto` is the default, it prints ETA
177 when requested if the output is a TTY. `always` disregards the output
178 type, and prints ETA when requested. `never` never prints ETA.
180 .. option:: --eta-interval=time
182 By default, fio requests client ETA status roughly every second. With
183 this option, the interval is configurable. Fio imposes a minimum
184 allowed time to avoid flooding the console, less than 250 msec is
187 .. option:: --eta-newline=time
189 Force a new line for every `time` period passed. When the unit is omitted,
190 the value is interpreted in seconds.
192 .. option:: --status-interval=time
194 Force a full status dump of cumulative (from job start) values at `time`
195 intervals. This option does *not* provide per-period measurements. So
196 values such as bandwidth are running averages. When the time unit is omitted,
197 `time` is interpreted in seconds.
199 .. option:: --section=name
201 Only run specified section `name` in job file. Multiple sections can be specified.
202 The ``--section`` option allows one to combine related jobs into one file.
203 E.g. one job file could define light, moderate, and heavy sections. Tell
204 fio to run only the "heavy" section by giving ``--section=heavy``
205 command line option. One can also specify the "write" operations in one
206 section and "verify" operation in another section. The ``--section`` option
207 only applies to job sections. The reserved *global* section is always
210 .. option:: --alloc-size=kb
212 Set the internal smalloc pool size to `kb` in KiB. The
213 ``--alloc-size`` switch allows one to use a larger pool size for smalloc.
214 If running large jobs with randommap enabled, fio can run out of memory.
215 Smalloc is an internal allocator for shared structures from a fixed size
216 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
218 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
221 .. option:: --warnings-fatal
223 All fio parser warnings are fatal, causing fio to exit with an
226 .. option:: --max-jobs=nr
228 Set the maximum number of threads/processes to support to `nr`.
229 NOTE: On Linux, it may be necessary to increase the shared-memory
230 limit (:file:`/proc/sys/kernel/shmmax`) if fio runs into errors while
233 .. option:: --server=args
235 Start a backend server, with `args` specifying what to listen to.
236 See `Client/Server`_ section.
238 .. option:: --daemonize=pidfile
240 Background a fio server, writing the pid to the given `pidfile` file.
242 .. option:: --client=hostname
244 Instead of running the jobs locally, send and run them on the given `hostname`
245 or set of `hostname`\s. See `Client/Server`_ section.
247 .. option:: --remote-config=file
249 Tell fio server to load this local `file`.
251 .. option:: --idle-prof=option
253 Report CPU idleness. `option` is one of the following:
256 Run unit work calibration only and exit.
259 Show aggregate system idleness and unit work.
262 As **system** but also show per CPU idleness.
264 .. option:: --inflate-log=log
266 Inflate and output compressed `log`.
268 .. option:: --trigger-file=file
270 Execute trigger command when `file` exists.
272 .. option:: --trigger-timeout=time
274 Execute trigger at this `time`.
276 .. option:: --trigger=command
278 Set this `command` as local trigger.
280 .. option:: --trigger-remote=command
282 Set this `command` as remote trigger.
284 .. option:: --aux-path=path
286 Use this `path` for fio state generated files.
288 Any parameters following the options will be assumed to be job files, unless
289 they match a job file parameter. Multiple job files can be listed and each job
290 file will be regarded as a separate group. Fio will :option:`stonewall`
291 execution between each group.
297 As previously described, fio accepts one or more job files describing what it is
298 supposed to do. The job file format is the classic ini file, where the names
299 enclosed in [] brackets define the job name. You are free to use any ASCII name
300 you want, except *global* which has special meaning. Following the job name is
301 a sequence of zero or more parameters, one per line, that define the behavior of
302 the job. If the first character in a line is a ';' or a '#', the entire line is
303 discarded as a comment.
305 A *global* section sets defaults for the jobs described in that file. A job may
306 override a *global* section parameter, and a job file may even have several
307 *global* sections if so desired. A job is only affected by a *global* section
310 The :option:`--cmdhelp` option also lists all options. If used with a `command`
311 argument, :option:`--cmdhelp` will detail the given `command`.
313 See the `examples/` directory for inspiration on how to write job files. Note
314 the copyright and license requirements currently apply to `examples/` files.
316 So let's look at a really simple job file that defines two processes, each
317 randomly reading from a 128MiB file:
321 ; -- start job file --
332 As you can see, the job file sections themselves are empty as all the described
333 parameters are shared. As no :option:`filename` option is given, fio makes up a
334 `filename` for each of the jobs as it sees fit. On the command line, this job
335 would look as follows::
337 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
340 Let's look at an example that has a number of processes writing randomly to
345 ; -- start job file --
356 Here we have no *global* section, as we only have one job defined anyway. We
357 want to use async I/O here, with a depth of 4 for each file. We also increased
358 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
359 jobs. The result is 4 processes each randomly writing to their own 64MiB
360 file. Instead of using the above job file, you could have given the parameters
361 on the command line. For this case, you would specify::
363 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
365 When fio is utilized as a basis of any reasonably large test suite, it might be
366 desirable to share a set of standardized settings across multiple job files.
367 Instead of copy/pasting such settings, any section may pull in an external
368 :file:`filename.fio` file with *include filename* directive, as in the following
371 ; -- start job file including.fio --
375 include glob-include.fio
382 include test-include.fio
383 ; -- end job file including.fio --
387 ; -- start job file glob-include.fio --
390 ; -- end job file glob-include.fio --
394 ; -- start job file test-include.fio --
397 ; -- end job file test-include.fio --
399 Settings pulled into a section apply to that section only (except *global*
400 section). Include directives may be nested in that any included file may contain
401 further include directive(s). Include files may not contain [] sections.
404 Environment variables
405 ~~~~~~~~~~~~~~~~~~~~~
407 Fio also supports environment variable expansion in job files. Any sub-string of
408 the form ``${VARNAME}`` as part of an option value (in other words, on the right
409 of the '='), will be expanded to the value of the environment variable called
410 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
411 empty string, the empty string will be substituted.
413 As an example, let's look at a sample fio invocation and job file::
415 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
419 ; -- start job file --
426 This will expand to the following equivalent job file at runtime:
430 ; -- start job file --
437 Fio ships with a few example job files, you can also look there for inspiration.
442 Additionally, fio has a set of reserved keywords that will be replaced
443 internally with the appropriate value. Those keywords are:
447 The architecture page size of the running system.
451 Megabytes of total memory in the system.
455 Number of online available CPUs.
457 These can be used on the command line or in the job file, and will be
458 automatically substituted with the current system values when the job is
459 run. Simple math is also supported on these keywords, so you can perform actions
464 and get that properly expanded to 8 times the size of memory in the machine.
470 This section describes in details each parameter associated with a job. Some
471 parameters take an option of a given type, such as an integer or a
472 string. Anywhere a numeric value is required, an arithmetic expression may be
473 used, provided it is surrounded by parentheses. Supported operators are:
482 For time values in expressions, units are microseconds by default. This is
483 different than for time values not in expressions (not enclosed in
484 parentheses). The following types are used:
491 String: A sequence of alphanumeric characters.
494 Integer with possible time suffix. Without a unit value is interpreted as
495 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
496 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
497 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
502 Integer. A whole number value, which may contain an integer prefix
503 and an integer suffix:
505 [*integer prefix*] **number** [*integer suffix*]
507 The optional *integer prefix* specifies the number's base. The default
508 is decimal. *0x* specifies hexadecimal.
510 The optional *integer suffix* specifies the number's units, and includes an
511 optional unit prefix and an optional unit. For quantities of data, the
512 default unit is bytes. For quantities of time, the default unit is seconds
513 unless otherwise specified.
515 With :option:`kb_base`\=1000, fio follows international standards for unit
516 prefixes. To specify power-of-10 decimal values defined in the
517 International System of Units (SI):
519 * *K* -- means kilo (K) or 1000
520 * *M* -- means mega (M) or 1000**2
521 * *G* -- means giga (G) or 1000**3
522 * *T* -- means tera (T) or 1000**4
523 * *P* -- means peta (P) or 1000**5
525 To specify power-of-2 binary values defined in IEC 80000-13:
527 * *Ki* -- means kibi (Ki) or 1024
528 * *Mi* -- means mebi (Mi) or 1024**2
529 * *Gi* -- means gibi (Gi) or 1024**3
530 * *Ti* -- means tebi (Ti) or 1024**4
531 * *Pi* -- means pebi (Pi) or 1024**5
533 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
534 from those specified in the SI and IEC 80000-13 standards to provide
535 compatibility with old scripts. For example, 4k means 4096.
537 For quantities of data, an optional unit of 'B' may be included
538 (e.g., 'kB' is the same as 'k').
540 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
541 not milli). 'b' and 'B' both mean byte, not bit.
543 Examples with :option:`kb_base`\=1000:
545 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
546 * *1 MiB*: 1048576, 1mi, 1024ki
547 * *1 MB*: 1000000, 1m, 1000k
548 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
549 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
551 Examples with :option:`kb_base`\=1024 (default):
553 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
554 * *1 MiB*: 1048576, 1m, 1024k
555 * *1 MB*: 1000000, 1mi, 1000ki
556 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
557 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
559 To specify times (units are not case sensitive):
563 * *M* -- means minutes
564 * *s* -- or sec means seconds (default)
565 * *ms* -- or *msec* means milliseconds
566 * *us* -- or *usec* means microseconds
568 If the option accepts an upper and lower range, use a colon ':' or
569 minus '-' to separate such values. See :ref:`irange <irange>`.
570 If the lower value specified happens to be larger than the upper value
571 the two values are swapped.
576 Boolean. Usually parsed as an integer, however only defined for
577 true and false (1 and 0).
582 Integer range with suffix. Allows value range to be given, such as
583 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
584 option allows two sets of ranges, they can be specified with a ',' or '/'
585 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
588 A list of floating point numbers, separated by a ':' character.
590 With the above in mind, here follows the complete list of fio job parameters.
596 .. option:: kb_base=int
598 Select the interpretation of unit prefixes in input parameters.
601 Inputs comply with IEC 80000-13 and the International
602 System of Units (SI). Use:
604 - power-of-2 values with IEC prefixes (e.g., KiB)
605 - power-of-10 values with SI prefixes (e.g., kB)
608 Compatibility mode (default). To avoid breaking old scripts:
610 - power-of-2 values with SI prefixes
611 - power-of-10 values with IEC prefixes
613 See :option:`bs` for more details on input parameters.
615 Outputs always use correct prefixes. Most outputs include both
618 bw=2383.3kB/s (2327.4KiB/s)
620 If only one value is reported, then kb_base selects the one to use:
622 **1000** -- SI prefixes
624 **1024** -- IEC prefixes
626 .. option:: unit_base=int
628 Base unit for reporting. Allowed values are:
631 Use auto-detection (default).
643 ASCII name of the job. This may be used to override the name printed by fio
644 for this job. Otherwise the job name is used. On the command line this
645 parameter has the special purpose of also signaling the start of a new job.
647 .. option:: description=str
649 Text description of the job. Doesn't do anything except dump this text
650 description when this job is run. It's not parsed.
652 .. option:: loops=int
654 Run the specified number of iterations of this job. Used to repeat the same
655 workload a given number of times. Defaults to 1.
657 .. option:: numjobs=int
659 Create the specified number of clones of this job. Each clone of job
660 is spawned as an independent thread or process. May be used to setup a
661 larger number of threads/processes doing the same thing. Each thread is
662 reported separately; to see statistics for all clones as a whole, use
663 :option:`group_reporting` in conjunction with :option:`new_group`.
664 See :option:`--max-jobs`. Default: 1.
667 Time related parameters
668 ~~~~~~~~~~~~~~~~~~~~~~~
670 .. option:: runtime=time
672 Tell fio to terminate processing after the specified period of time. It
673 can be quite hard to determine for how long a specified job will run, so
674 this parameter is handy to cap the total runtime to a given time. When
675 the unit is omitted, the value is interpreted in seconds.
677 .. option:: time_based
679 If set, fio will run for the duration of the :option:`runtime` specified
680 even if the file(s) are completely read or written. It will simply loop over
681 the same workload as many times as the :option:`runtime` allows.
683 .. option:: startdelay=irange(time)
685 Delay the start of job for the specified amount of time. Can be a single
686 value or a range. When given as a range, each thread will choose a value
687 randomly from within the range. Value is in seconds if a unit is omitted.
689 .. option:: ramp_time=time
691 If set, fio will run the specified workload for this amount of time before
692 logging any performance numbers. Useful for letting performance settle
693 before logging results, thus minimizing the runtime required for stable
694 results. Note that the ``ramp_time`` is considered lead in time for a job,
695 thus it will increase the total runtime if a special timeout or
696 :option:`runtime` is specified. When the unit is omitted, the value is
699 .. option:: clocksource=str
701 Use the given clocksource as the base of timing. The supported options are:
704 :manpage:`gettimeofday(2)`
707 :manpage:`clock_gettime(2)`
710 Internal CPU clock source
712 cpu is the preferred clocksource if it is reliable, as it is very fast (and
713 fio is heavy on time calls). Fio will automatically use this clocksource if
714 it's supported and considered reliable on the system it is running on,
715 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
716 means supporting TSC Invariant.
718 .. option:: gtod_reduce=bool
720 Enable all of the :manpage:`gettimeofday(2)` reducing options
721 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
722 reduce precision of the timeout somewhat to really shrink the
723 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
724 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
725 time keeping was enabled.
727 .. option:: gtod_cpu=int
729 Sometimes it's cheaper to dedicate a single thread of execution to just
730 getting the current time. Fio (and databases, for instance) are very
731 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
732 one CPU aside for doing nothing but logging current time to a shared memory
733 location. Then the other threads/processes that run I/O workloads need only
734 copy that segment, instead of entering the kernel with a
735 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
736 calls will be excluded from other uses. Fio will manually clear it from the
737 CPU mask of other jobs.
743 .. option:: directory=str
745 Prefix filenames with this directory. Used to place files in a different
746 location than :file:`./`. You can specify a number of directories by
747 separating the names with a ':' character. These directories will be
748 assigned equally distributed to job clones created by :option:`numjobs` as
749 long as they are using generated filenames. If specific `filename(s)` are
750 set fio will use the first listed directory, and thereby matching the
751 `filename` semantic which generates a file each clone if not specified, but
752 let all clones use the same if set.
754 See the :option:`filename` option for information on how to escape "``:``" and
755 "``\``" characters within the directory path itself.
757 .. option:: filename=str
759 Fio normally makes up a `filename` based on the job name, thread number, and
760 file number (see :option:`filename_format`). If you want to share files
761 between threads in a job or several
762 jobs with fixed file paths, specify a `filename` for each of them to override
763 the default. If the ioengine is file based, you can specify a number of files
764 by separating the names with a ':' colon. So if you wanted a job to open
765 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
766 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
767 specified, :option:`nrfiles` is ignored. The size of regular files specified
768 by this option will be :option:`size` divided by number of files unless an
769 explicit size is specified by :option:`filesize`.
771 Each colon and backslash in the wanted path must be escaped with a ``\``
772 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
773 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
774 :file:`F:\\filename` then you would use ``filename=F\:\\filename``.
776 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
777 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
778 Note: Windows and FreeBSD prevent write access to areas
779 of the disk containing in-use data (e.g. filesystems).
781 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
782 of the two depends on the read/write direction set.
784 .. option:: filename_format=str
786 If sharing multiple files between jobs, it is usually necessary to have fio
787 generate the exact names that you want. By default, fio will name a file
788 based on the default file format specification of
789 :file:`jobname.jobnumber.filenumber`. With this option, that can be
790 customized. Fio will recognize and replace the following keywords in this
794 The name of the worker thread or process.
796 The incremental number of the worker thread or process.
798 The incremental number of the file for that worker thread or
801 To have dependent jobs share a set of files, this option can be set to have
802 fio generate filenames that are shared between the two. For instance, if
803 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
804 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
805 will be used if no other format specifier is given.
807 If you specify a path then the directories will be created up to the
808 main directory for the file. So for example if you specify
809 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
810 created before the file setup part of the job. If you specify
811 :option:`directory` then the path will be relative that directory,
812 otherwise it is treated as the absolute path.
814 .. option:: unique_filename=bool
816 To avoid collisions between networked clients, fio defaults to prefixing any
817 generated filenames (with a directory specified) with the source of the
818 client connecting. To disable this behavior, set this option to 0.
820 .. option:: opendir=str
822 Recursively open any files below directory `str`.
824 .. option:: lockfile=str
826 Fio defaults to not locking any files before it does I/O to them. If a file
827 or file descriptor is shared, fio can serialize I/O to that file to make the
828 end result consistent. This is usual for emulating real workloads that share
829 files. The lock modes are:
832 No locking. The default.
834 Only one thread or process may do I/O at a time, excluding all
837 Read-write locking on the file. Many readers may
838 access the file at the same time, but writes get exclusive access.
840 .. option:: nrfiles=int
842 Number of files to use for this job. Defaults to 1. The size of files
843 will be :option:`size` divided by this unless explicit size is specified by
844 :option:`filesize`. Files are created for each thread separately, and each
845 file will have a file number within its name by default, as explained in
846 :option:`filename` section.
849 .. option:: openfiles=int
851 Number of files to keep open at the same time. Defaults to the same as
852 :option:`nrfiles`, can be set smaller to limit the number simultaneous
855 .. option:: file_service_type=str
857 Defines how fio decides which file from a job to service next. The following
861 Choose a file at random.
864 Round robin over opened files. This is the default.
867 Finish one file before moving on to the next. Multiple files can
868 still be open depending on :option:`openfiles`.
871 Use a *Zipf* distribution to decide what file to access.
874 Use a *Pareto* distribution to decide what file to access.
877 Use a *Gaussian* (normal) distribution to decide what file to
883 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
884 tell fio how many I/Os to issue before switching to a new file. For example,
885 specifying ``file_service_type=random:8`` would cause fio to issue
886 8 I/Os before selecting a new file at random. For the non-uniform
887 distributions, a floating point postfix can be given to influence how the
888 distribution is skewed. See :option:`random_distribution` for a description
889 of how that would work.
891 .. option:: ioscheduler=str
893 Attempt to switch the device hosting the file to the specified I/O scheduler
896 .. option:: create_serialize=bool
898 If true, serialize the file creation for the jobs. This may be handy to
899 avoid interleaving of data files, which may greatly depend on the filesystem
900 used and even the number of processors in the system. Default: true.
902 .. option:: create_fsync=bool
904 :manpage:`fsync(2)` the data file after creation. This is the default.
906 .. option:: create_on_open=bool
908 If true, don't pre-create files but allow the job's open() to create a file
909 when it's time to do I/O. Default: false -- pre-create all necessary files
912 .. option:: create_only=bool
914 If true, fio will only run the setup phase of the job. If files need to be
915 laid out or updated on disk, only that will be done -- the actual job contents
916 are not executed. Default: false.
918 .. option:: allow_file_create=bool
920 If true, fio is permitted to create files as part of its workload. If this
921 option is false, then fio will error out if
922 the files it needs to use don't already exist. Default: true.
924 .. option:: allow_mounted_write=bool
926 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
927 to what appears to be a mounted device or partition. This should help catch
928 creating inadvertently destructive tests, not realizing that the test will
929 destroy data on the mounted file system. Note that some platforms don't allow
930 writing against a mounted device regardless of this option. Default: false.
932 .. option:: pre_read=bool
934 If this is given, files will be pre-read into memory before starting the
935 given I/O operation. This will also clear the :option:`invalidate` flag,
936 since it is pointless to pre-read and then drop the cache. This will only
937 work for I/O engines that are seek-able, since they allow you to read the
938 same data multiple times. Thus it will not work on non-seekable I/O engines
939 (e.g. network, splice). Default: false.
941 .. option:: unlink=bool
943 Unlink the job files when done. Not the default, as repeated runs of that
944 job would then waste time recreating the file set again and again. Default:
947 .. option:: unlink_each_loop=bool
949 Unlink job files after each iteration or loop. Default: false.
951 .. option:: zonesize=int
953 Divide a file into zones of the specified size. See :option:`zoneskip`.
955 .. option:: zonerange=int
957 Give size of an I/O zone. See :option:`zoneskip`.
959 .. option:: zoneskip=int
961 Skip the specified number of bytes when :option:`zonesize` data has been
962 read. The two zone options can be used to only do I/O on zones of a file.
968 .. option:: direct=bool
970 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
971 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
972 ioengines don't support direct I/O. Default: false.
974 .. option:: atomic=bool
976 If value is true, attempt to use atomic direct I/O. Atomic writes are
977 guaranteed to be stable once acknowledged by the operating system. Only
978 Linux supports O_ATOMIC right now.
980 .. option:: buffered=bool
982 If value is true, use buffered I/O. This is the opposite of the
983 :option:`direct` option. Defaults to true.
985 .. option:: readwrite=str, rw=str
987 Type of I/O pattern. Accepted values are:
994 Sequential trims (Linux block devices only).
1000 Random trims (Linux block devices only).
1002 Sequential mixed reads and writes.
1004 Random mixed reads and writes.
1006 Sequential trim+write sequences. Blocks will be trimmed first,
1007 then the same blocks will be written to.
1009 Fio defaults to read if the option is not specified. For the mixed I/O
1010 types, the default is to split them 50/50. For certain types of I/O the
1011 result may still be skewed a bit, since the speed may be different.
1013 It is possible to specify the number of I/Os to do before getting a new
1014 offset by appending ``:<nr>`` to the end of the string given. For a
1015 random read, it would look like ``rw=randread:8`` for passing in an offset
1016 modifier with a value of 8. If the suffix is used with a sequential I/O
1017 pattern, then the *<nr>* value specified will be **added** to the generated
1018 offset for each I/O turning sequential I/O into sequential I/O with holes.
1019 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1020 the :option:`rw_sequencer` option.
1022 .. option:: rw_sequencer=str
1024 If an offset modifier is given by appending a number to the ``rw=<str>``
1025 line, then this option controls how that number modifies the I/O offset
1026 being generated. Accepted values are:
1029 Generate sequential offset.
1031 Generate the same offset.
1033 ``sequential`` is only useful for random I/O, where fio would normally
1034 generate a new random offset for every I/O. If you append e.g. 8 to randread,
1035 you would get a new random offset for every 8 I/Os. The result would be a
1036 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
1037 to specify that. As sequential I/O is already sequential, setting
1038 ``sequential`` for that would not result in any differences. ``identical``
1039 behaves in a similar fashion, except it sends the same offset 8 number of
1040 times before generating a new offset.
1042 .. option:: unified_rw_reporting=bool
1044 Fio normally reports statistics on a per data direction basis, meaning that
1045 reads, writes, and trims are accounted and reported separately. If this
1046 option is set fio sums the results and report them as "mixed" instead.
1048 .. option:: randrepeat=bool
1050 Seed the random number generator used for random I/O patterns in a
1051 predictable way so the pattern is repeatable across runs. Default: true.
1053 .. option:: allrandrepeat=bool
1055 Seed all random number generators in a predictable way so results are
1056 repeatable across runs. Default: false.
1058 .. option:: randseed=int
1060 Seed the random number generators based on this seed value, to be able to
1061 control what sequence of output is being generated. If not set, the random
1062 sequence depends on the :option:`randrepeat` setting.
1064 .. option:: fallocate=str
1066 Whether pre-allocation is performed when laying down files.
1067 Accepted values are:
1070 Do not pre-allocate space.
1073 Use a platform's native pre-allocation call but fall back to
1074 **none** behavior if it fails/is not implemented.
1077 Pre-allocate via :manpage:`posix_fallocate(3)`.
1080 Pre-allocate via :manpage:`fallocate(2)` with
1081 FALLOC_FL_KEEP_SIZE set.
1084 Backward-compatible alias for **none**.
1087 Backward-compatible alias for **posix**.
1089 May not be available on all supported platforms. **keep** is only available
1090 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1091 because ZFS doesn't support pre-allocation. Default: **native** if any
1092 pre-allocation methods are available, **none** if not.
1094 .. option:: fadvise_hint=str
1096 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1097 advise the kernel on what I/O patterns are likely to be issued.
1098 Accepted values are:
1101 Backwards-compatible hint for "no hint".
1104 Backwards compatible hint for "advise with fio workload type". This
1105 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1106 for a sequential workload.
1109 Advise using **FADV_SEQUENTIAL**.
1112 Advise using **FADV_RANDOM**.
1114 .. option:: write_hint=str
1116 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1117 from a write. Only supported on Linux, as of version 4.13. Accepted
1121 No particular life time associated with this file.
1124 Data written to this file has a short life time.
1127 Data written to this file has a medium life time.
1130 Data written to this file has a long life time.
1133 Data written to this file has a very long life time.
1135 The values are all relative to each other, and no absolute meaning
1136 should be associated with them.
1138 .. option:: offset=int
1140 Start I/O at the provided offset in the file, given as either a fixed size in
1141 bytes or a percentage. If a percentage is given, the generated offset will be
1142 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1143 provided. Data before the given offset will not be touched. This
1144 effectively caps the file size at `real_size - offset`. Can be combined with
1145 :option:`size` to constrain the start and end range of the I/O workload.
1146 A percentage can be specified by a number between 1 and 100 followed by '%',
1147 for example, ``offset=20%`` to specify 20%.
1149 .. option:: offset_align=int
1151 If set to non-zero value, the byte offset generated by a percentage ``offset``
1152 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1153 offset is aligned to the minimum block size.
1155 .. option:: offset_increment=int
1157 If this is provided, then the real offset becomes `offset + offset_increment
1158 * thread_number`, where the thread number is a counter that starts at 0 and
1159 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1160 specified). This option is useful if there are several jobs which are
1161 intended to operate on a file in parallel disjoint segments, with even
1162 spacing between the starting points.
1164 .. option:: number_ios=int
1166 Fio will normally perform I/Os until it has exhausted the size of the region
1167 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1168 condition). With this setting, the range/size can be set independently of
1169 the number of I/Os to perform. When fio reaches this number, it will exit
1170 normally and report status. Note that this does not extend the amount of I/O
1171 that will be done, it will only stop fio if this condition is met before
1172 other end-of-job criteria.
1174 .. option:: fsync=int
1176 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1177 the dirty data for every number of blocks given. For example, if you give 32
1178 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1179 using non-buffered I/O, we may not sync the file. The exception is the sg
1180 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1181 means fio does not periodically issue and wait for a sync to complete. Also
1182 see :option:`end_fsync` and :option:`fsync_on_close`.
1184 .. option:: fdatasync=int
1186 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1187 not metadata blocks. In Windows, FreeBSD, and DragonFlyBSD there is no
1188 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1189 Defaults to 0, which means fio does not periodically issue and wait for a
1190 data-only sync to complete.
1192 .. option:: write_barrier=int
1194 Make every `N-th` write a barrier write.
1196 .. option:: sync_file_range=str:int
1198 Use :manpage:`sync_file_range(2)` for every `int` number of write
1199 operations. Fio will track range of writes that have happened since the last
1200 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1203 SYNC_FILE_RANGE_WAIT_BEFORE
1205 SYNC_FILE_RANGE_WRITE
1207 SYNC_FILE_RANGE_WAIT_AFTER
1209 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1210 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1211 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1214 .. option:: overwrite=bool
1216 If true, writes to a file will always overwrite existing data. If the file
1217 doesn't already exist, it will be created before the write phase begins. If
1218 the file exists and is large enough for the specified write phase, nothing
1219 will be done. Default: false.
1221 .. option:: end_fsync=bool
1223 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1226 .. option:: fsync_on_close=bool
1228 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1229 from :option:`end_fsync` in that it will happen on every file close, not
1230 just at the end of the job. Default: false.
1232 .. option:: rwmixread=int
1234 Percentage of a mixed workload that should be reads. Default: 50.
1236 .. option:: rwmixwrite=int
1238 Percentage of a mixed workload that should be writes. If both
1239 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1240 add up to 100%, the latter of the two will be used to override the
1241 first. This may interfere with a given rate setting, if fio is asked to
1242 limit reads or writes to a certain rate. If that is the case, then the
1243 distribution may be skewed. Default: 50.
1245 .. option:: random_distribution=str:float[,str:float][,str:float]
1247 By default, fio will use a completely uniform random distribution when asked
1248 to perform random I/O. Sometimes it is useful to skew the distribution in
1249 specific ways, ensuring that some parts of the data is more hot than others.
1250 fio includes the following distribution models:
1253 Uniform random distribution
1262 Normal (Gaussian) distribution
1265 Zoned random distribution
1268 Zone absolute random distribution
1270 When using a **zipf** or **pareto** distribution, an input value is also
1271 needed to define the access pattern. For **zipf**, this is the `Zipf
1272 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1273 program, :command:`fio-genzipf`, that can be used visualize what the given input
1274 values will yield in terms of hit rates. If you wanted to use **zipf** with
1275 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1276 option. If a non-uniform model is used, fio will disable use of the random
1277 map. For the **normal** distribution, a normal (Gaussian) deviation is
1278 supplied as a value between 0 and 100.
1280 For a **zoned** distribution, fio supports specifying percentages of I/O
1281 access that should fall within what range of the file or device. For
1282 example, given a criteria of:
1284 * 60% of accesses should be to the first 10%
1285 * 30% of accesses should be to the next 20%
1286 * 8% of accesses should be to the next 30%
1287 * 2% of accesses should be to the next 40%
1289 we can define that through zoning of the random accesses. For the above
1290 example, the user would do::
1292 random_distribution=zoned:60/10:30/20:8/30:2/40
1294 A **zoned_abs** distribution works exactly like the **zoned**, except
1295 that it takes absolute sizes. For example, let's say you wanted to
1296 define access according to the following criteria:
1298 * 60% of accesses should be to the first 20G
1299 * 30% of accesses should be to the next 100G
1300 * 10% of accesses should be to the next 500G
1302 we can define an absolute zoning distribution with:
1304 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1306 For both **zoned** and **zoned_abs**, fio supports defining up to
1309 Similarly to how :option:`bssplit` works for setting ranges and
1310 percentages of block sizes. Like :option:`bssplit`, it's possible to
1311 specify separate zones for reads, writes, and trims. If just one set
1312 is given, it'll apply to all of them. This goes for both **zoned**
1313 **zoned_abs** distributions.
1315 .. option:: percentage_random=int[,int][,int]
1317 For a random workload, set how big a percentage should be random. This
1318 defaults to 100%, in which case the workload is fully random. It can be set
1319 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1320 sequential. Any setting in between will result in a random mix of sequential
1321 and random I/O, at the given percentages. Comma-separated values may be
1322 specified for reads, writes, and trims as described in :option:`blocksize`.
1324 .. option:: norandommap
1326 Normally fio will cover every block of the file when doing random I/O. If
1327 this option is given, fio will just get a new random offset without looking
1328 at past I/O history. This means that some blocks may not be read or written,
1329 and that some blocks may be read/written more than once. If this option is
1330 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1331 only intact blocks are verified, i.e., partially-overwritten blocks are
1332 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1333 the same block to be overwritten, which can cause verification errors. Either
1334 do not use norandommap in this case, or also use the lfsr random generator.
1336 .. option:: softrandommap=bool
1338 See :option:`norandommap`. If fio runs with the random block map enabled and
1339 it fails to allocate the map, if this option is set it will continue without
1340 a random block map. As coverage will not be as complete as with random maps,
1341 this option is disabled by default.
1343 .. option:: random_generator=str
1345 Fio supports the following engines for generating I/O offsets for random I/O:
1348 Strong 2^88 cycle random number generator.
1350 Linear feedback shift register generator.
1352 Strong 64-bit 2^258 cycle random number generator.
1354 **tausworthe** is a strong random number generator, but it requires tracking
1355 on the side if we want to ensure that blocks are only read or written
1356 once. **lfsr** guarantees that we never generate the same offset twice, and
1357 it's also less computationally expensive. It's not a true random generator,
1358 however, though for I/O purposes it's typically good enough. **lfsr** only
1359 works with single block sizes, not with workloads that use multiple block
1360 sizes. If used with such a workload, fio may read or write some blocks
1361 multiple times. The default value is **tausworthe**, unless the required
1362 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1363 selected automatically.
1369 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1371 The block size in bytes used for I/O units. Default: 4096. A single value
1372 applies to reads, writes, and trims. Comma-separated values may be
1373 specified for reads, writes, and trims. A value not terminated in a comma
1374 applies to subsequent types.
1379 means 256k for reads, writes and trims.
1382 means 8k for reads, 32k for writes and trims.
1385 means 8k for reads, 32k for writes, and default for trims.
1388 means default for reads, 8k for writes and trims.
1391 means default for reads, 8k for writes, and default for trims.
1393 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1395 A range of block sizes in bytes for I/O units. The issued I/O unit will
1396 always be a multiple of the minimum size, unless
1397 :option:`blocksize_unaligned` is set.
1399 Comma-separated ranges may be specified for reads, writes, and trims as
1400 described in :option:`blocksize`.
1402 Example: ``bsrange=1k-4k,2k-8k``.
1404 .. option:: bssplit=str[,str][,str]
1406 Sometimes you want even finer grained control of the block sizes
1407 issued, not just an even split between them. This option allows you to
1408 weight various block sizes, so that you are able to define a specific
1409 amount of block sizes issued. The format for this option is::
1411 bssplit=blocksize/percentage:blocksize/percentage
1413 for as many block sizes as needed. So if you want to define a workload
1414 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1417 bssplit=4k/10:64k/50:32k/40
1419 Ordering does not matter. If the percentage is left blank, fio will
1420 fill in the remaining values evenly. So a bssplit option like this one::
1422 bssplit=4k/50:1k/:32k/
1424 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1425 add up to 100, if bssplit is given a range that adds up to more, it
1428 Comma-separated values may be specified for reads, writes, and trims as
1429 described in :option:`blocksize`.
1431 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1432 having 90% 4k writes and 10% 8k writes, you would specify::
1434 bssplit=2k/50:4k/50,4k/90:8k/10
1436 Fio supports defining up to 64 different weights for each data
1439 .. option:: blocksize_unaligned, bs_unaligned
1441 If set, fio will issue I/O units with any size within
1442 :option:`blocksize_range`, not just multiples of the minimum size. This
1443 typically won't work with direct I/O, as that normally requires sector
1446 .. option:: bs_is_seq_rand=bool
1448 If this option is set, fio will use the normal read,write blocksize settings
1449 as sequential,random blocksize settings instead. Any random read or write
1450 will use the WRITE blocksize settings, and any sequential read or write will
1451 use the READ blocksize settings.
1453 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1455 Boundary to which fio will align random I/O units. Default:
1456 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1457 I/O, though it usually depends on the hardware block size. This option is
1458 mutually exclusive with using a random map for files, so it will turn off
1459 that option. Comma-separated values may be specified for reads, writes, and
1460 trims as described in :option:`blocksize`.
1466 .. option:: zero_buffers
1468 Initialize buffers with all zeros. Default: fill buffers with random data.
1470 .. option:: refill_buffers
1472 If this option is given, fio will refill the I/O buffers on every
1473 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1474 naturally. Defaults to being unset i.e., the buffer is only filled at
1475 init time and the data in it is reused when possible but if any of
1476 :option:`verify`, :option:`buffer_compress_percentage` or
1477 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1478 automatically enabled.
1480 .. option:: scramble_buffers=bool
1482 If :option:`refill_buffers` is too costly and the target is using data
1483 deduplication, then setting this option will slightly modify the I/O buffer
1484 contents to defeat normal de-dupe attempts. This is not enough to defeat
1485 more clever block compression attempts, but it will stop naive dedupe of
1486 blocks. Default: true.
1488 .. option:: buffer_compress_percentage=int
1490 If this is set, then fio will attempt to provide I/O buffer content
1491 (on WRITEs) that compresses to the specified level. Fio does this by
1492 providing a mix of random data followed by fixed pattern data. The
1493 fixed pattern is either zeros, or the pattern specified by
1494 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1495 might skew the compression ratio slightly. Setting
1496 `buffer_compress_percentage` to a value other than 100 will also
1497 enable :option:`refill_buffers` in order to reduce the likelihood that
1498 adjacent blocks are so similar that they over compress when seen
1499 together. See :option:`buffer_compress_chunk` for how to set a finer or
1500 coarser granularity for the random/fixed data region. Defaults to unset
1501 i.e., buffer data will not adhere to any compression level.
1503 .. option:: buffer_compress_chunk=int
1505 This setting allows fio to manage how big the random/fixed data region
1506 is when using :option:`buffer_compress_percentage`. When
1507 `buffer_compress_chunk` is set to some non-zero value smaller than the
1508 block size, fio can repeat the random/fixed region throughout the I/O
1509 buffer at the specified interval (which particularly useful when
1510 bigger block sizes are used for a job). When set to 0, fio will use a
1511 chunk size that matches the block size resulting in a single
1512 random/fixed region within the I/O buffer. Defaults to 512. When the
1513 unit is omitted, the value is interpreted in bytes.
1515 .. option:: buffer_pattern=str
1517 If set, fio will fill the I/O buffers with this pattern or with the contents
1518 of a file. If not set, the contents of I/O buffers are defined by the other
1519 options related to buffer contents. The setting can be any pattern of bytes,
1520 and can be prefixed with 0x for hex values. It may also be a string, where
1521 the string must then be wrapped with ``""``. Or it may also be a filename,
1522 where the filename must be wrapped with ``''`` in which case the file is
1523 opened and read. Note that not all the file contents will be read if that
1524 would cause the buffers to overflow. So, for example::
1526 buffer_pattern='filename'
1530 buffer_pattern="abcd"
1538 buffer_pattern=0xdeadface
1540 Also you can combine everything together in any order::
1542 buffer_pattern=0xdeadface"abcd"-12'filename'
1544 .. option:: dedupe_percentage=int
1546 If set, fio will generate this percentage of identical buffers when
1547 writing. These buffers will be naturally dedupable. The contents of the
1548 buffers depend on what other buffer compression settings have been set. It's
1549 possible to have the individual buffers either fully compressible, or not at
1550 all -- this option only controls the distribution of unique buffers. Setting
1551 this option will also enable :option:`refill_buffers` to prevent every buffer
1554 .. option:: invalidate=bool
1556 Invalidate the buffer/page cache parts of the files to be used prior to
1557 starting I/O if the platform and file type support it. Defaults to true.
1558 This will be ignored if :option:`pre_read` is also specified for the
1561 .. option:: sync=bool
1563 Use synchronous I/O for buffered writes. For the majority of I/O engines,
1564 this means using O_SYNC. Default: false.
1566 .. option:: iomem=str, mem=str
1568 Fio can use various types of memory as the I/O unit buffer. The allowed
1572 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1576 Use shared memory as the buffers. Allocated through
1577 :manpage:`shmget(2)`.
1580 Same as shm, but use huge pages as backing.
1583 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1584 be file backed if a filename is given after the option. The format
1585 is `mem=mmap:/path/to/file`.
1588 Use a memory mapped huge file as the buffer backing. Append filename
1589 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1592 Same as mmap, but use a MMAP_SHARED mapping.
1595 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1596 The :option:`ioengine` must be `rdma`.
1598 The area allocated is a function of the maximum allowed bs size for the job,
1599 multiplied by the I/O depth given. Note that for **shmhuge** and
1600 **mmaphuge** to work, the system must have free huge pages allocated. This
1601 can normally be checked and set by reading/writing
1602 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1603 is 4MiB in size. So to calculate the number of huge pages you need for a
1604 given job file, add up the I/O depth of all jobs (normally one unless
1605 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1606 that number by the huge page size. You can see the size of the huge pages in
1607 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1608 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1609 see :option:`hugepage-size`.
1611 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1612 should point there. So if it's mounted in :file:`/huge`, you would use
1613 `mem=mmaphuge:/huge/somefile`.
1615 .. option:: iomem_align=int, mem_align=int
1617 This indicates the memory alignment of the I/O memory buffers. Note that
1618 the given alignment is applied to the first I/O unit buffer, if using
1619 :option:`iodepth` the alignment of the following buffers are given by the
1620 :option:`bs` used. In other words, if using a :option:`bs` that is a
1621 multiple of the page sized in the system, all buffers will be aligned to
1622 this value. If using a :option:`bs` that is not page aligned, the alignment
1623 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1626 .. option:: hugepage-size=int
1628 Defines the size of a huge page. Must at least be equal to the system
1629 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1630 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1631 preferred way to set this to avoid setting a non-pow-2 bad value.
1633 .. option:: lockmem=int
1635 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1636 simulate a smaller amount of memory. The amount specified is per worker.
1642 .. option:: size=int
1644 The total size of file I/O for each thread of this job. Fio will run until
1645 this many bytes has been transferred, unless runtime is limited by other options
1646 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1647 Fio will divide this size between the available files determined by options
1648 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1649 specified by the job. If the result of division happens to be 0, the size is
1650 set to the physical size of the given files or devices if they exist.
1651 If this option is not specified, fio will use the full size of the given
1652 files or devices. If the files do not exist, size must be given. It is also
1653 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1654 given, fio will use 20% of the full size of the given files or devices.
1655 Can be combined with :option:`offset` to constrain the start and end range
1656 that I/O will be done within.
1658 .. option:: io_size=int, io_limit=int
1660 Normally fio operates within the region set by :option:`size`, which means
1661 that the :option:`size` option sets both the region and size of I/O to be
1662 performed. Sometimes that is not what you want. With this option, it is
1663 possible to define just the amount of I/O that fio should do. For instance,
1664 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1665 will perform I/O within the first 20GiB but exit when 5GiB have been
1666 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1667 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1668 the 0..20GiB region.
1670 .. option:: filesize=irange(int)
1672 Individual file sizes. May be a range, in which case fio will select sizes
1673 for files at random within the given range and limited to :option:`size` in
1674 total (if that is given). If not given, each created file is the same size.
1675 This option overrides :option:`size` in terms of file size, which means
1676 this value is used as a fixed size or possible range of each file.
1678 .. option:: file_append=bool
1680 Perform I/O after the end of the file. Normally fio will operate within the
1681 size of a file. If this option is set, then fio will append to the file
1682 instead. This has identical behavior to setting :option:`offset` to the size
1683 of a file. This option is ignored on non-regular files.
1685 .. option:: fill_device=bool, fill_fs=bool
1687 Sets size to something really large and waits for ENOSPC (no space left on
1688 device) as the terminating condition. Only makes sense with sequential
1689 write. For a read workload, the mount point will be filled first then I/O
1690 started on the result. This option doesn't make sense if operating on a raw
1691 device node, since the size of that is already known by the file system.
1692 Additionally, writing beyond end-of-device will not return ENOSPC there.
1698 .. option:: ioengine=str
1700 Defines how the job issues I/O to the file. The following types are defined:
1703 Basic :manpage:`read(2)` or :manpage:`write(2)`
1704 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1705 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1708 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1709 all supported operating systems except for Windows.
1712 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1713 queuing by coalescing adjacent I/Os into a single submission.
1716 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1719 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1722 Linux native asynchronous I/O. Note that Linux may only support
1723 queued behavior with non-buffered I/O (set ``direct=1`` or
1725 This engine defines engine specific options.
1728 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1729 :manpage:`aio_write(3)`.
1732 Solaris native asynchronous I/O.
1735 Windows native asynchronous I/O. Default on Windows.
1738 File is memory mapped with :manpage:`mmap(2)` and data copied
1739 to/from using :manpage:`memcpy(3)`.
1742 :manpage:`splice(2)` is used to transfer the data and
1743 :manpage:`vmsplice(2)` to transfer data from user space to the
1747 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1748 ioctl, or if the target is an sg character device we use
1749 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1750 I/O. Requires :option:`filename` option to specify either block or
1752 The sg engine includes engine specific options.
1755 Doesn't transfer any data, just pretends to. This is mainly used to
1756 exercise fio itself and for debugging/testing purposes.
1759 Transfer over the network to given ``host:port``. Depending on the
1760 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1761 :option:`listen` and :option:`filename` options are used to specify
1762 what sort of connection to make, while the :option:`protocol` option
1763 determines which protocol will be used. This engine defines engine
1767 Like **net**, but uses :manpage:`splice(2)` and
1768 :manpage:`vmsplice(2)` to map data and send/receive.
1769 This engine defines engine specific options.
1772 Doesn't transfer any data, but burns CPU cycles according to the
1773 :option:`cpuload` and :option:`cpuchunks` options. Setting
1774 :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
1775 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
1776 to get desired CPU usage, as the cpuload only loads a
1777 single CPU at the desired rate. A job never finishes unless there is
1778 at least one non-cpuio job.
1781 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall
1782 Interface approach to async I/O. See
1784 http://www.xmailserver.org/guasi-lib.html
1786 for more info on GUASI.
1789 The RDMA I/O engine supports both RDMA memory semantics
1790 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1791 InfiniBand, RoCE and iWARP protocols. This engine defines engine
1795 I/O engine that does regular fallocate to simulate data transfer as
1799 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1802 does fallocate(,mode = 0).
1805 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1808 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1809 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1810 size to the current block offset. :option:`blocksize` is ignored.
1813 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1814 defragment activity in request to DDIR_WRITE event.
1817 I/O engine supporting direct access to Ceph Reliable Autonomic
1818 Distributed Object Store (RADOS) via librados. This ioengine
1819 defines engine specific options.
1822 I/O engine supporting direct access to Ceph Rados Block Devices
1823 (RBD) via librbd without the need to use the kernel rbd driver. This
1824 ioengine defines engine specific options.
1827 Using GlusterFS libgfapi sync interface to direct access to
1828 GlusterFS volumes without having to go through FUSE. This ioengine
1829 defines engine specific options.
1832 Using GlusterFS libgfapi async interface to direct access to
1833 GlusterFS volumes without having to go through FUSE. This ioengine
1834 defines engine specific options.
1837 Read and write through Hadoop (HDFS). The :option:`filename` option
1838 is used to specify host,port of the hdfs name-node to connect. This
1839 engine interprets offsets a little differently. In HDFS, files once
1840 created cannot be modified so random writes are not possible. To
1841 imitate this the libhdfs engine expects a bunch of small files to be
1842 created over HDFS and will randomly pick a file from them
1843 based on the offset generated by fio backend (see the example
1844 job file to create such files, use ``rw=write`` option). Please
1845 note, it may be necessary to set environment variables to work
1846 with HDFS/libhdfs properly. Each job uses its own connection to
1850 Read, write and erase an MTD character device (e.g.,
1851 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
1852 underlying device type, the I/O may have to go in a certain pattern,
1853 e.g., on NAND, writing sequentially to erase blocks and discarding
1854 before overwriting. The `trimwrite` mode works well for this
1858 Read and write using filesystem DAX to a file on a filesystem
1859 mounted with DAX on a persistent memory device through the PMDK
1863 Read and write using device DAX to a persistent memory device (e.g.,
1864 /dev/dax0.0) through the PMDK libpmem library.
1867 Prefix to specify loading an external I/O engine object file. Append
1868 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
1869 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
1870 absolute or relative. See :file:`engines/skeleton_external.c` for
1871 details of writing an external I/O engine.
1874 Simply create the files and do no I/O to them. You still need to
1875 set `filesize` so that all the accounting still occurs, but no
1876 actual I/O will be done other than creating the file.
1879 Read and write using mmap I/O to a file on a filesystem
1880 mounted with DAX on a persistent memory device through the PMDK
1883 I/O engine specific parameters
1884 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1886 In addition, there are some parameters which are only valid when a specific
1887 :option:`ioengine` is in use. These are used identically to normal parameters,
1888 with the caveat that when used on the command line, they must come after the
1889 :option:`ioengine` that defines them is selected.
1891 .. option:: userspace_reap : [libaio]
1893 Normally, with the libaio engine in use, fio will use the
1894 :manpage:`io_getevents(2)` system call to reap newly returned events. With
1895 this flag turned on, the AIO ring will be read directly from user-space to
1896 reap events. The reaping mode is only enabled when polling for a minimum of
1897 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
1899 .. option:: hipri : [pvsync2]
1901 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
1904 .. option:: hipri_percentage : [pvsync2]
1906 When hipri is set this determines the probability of a pvsync2 I/O being high
1907 priority. The default is 100%.
1909 .. option:: cpuload=int : [cpuio]
1911 Attempt to use the specified percentage of CPU cycles. This is a mandatory
1912 option when using cpuio I/O engine.
1914 .. option:: cpuchunks=int : [cpuio]
1916 Split the load into cycles of the given time. In microseconds.
1918 .. option:: exit_on_io_done=bool : [cpuio]
1920 Detect when I/O threads are done, then exit.
1922 .. option:: namenode=str : [libhdfs]
1924 The hostname or IP address of a HDFS cluster namenode to contact.
1926 .. option:: port=int
1930 The listening port of the HFDS cluster namenode.
1934 The TCP or UDP port to bind to or connect to. If this is used with
1935 :option:`numjobs` to spawn multiple instances of the same job type, then
1936 this will be the starting port number since fio will use a range of
1941 The port to use for RDMA-CM communication. This should be the same value
1942 on the client and the server side.
1944 .. option:: hostname=str : [netsplice] [net] [rdma]
1946 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
1947 is a TCP listener or UDP reader, the hostname is not used and must be omitted
1948 unless it is a valid UDP multicast address.
1950 .. option:: interface=str : [netsplice] [net]
1952 The IP address of the network interface used to send or receive UDP
1955 .. option:: ttl=int : [netsplice] [net]
1957 Time-to-live value for outgoing UDP multicast packets. Default: 1.
1959 .. option:: nodelay=bool : [netsplice] [net]
1961 Set TCP_NODELAY on TCP connections.
1963 .. option:: protocol=str, proto=str : [netsplice] [net]
1965 The network protocol to use. Accepted values are:
1968 Transmission control protocol.
1970 Transmission control protocol V6.
1972 User datagram protocol.
1974 User datagram protocol V6.
1978 When the protocol is TCP or UDP, the port must also be given, as well as the
1979 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
1980 normal :option:`filename` option should be used and the port is invalid.
1982 .. option:: listen : [netsplice] [net]
1984 For TCP network connections, tell fio to listen for incoming connections
1985 rather than initiating an outgoing connection. The :option:`hostname` must
1986 be omitted if this option is used.
1988 .. option:: pingpong : [netsplice] [net]
1990 Normally a network writer will just continue writing data, and a network
1991 reader will just consume packages. If ``pingpong=1`` is set, a writer will
1992 send its normal payload to the reader, then wait for the reader to send the
1993 same payload back. This allows fio to measure network latencies. The
1994 submission and completion latencies then measure local time spent sending or
1995 receiving, and the completion latency measures how long it took for the
1996 other end to receive and send back. For UDP multicast traffic
1997 ``pingpong=1`` should only be set for a single reader when multiple readers
1998 are listening to the same address.
2000 .. option:: window_size : [netsplice] [net]
2002 Set the desired socket buffer size for the connection.
2004 .. option:: mss : [netsplice] [net]
2006 Set the TCP maximum segment size (TCP_MAXSEG).
2008 .. option:: donorname=str : [e4defrag]
2010 File will be used as a block donor (swap extents between files).
2012 .. option:: inplace=int : [e4defrag]
2014 Configure donor file blocks allocation strategy:
2017 Default. Preallocate donor's file on init.
2019 Allocate space immediately inside defragment event, and free right
2022 .. option:: clustername=str : [rbd,rados]
2024 Specifies the name of the Ceph cluster.
2026 .. option:: rbdname=str : [rbd]
2028 Specifies the name of the RBD.
2030 .. option:: pool=str : [rbd,rados]
2032 Specifies the name of the Ceph pool containing RBD or RADOS data.
2034 .. option:: clientname=str : [rbd,rados]
2036 Specifies the username (without the 'client.' prefix) used to access the
2037 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2038 the full *type.id* string. If no type. prefix is given, fio will add
2039 'client.' by default.
2041 .. option:: busy_poll=bool : [rbd,rados]
2043 Poll store instead of waiting for completion. Usually this provides better
2044 throughput at cost of higher(up to 100%) CPU utilization.
2046 .. option:: skip_bad=bool : [mtd]
2048 Skip operations against known bad blocks.
2050 .. option:: hdfsdirectory : [libhdfs]
2052 libhdfs will create chunk in this HDFS directory.
2054 .. option:: chunk_size : [libhdfs]
2056 The size of the chunk to use for each file.
2058 .. option:: verb=str : [rdma]
2060 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2061 values are write, read, send and recv. These correspond to the equivalent
2062 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2063 specified on the client side of the connection. See the examples folder.
2065 .. option:: bindname=str : [rdma]
2067 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2068 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2069 will be passed into the rdma_bind_addr() function and on the client site it
2070 will be used in the rdma_resolve_add() function. This can be useful when
2071 multiple paths exist between the client and the server or in certain loopback
2074 .. option:: readfua=bool : [sg]
2076 With readfua option set to 1, read operations include
2077 the force unit access (fua) flag. Default is 0.
2079 .. option:: writefua=bool : [sg]
2081 With writefua option set to 1, write operations include
2082 the force unit access (fua) flag. Default is 0.
2084 .. option:: sg_write_mode=str : [sg]
2085 Specify the type of write commands to issue. This option can take three values:
2088 This is the default where write opcodes are issued as usual.
2090 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2091 directs the device to carry out a medium verification with no data
2092 comparison. The writefua option is ignored with this selection.
2094 Issue WRITE SAME commands. This transfers a single block to the device
2095 and writes this same block of data to a contiguous sequence of LBAs
2096 beginning at the specified offset. fio's block size parameter specifies
2097 the amount of data written with each command. However, the amount of data
2098 actually transferred to the device is equal to the device's block
2099 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2100 write 16 sectors with each command. fio will still generate 8k of data
2101 for each command but only the first 512 bytes will be used and
2102 transferred to the device. The writefua option is ignored with this
2108 .. option:: iodepth=int
2110 Number of I/O units to keep in flight against the file. Note that
2111 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
2112 for small degrees when :option:`verify_async` is in use). Even async
2113 engines may impose OS restrictions causing the desired depth not to be
2114 achieved. This may happen on Linux when using libaio and not setting
2115 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
2116 eye on the I/O depth distribution in the fio output to verify that the
2117 achieved depth is as expected. Default: 1.
2119 .. option:: iodepth_batch_submit=int, iodepth_batch=int
2121 This defines how many pieces of I/O to submit at once. It defaults to 1
2122 which means that we submit each I/O as soon as it is available, but can be
2123 raised to submit bigger batches of I/O at the time. If it is set to 0 the
2124 :option:`iodepth` value will be used.
2126 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
2128 This defines how many pieces of I/O to retrieve at once. It defaults to 1
2129 which means that we'll ask for a minimum of 1 I/O in the retrieval process
2130 from the kernel. The I/O retrieval will go on until we hit the limit set by
2131 :option:`iodepth_low`. If this variable is set to 0, then fio will always
2132 check for completed events before queuing more I/O. This helps reduce I/O
2133 latency, at the cost of more retrieval system calls.
2135 .. option:: iodepth_batch_complete_max=int
2137 This defines maximum pieces of I/O to retrieve at once. This variable should
2138 be used along with :option:`iodepth_batch_complete_min`\=int variable,
2139 specifying the range of min and max amount of I/O which should be
2140 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
2145 iodepth_batch_complete_min=1
2146 iodepth_batch_complete_max=<iodepth>
2148 which means that we will retrieve at least 1 I/O and up to the whole
2149 submitted queue depth. If none of I/O has been completed yet, we will wait.
2153 iodepth_batch_complete_min=0
2154 iodepth_batch_complete_max=<iodepth>
2156 which means that we can retrieve up to the whole submitted queue depth, but
2157 if none of I/O has been completed yet, we will NOT wait and immediately exit
2158 the system call. In this example we simply do polling.
2160 .. option:: iodepth_low=int
2162 The low water mark indicating when to start filling the queue
2163 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2164 attempt to keep the queue full at all times. If :option:`iodepth` is set to
2165 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
2166 16 requests, it will let the depth drain down to 4 before starting to fill
2169 .. option:: serialize_overlap=bool
2171 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2172 When two or more I/Os are submitted simultaneously, there is no guarantee that
2173 the I/Os will be processed or completed in the submitted order. Further, if
2174 two or more of those I/Os are writes, any overlapping region between them can
2175 become indeterminate/undefined on certain storage. These issues can cause
2176 verification to fail erratically when at least one of the racing I/Os is
2177 changing data and the overlapping region has a non-zero size. Setting
2178 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2179 serializing in-flight I/Os that have a non-zero overlap. Note that setting
2180 this option can reduce both performance and the :option:`iodepth` achieved.
2181 Additionally this option does not work when :option:`io_submit_mode` is set to
2182 offload. Default: false.
2184 .. option:: io_submit_mode=str
2186 This option controls how fio submits the I/O to the I/O engine. The default
2187 is `inline`, which means that the fio job threads submit and reap I/O
2188 directly. If set to `offload`, the job threads will offload I/O submission
2189 to a dedicated pool of I/O threads. This requires some coordination and thus
2190 has a bit of extra overhead, especially for lower queue depth I/O where it
2191 can increase latencies. The benefit is that fio can manage submission rates
2192 independently of the device completion rates. This avoids skewed latency
2193 reporting if I/O gets backed up on the device side (the coordinated omission
2200 .. option:: thinktime=time
2202 Stall the job for the specified period of time after an I/O has completed before issuing the
2203 next. May be used to simulate processing being done by an application.
2204 When the unit is omitted, the value is interpreted in microseconds. See
2205 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2207 .. option:: thinktime_spin=time
2209 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2210 something with the data received, before falling back to sleeping for the
2211 rest of the period specified by :option:`thinktime`. When the unit is
2212 omitted, the value is interpreted in microseconds.
2214 .. option:: thinktime_blocks=int
2216 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2217 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
2218 fio wait :option:`thinktime` usecs after every block. This effectively makes any
2219 queue depth setting redundant, since no more than 1 I/O will be queued
2220 before we have to complete it and do our :option:`thinktime`. In other words, this
2221 setting effectively caps the queue depth if the latter is larger.
2223 .. option:: rate=int[,int][,int]
2225 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2226 suffix rules apply. Comma-separated values may be specified for reads,
2227 writes, and trims as described in :option:`blocksize`.
2229 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2230 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2231 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2232 latter will only limit reads.
2234 .. option:: rate_min=int[,int][,int]
2236 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2237 to meet this requirement will cause the job to exit. Comma-separated values
2238 may be specified for reads, writes, and trims as described in
2239 :option:`blocksize`.
2241 .. option:: rate_iops=int[,int][,int]
2243 Cap the bandwidth to this number of IOPS. Basically the same as
2244 :option:`rate`, just specified independently of bandwidth. If the job is
2245 given a block size range instead of a fixed value, the smallest block size
2246 is used as the metric. Comma-separated values may be specified for reads,
2247 writes, and trims as described in :option:`blocksize`.
2249 .. option:: rate_iops_min=int[,int][,int]
2251 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2252 Comma-separated values may be specified for reads, writes, and trims as
2253 described in :option:`blocksize`.
2255 .. option:: rate_process=str
2257 This option controls how fio manages rated I/O submissions. The default is
2258 `linear`, which submits I/O in a linear fashion with fixed delays between
2259 I/Os that gets adjusted based on I/O completion rates. If this is set to
2260 `poisson`, fio will submit I/O based on a more real world random request
2261 flow, known as the Poisson process
2262 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2263 10^6 / IOPS for the given workload.
2265 .. option:: rate_ignore_thinktime=bool
2267 By default, fio will attempt to catch up to the specified rate setting,
2268 if any kind of thinktime setting was used. If this option is set, then
2269 fio will ignore the thinktime and continue doing IO at the specified
2270 rate, instead of entering a catch-up mode after thinktime is done.
2276 .. option:: latency_target=time
2278 If set, fio will attempt to find the max performance point that the given
2279 workload will run at while maintaining a latency below this target. When
2280 the unit is omitted, the value is interpreted in microseconds. See
2281 :option:`latency_window` and :option:`latency_percentile`.
2283 .. option:: latency_window=time
2285 Used with :option:`latency_target` to specify the sample window that the job
2286 is run at varying queue depths to test the performance. When the unit is
2287 omitted, the value is interpreted in microseconds.
2289 .. option:: latency_percentile=float
2291 The percentage of I/Os that must fall within the criteria specified by
2292 :option:`latency_target` and :option:`latency_window`. If not set, this
2293 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2294 set by :option:`latency_target`.
2296 .. option:: max_latency=time
2298 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2299 maximum latency. When the unit is omitted, the value is interpreted in
2302 .. option:: rate_cycle=int
2304 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2305 of milliseconds. Defaults to 1000.
2311 .. option:: write_iolog=str
2313 Write the issued I/O patterns to the specified file. See
2314 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2315 iologs will be interspersed and the file may be corrupt.
2317 .. option:: read_iolog=str
2319 Open an iolog with the specified filename and replay the I/O patterns it
2320 contains. This can be used to store a workload and replay it sometime
2321 later. The iolog given may also be a blktrace binary file, which allows fio
2322 to replay a workload captured by :command:`blktrace`. See
2323 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2324 replay, the file needs to be turned into a blkparse binary data file first
2325 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2327 .. option:: read_iolog_chunked=bool
2329 Determines how iolog is read. If false(default) entire :option:`read_iolog`
2330 will be read at once. If selected true, input from iolog will be read
2331 gradually. Useful when iolog is very large, or it is generated.
2333 .. option:: replay_no_stall=bool
2335 When replaying I/O with :option:`read_iolog` the default behavior is to
2336 attempt to respect the timestamps within the log and replay them with the
2337 appropriate delay between IOPS. By setting this variable fio will not
2338 respect the timestamps and attempt to replay them as fast as possible while
2339 still respecting ordering. The result is the same I/O pattern to a given
2340 device, but different timings.
2342 .. option:: replay_time_scale=int
2344 When replaying I/O with :option:`read_iolog`, fio will honor the
2345 original timing in the trace. With this option, it's possible to scale
2346 the time. It's a percentage option, if set to 50 it means run at 50%
2347 the original IO rate in the trace. If set to 200, run at twice the
2348 original IO rate. Defaults to 100.
2350 .. option:: replay_redirect=str
2352 While replaying I/O patterns using :option:`read_iolog` the default behavior
2353 is to replay the IOPS onto the major/minor device that each IOP was recorded
2354 from. This is sometimes undesirable because on a different machine those
2355 major/minor numbers can map to a different device. Changing hardware on the
2356 same system can also result in a different major/minor mapping.
2357 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
2358 device regardless of the device it was recorded
2359 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
2360 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2361 multiple devices will be replayed onto a single device, if the trace
2362 contains multiple devices. If you want multiple devices to be replayed
2363 concurrently to multiple redirected devices you must blkparse your trace
2364 into separate traces and replay them with independent fio invocations.
2365 Unfortunately this also breaks the strict time ordering between multiple
2368 .. option:: replay_align=int
2370 Force alignment of I/O offsets and lengths in a trace to this power of 2
2373 .. option:: replay_scale=int
2375 Scale sector offsets down by this factor when replaying traces.
2377 .. option:: replay_skip=str
2379 Sometimes it's useful to skip certain IO types in a replay trace.
2380 This could be, for instance, eliminating the writes in the trace.
2381 Or not replaying the trims/discards, if you are redirecting to
2382 a device that doesn't support them. This option takes a comma
2383 separated list of read, write, trim, sync.
2386 Threads, processes and job synchronization
2387 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2391 Fio defaults to creating jobs by using fork, however if this option is
2392 given, fio will create jobs by using POSIX Threads' function
2393 :manpage:`pthread_create(3)` to create threads instead.
2395 .. option:: wait_for=str
2397 If set, the current job won't be started until all workers of the specified
2398 waitee job are done.
2400 ``wait_for`` operates on the job name basis, so there are a few
2401 limitations. First, the waitee must be defined prior to the waiter job
2402 (meaning no forward references). Second, if a job is being referenced as a
2403 waitee, it must have a unique name (no duplicate waitees).
2405 .. option:: nice=int
2407 Run the job with the given nice value. See man :manpage:`nice(2)`.
2409 On Windows, values less than -15 set the process class to "High"; -1 through
2410 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2413 .. option:: prio=int
2415 Set the I/O priority value of this job. Linux limits us to a positive value
2416 between 0 and 7, with 0 being the highest. See man
2417 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2418 systems since meaning of priority may differ.
2420 .. option:: prioclass=int
2422 Set the I/O priority class. See man :manpage:`ionice(1)`.
2424 .. option:: cpus_allowed=str
2426 Controls the same options as :option:`cpumask`, but accepts a textual
2427 specification of the permitted CPUs instead and CPUs are indexed from 0. So
2428 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
2429 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
2430 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
2432 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
2433 processor group will be used and affinity settings are inherited from the
2434 system. An fio build configured to target Windows 7 makes options that set
2435 CPUs processor group aware and values will set both the processor group
2436 and a CPU from within that group. For example, on a system where processor
2437 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
2438 values between 0 and 39 will bind CPUs from processor group 0 and
2439 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
2440 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
2441 single ``cpus_allowed`` option must be from the same processor group. For
2442 Windows fio builds not built for Windows 7, CPUs will only be selected from
2443 (and be relative to) whatever processor group fio happens to be running in
2444 and CPUs from other processor groups cannot be used.
2446 .. option:: cpus_allowed_policy=str
2448 Set the policy of how fio distributes the CPUs specified by
2449 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
2452 All jobs will share the CPU set specified.
2454 Each job will get a unique CPU from the CPU set.
2456 **shared** is the default behavior, if the option isn't specified. If
2457 **split** is specified, then fio will will assign one cpu per job. If not
2458 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2461 .. option:: cpumask=int
2463 Set the CPU affinity of this job. The parameter given is a bit mask of
2464 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2465 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2466 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2467 operating systems or kernel versions. This option doesn't work well for a
2468 higher CPU count than what you can store in an integer mask, so it can only
2469 control cpus 1-32. For boxes with larger CPU counts, use
2470 :option:`cpus_allowed`.
2472 .. option:: numa_cpu_nodes=str
2474 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2475 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2476 NUMA options support, fio must be built on a system with libnuma-dev(el)
2479 .. option:: numa_mem_policy=str
2481 Set this job's memory policy and corresponding NUMA nodes. Format of the
2486 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
2487 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2488 policies, no node needs to be specified. For ``prefer``, only one node is
2489 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2490 follows: a comma delimited list of numbers, A-B ranges, or `all`.
2492 .. option:: cgroup=str
2494 Add job to this control group. If it doesn't exist, it will be created. The
2495 system must have a mounted cgroup blkio mount point for this to work. If
2496 your system doesn't have it mounted, you can do so with::
2498 # mount -t cgroup -o blkio none /cgroup
2500 .. option:: cgroup_weight=int
2502 Set the weight of the cgroup to this value. See the documentation that comes
2503 with the kernel, allowed values are in the range of 100..1000.
2505 .. option:: cgroup_nodelete=bool
2507 Normally fio will delete the cgroups it has created after the job
2508 completion. To override this behavior and to leave cgroups around after the
2509 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2510 to inspect various cgroup files after job completion. Default: false.
2512 .. option:: flow_id=int
2514 The ID of the flow. If not specified, it defaults to being a global
2515 flow. See :option:`flow`.
2517 .. option:: flow=int
2519 Weight in token-based flow control. If this value is used, then there is a
2520 'flow counter' which is used to regulate the proportion of activity between
2521 two or more jobs. Fio attempts to keep this flow counter near zero. The
2522 ``flow`` parameter stands for how much should be added or subtracted to the
2523 flow counter on each iteration of the main I/O loop. That is, if one job has
2524 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2525 ratio in how much one runs vs the other.
2527 .. option:: flow_watermark=int
2529 The maximum value that the absolute value of the flow counter is allowed to
2530 reach before the job must wait for a lower value of the counter.
2532 .. option:: flow_sleep=int
2534 The period of time, in microseconds, to wait after the flow watermark has
2535 been exceeded before retrying operations.
2537 .. option:: stonewall, wait_for_previous
2539 Wait for preceding jobs in the job file to exit, before starting this
2540 one. Can be used to insert serialization points in the job file. A stone
2541 wall also implies starting a new reporting group, see
2542 :option:`group_reporting`.
2546 By default, fio will continue running all other jobs when one job finishes
2547 but sometimes this is not the desired action. Setting ``exitall`` will
2548 instead make fio terminate all other jobs when one job finishes.
2550 .. option:: exec_prerun=str
2552 Before running this job, issue the command specified through
2553 :manpage:`system(3)`. Output is redirected in a file called
2554 :file:`jobname.prerun.txt`.
2556 .. option:: exec_postrun=str
2558 After the job completes, issue the command specified though
2559 :manpage:`system(3)`. Output is redirected in a file called
2560 :file:`jobname.postrun.txt`.
2564 Instead of running as the invoking user, set the user ID to this value
2565 before the thread/process does any work.
2569 Set group ID, see :option:`uid`.
2575 .. option:: verify_only
2577 Do not perform specified workload, only verify data still matches previous
2578 invocation of this workload. This option allows one to check data multiple
2579 times at a later date without overwriting it. This option makes sense only
2580 for workloads that write data, and does not support workloads with the
2581 :option:`time_based` option set.
2583 .. option:: do_verify=bool
2585 Run the verify phase after a write phase. Only valid if :option:`verify` is
2588 .. option:: verify=str
2590 If writing to a file, fio can verify the file contents after each iteration
2591 of the job. Each verification method also implies verification of special
2592 header, which is written to the beginning of each block. This header also
2593 includes meta information, like offset of the block, block number, timestamp
2594 when block was written, etc. :option:`verify` can be combined with
2595 :option:`verify_pattern` option. The allowed values are:
2598 Use an md5 sum of the data area and store it in the header of
2602 Use an experimental crc64 sum of the data area and store it in the
2603 header of each block.
2606 Use a crc32c sum of the data area and store it in the header of
2607 each block. This will automatically use hardware acceleration
2608 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
2609 fall back to software crc32c if none is found. Generally the
2610 fastest checksum fio supports when hardware accelerated.
2616 Use a crc32 sum of the data area and store it in the header of each
2620 Use a crc16 sum of the data area and store it in the header of each
2624 Use a crc7 sum of the data area and store it in the header of each
2628 Use xxhash as the checksum function. Generally the fastest software
2629 checksum that fio supports.
2632 Use sha512 as the checksum function.
2635 Use sha256 as the checksum function.
2638 Use optimized sha1 as the checksum function.
2641 Use optimized sha3-224 as the checksum function.
2644 Use optimized sha3-256 as the checksum function.
2647 Use optimized sha3-384 as the checksum function.
2650 Use optimized sha3-512 as the checksum function.
2653 This option is deprecated, since now meta information is included in
2654 generic verification header and meta verification happens by
2655 default. For detailed information see the description of the
2656 :option:`verify` setting. This option is kept because of
2657 compatibility's sake with old configurations. Do not use it.
2660 Verify a strict pattern. Normally fio includes a header with some
2661 basic information and checksumming, but if this option is set, only
2662 the specific pattern set with :option:`verify_pattern` is verified.
2665 Only pretend to verify. Useful for testing internals with
2666 :option:`ioengine`\=null, not for much else.
2668 This option can be used for repeated burn-in tests of a system to make sure
2669 that the written data is also correctly read back. If the data direction
2670 given is a read or random read, fio will assume that it should verify a
2671 previously written file. If the data direction includes any form of write,
2672 the verify will be of the newly written data.
2674 To avoid false verification errors, do not use the norandommap option when
2675 verifying data with async I/O engines and I/O depths > 1. Or use the
2676 norandommap and the lfsr random generator together to avoid writing to the
2677 same offset with muliple outstanding I/Os.
2679 .. option:: verify_offset=int
2681 Swap the verification header with data somewhere else in the block before
2682 writing. It is swapped back before verifying.
2684 .. option:: verify_interval=int
2686 Write the verification header at a finer granularity than the
2687 :option:`blocksize`. It will be written for chunks the size of
2688 ``verify_interval``. :option:`blocksize` should divide this evenly.
2690 .. option:: verify_pattern=str
2692 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
2693 filling with totally random bytes, but sometimes it's interesting to fill
2694 with a known pattern for I/O verification purposes. Depending on the width
2695 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
2696 be either a decimal or a hex number). The ``verify_pattern`` if larger than
2697 a 32-bit quantity has to be a hex number that starts with either "0x" or
2698 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
2699 format, which means that for each block offset will be written and then
2700 verified back, e.g.::
2704 Or use combination of everything::
2706 verify_pattern=0xff%o"abcd"-12
2708 .. option:: verify_fatal=bool
2710 Normally fio will keep checking the entire contents before quitting on a
2711 block verification failure. If this option is set, fio will exit the job on
2712 the first observed failure. Default: false.
2714 .. option:: verify_dump=bool
2716 If set, dump the contents of both the original data block and the data block
2717 we read off disk to files. This allows later analysis to inspect just what
2718 kind of data corruption occurred. Off by default.
2720 .. option:: verify_async=int
2722 Fio will normally verify I/O inline from the submitting thread. This option
2723 takes an integer describing how many async offload threads to create for I/O
2724 verification instead, causing fio to offload the duty of verifying I/O
2725 contents to one or more separate threads. If using this offload option, even
2726 sync I/O engines can benefit from using an :option:`iodepth` setting higher
2727 than 1, as it allows them to have I/O in flight while verifies are running.
2728 Defaults to 0 async threads, i.e. verification is not asynchronous.
2730 .. option:: verify_async_cpus=str
2732 Tell fio to set the given CPU affinity on the async I/O verification
2733 threads. See :option:`cpus_allowed` for the format used.
2735 .. option:: verify_backlog=int
2737 Fio will normally verify the written contents of a job that utilizes verify
2738 once that job has completed. In other words, everything is written then
2739 everything is read back and verified. You may want to verify continually
2740 instead for a variety of reasons. Fio stores the meta data associated with
2741 an I/O block in memory, so for large verify workloads, quite a bit of memory
2742 would be used up holding this meta data. If this option is enabled, fio will
2743 write only N blocks before verifying these blocks.
2745 .. option:: verify_backlog_batch=int
2747 Control how many blocks fio will verify if :option:`verify_backlog` is
2748 set. If not set, will default to the value of :option:`verify_backlog`
2749 (meaning the entire queue is read back and verified). If
2750 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
2751 blocks will be verified, if ``verify_backlog_batch`` is larger than
2752 :option:`verify_backlog`, some blocks will be verified more than once.
2754 .. option:: verify_state_save=bool
2756 When a job exits during the write phase of a verify workload, save its
2757 current state. This allows fio to replay up until that point, if the verify
2758 state is loaded for the verify read phase. The format of the filename is,
2761 <type>-<jobname>-<jobindex>-verify.state.
2763 <type> is "local" for a local run, "sock" for a client/server socket
2764 connection, and "ip" (192.168.0.1, for instance) for a networked
2765 client/server connection. Defaults to true.
2767 .. option:: verify_state_load=bool
2769 If a verify termination trigger was used, fio stores the current write state
2770 of each thread. This can be used at verification time so that fio knows how
2771 far it should verify. Without this information, fio will run a full
2772 verification pass, according to the settings in the job file used. Default
2775 .. option:: trim_percentage=int
2777 Number of verify blocks to discard/trim.
2779 .. option:: trim_verify_zero=bool
2781 Verify that trim/discarded blocks are returned as zeros.
2783 .. option:: trim_backlog=int
2785 Trim after this number of blocks are written.
2787 .. option:: trim_backlog_batch=int
2789 Trim this number of I/O blocks.
2791 .. option:: experimental_verify=bool
2793 Enable experimental verification.
2798 .. option:: steadystate=str:float, ss=str:float
2800 Define the criterion and limit for assessing steady state performance. The
2801 first parameter designates the criterion whereas the second parameter sets
2802 the threshold. When the criterion falls below the threshold for the
2803 specified duration, the job will stop. For example, `iops_slope:0.1%` will
2804 direct fio to terminate the job when the least squares regression slope
2805 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
2806 this will apply to all jobs in the group. Below is the list of available
2807 steady state assessment criteria. All assessments are carried out using only
2808 data from the rolling collection window. Threshold limits can be expressed
2809 as a fixed value or as a percentage of the mean in the collection window.
2812 Collect IOPS data. Stop the job if all individual IOPS measurements
2813 are within the specified limit of the mean IOPS (e.g., ``iops:2``
2814 means that all individual IOPS values must be within 2 of the mean,
2815 whereas ``iops:0.2%`` means that all individual IOPS values must be
2816 within 0.2% of the mean IOPS to terminate the job).
2819 Collect IOPS data and calculate the least squares regression
2820 slope. Stop the job if the slope falls below the specified limit.
2823 Collect bandwidth data. Stop the job if all individual bandwidth
2824 measurements are within the specified limit of the mean bandwidth.
2827 Collect bandwidth data and calculate the least squares regression
2828 slope. Stop the job if the slope falls below the specified limit.
2830 .. option:: steadystate_duration=time, ss_dur=time
2832 A rolling window of this duration will be used to judge whether steady state
2833 has been reached. Data will be collected once per second. The default is 0
2834 which disables steady state detection. When the unit is omitted, the
2835 value is interpreted in seconds.
2837 .. option:: steadystate_ramp_time=time, ss_ramp=time
2839 Allow the job to run for the specified duration before beginning data
2840 collection for checking the steady state job termination criterion. The
2841 default is 0. When the unit is omitted, the value is interpreted in seconds.
2844 Measurements and reporting
2845 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2847 .. option:: per_job_logs=bool
2849 If set, this generates bw/clat/iops log with per file private filenames. If
2850 not set, jobs with identical names will share the log filename. Default:
2853 .. option:: group_reporting
2855 It may sometimes be interesting to display statistics for groups of jobs as
2856 a whole instead of for each individual job. This is especially true if
2857 :option:`numjobs` is used; looking at individual thread/process output
2858 quickly becomes unwieldy. To see the final report per-group instead of
2859 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
2860 same reporting group, unless if separated by a :option:`stonewall`, or by
2861 using :option:`new_group`.
2863 .. option:: new_group
2865 Start a new reporting group. See: :option:`group_reporting`. If not given,
2866 all jobs in a file will be part of the same reporting group, unless
2867 separated by a :option:`stonewall`.
2869 .. option:: stats=bool
2871 By default, fio collects and shows final output results for all jobs
2872 that run. If this option is set to 0, then fio will ignore it in
2873 the final stat output.
2875 .. option:: write_bw_log=str
2877 If given, write a bandwidth log for this job. Can be used to store data of
2878 the bandwidth of the jobs in their lifetime.
2880 If no str argument is given, the default filename of
2881 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
2882 will still append the type of log. So if one specifies::
2886 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
2887 of the job (`1..N`, where `N` is the number of jobs). If
2888 :option:`per_job_logs` is false, then the filename will not include the
2891 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
2892 text files into nice graphs. See `Log File Formats`_ for how data is
2893 structured within the file.
2895 .. option:: write_lat_log=str
2897 Same as :option:`write_bw_log`, except this option creates I/O
2898 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
2899 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
2900 latency files instead. See :option:`write_bw_log` for details about
2901 the filename format and `Log File Formats`_ for how data is structured
2904 .. option:: write_hist_log=str
2906 Same as :option:`write_bw_log` but writes an I/O completion latency
2907 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
2908 file will be empty unless :option:`log_hist_msec` has also been set.
2909 See :option:`write_bw_log` for details about the filename format and
2910 `Log File Formats`_ for how data is structured within the file.
2912 .. option:: write_iops_log=str
2914 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
2915 :file:`name_iops.x.log`) instead. See :option:`write_bw_log` for
2916 details about the filename format and `Log File Formats`_ for how data
2917 is structured within the file.
2919 .. option:: log_avg_msec=int
2921 By default, fio will log an entry in the iops, latency, or bw log for every
2922 I/O that completes. When writing to the disk log, that can quickly grow to a
2923 very large size. Setting this option makes fio average the each log entry
2924 over the specified period of time, reducing the resolution of the log. See
2925 :option:`log_max_value` as well. Defaults to 0, logging all entries.
2926 Also see `Log File Formats`_.
2928 .. option:: log_hist_msec=int
2930 Same as :option:`log_avg_msec`, but logs entries for completion latency
2931 histograms. Computing latency percentiles from averages of intervals using
2932 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
2933 histogram entries over the specified period of time, reducing log sizes for
2934 high IOPS devices while retaining percentile accuracy. See
2935 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
2936 Defaults to 0, meaning histogram logging is disabled.
2938 .. option:: log_hist_coarseness=int
2940 Integer ranging from 0 to 6, defining the coarseness of the resolution of
2941 the histogram logs enabled with :option:`log_hist_msec`. For each increment
2942 in coarseness, fio outputs half as many bins. Defaults to 0, for which
2943 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
2944 and `Log File Formats`_.
2946 .. option:: log_max_value=bool
2948 If :option:`log_avg_msec` is set, fio logs the average over that window. If
2949 you instead want to log the maximum value, set this option to 1. Defaults to
2950 0, meaning that averaged values are logged.
2952 .. option:: log_offset=bool
2954 If this is set, the iolog options will include the byte offset for the I/O
2955 entry as well as the other data values. Defaults to 0 meaning that
2956 offsets are not present in logs. Also see `Log File Formats`_.
2958 .. option:: log_compression=int
2960 If this is set, fio will compress the I/O logs as it goes, to keep the
2961 memory footprint lower. When a log reaches the specified size, that chunk is
2962 removed and compressed in the background. Given that I/O logs are fairly
2963 highly compressible, this yields a nice memory savings for longer runs. The
2964 downside is that the compression will consume some background CPU cycles, so
2965 it may impact the run. This, however, is also true if the logging ends up
2966 consuming most of the system memory. So pick your poison. The I/O logs are
2967 saved normally at the end of a run, by decompressing the chunks and storing
2968 them in the specified log file. This feature depends on the availability of
2971 .. option:: log_compression_cpus=str
2973 Define the set of CPUs that are allowed to handle online log compression for
2974 the I/O jobs. This can provide better isolation between performance
2975 sensitive jobs, and background compression work. See
2976 :option:`cpus_allowed` for the format used.
2978 .. option:: log_store_compressed=bool
2980 If set, fio will store the log files in a compressed format. They can be
2981 decompressed with fio, using the :option:`--inflate-log` command line
2982 parameter. The files will be stored with a :file:`.fz` suffix.
2984 .. option:: log_unix_epoch=bool
2986 If set, fio will log Unix timestamps to the log files produced by enabling
2987 write_type_log for each log type, instead of the default zero-based
2990 .. option:: block_error_percentiles=bool
2992 If set, record errors in trim block-sized units from writes and trims and
2993 output a histogram of how many trims it took to get to errors, and what kind
2994 of error was encountered.
2996 .. option:: bwavgtime=int
2998 Average the calculated bandwidth over the given time. Value is specified in
2999 milliseconds. If the job also does bandwidth logging through
3000 :option:`write_bw_log`, then the minimum of this option and
3001 :option:`log_avg_msec` will be used. Default: 500ms.
3003 .. option:: iopsavgtime=int
3005 Average the calculated IOPS over the given time. Value is specified in
3006 milliseconds. If the job also does IOPS logging through
3007 :option:`write_iops_log`, then the minimum of this option and
3008 :option:`log_avg_msec` will be used. Default: 500ms.
3010 .. option:: disk_util=bool
3012 Generate disk utilization statistics, if the platform supports it.
3015 .. option:: disable_lat=bool
3017 Disable measurements of total latency numbers. Useful only for cutting back
3018 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
3019 performance at really high IOPS rates. Note that to really get rid of a
3020 large amount of these calls, this option must be used with
3021 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
3023 .. option:: disable_clat=bool
3025 Disable measurements of completion latency numbers. See
3026 :option:`disable_lat`.
3028 .. option:: disable_slat=bool
3030 Disable measurements of submission latency numbers. See
3031 :option:`disable_lat`.
3033 .. option:: disable_bw_measurement=bool, disable_bw=bool
3035 Disable measurements of throughput/bandwidth numbers. See
3036 :option:`disable_lat`.
3038 .. option:: clat_percentiles=bool
3040 Enable the reporting of percentiles of completion latencies. This
3041 option is mutually exclusive with :option:`lat_percentiles`.
3043 .. option:: lat_percentiles=bool
3045 Enable the reporting of percentiles of I/O latencies. This is similar
3046 to :option:`clat_percentiles`, except that this includes the
3047 submission latency. This option is mutually exclusive with
3048 :option:`clat_percentiles`.
3050 .. option:: percentile_list=float_list
3052 Overwrite the default list of percentiles for completion latencies and
3053 the block error histogram. Each number is a floating number in the
3054 range (0,100], and the maximum length of the list is 20. Use ``:`` to
3055 separate the numbers, and list the numbers in ascending order. For
3056 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
3057 values of completion latency below which 99.5% and 99.9% of the observed
3058 latencies fell, respectively.
3060 .. option:: significant_figures=int
3062 If using :option:`--output-format` of `normal`, set the significant
3063 figures to this value. Higher values will yield more precise IOPS and
3064 throughput units, while lower values will round. Requires a minimum
3065 value of 1 and a maximum value of 10. Defaults to 4.
3071 .. option:: exitall_on_error
3073 When one job finishes in error, terminate the rest. The default is to wait
3074 for each job to finish.
3076 .. option:: continue_on_error=str
3078 Normally fio will exit the job on the first observed failure. If this option
3079 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
3080 EILSEQ) until the runtime is exceeded or the I/O size specified is
3081 completed. If this option is used, there are two more stats that are
3082 appended, the total error count and the first error. The error field given
3083 in the stats is the first error that was hit during the run.
3085 The allowed values are:
3088 Exit on any I/O or verify errors.
3091 Continue on read errors, exit on all others.
3094 Continue on write errors, exit on all others.
3097 Continue on any I/O error, exit on all others.
3100 Continue on verify errors, exit on all others.
3103 Continue on all errors.
3106 Backward-compatible alias for 'none'.
3109 Backward-compatible alias for 'all'.
3111 .. option:: ignore_error=str
3113 Sometimes you want to ignore some errors during test in that case you can
3114 specify error list for each error type, instead of only being able to
3115 ignore the default 'non-fatal error' using :option:`continue_on_error`.
3116 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
3117 given error type is separated with ':'. Error may be symbol ('ENOSPC',
3118 'ENOMEM') or integer. Example::
3120 ignore_error=EAGAIN,ENOSPC:122
3122 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
3123 WRITE. This option works by overriding :option:`continue_on_error` with
3124 the list of errors for each error type if any.
3126 .. option:: error_dump=bool
3128 If set dump every error even if it is non fatal, true by default. If
3129 disabled only fatal error will be dumped.
3131 Running predefined workloads
3132 ----------------------------
3134 Fio includes predefined profiles that mimic the I/O workloads generated by
3137 .. option:: profile=str
3139 The predefined workload to run. Current profiles are:
3142 Threaded I/O bench (tiotest/tiobench) like workload.
3145 Aerospike Certification Tool (ACT) like workload.
3147 To view a profile's additional options use :option:`--cmdhelp` after specifying
3148 the profile. For example::
3150 $ fio --profile=act --cmdhelp
3155 .. option:: device-names=str
3160 .. option:: load=int
3163 ACT load multiplier. Default: 1.
3165 .. option:: test-duration=time
3168 How long the entire test takes to run. When the unit is omitted, the value
3169 is given in seconds. Default: 24h.
3171 .. option:: threads-per-queue=int
3174 Number of read I/O threads per device. Default: 8.
3176 .. option:: read-req-num-512-blocks=int
3179 Number of 512B blocks to read at the time. Default: 3.
3181 .. option:: large-block-op-kbytes=int
3184 Size of large block ops in KiB (writes). Default: 131072.
3189 Set to run ACT prep phase.
3191 Tiobench profile options
3192 ~~~~~~~~~~~~~~~~~~~~~~~~
3194 .. option:: size=str
3199 .. option:: block=int
3202 Block size in bytes. Default: 4096.
3204 .. option:: numruns=int
3214 .. option:: threads=int
3219 Interpreting the output
3220 -----------------------
3223 Example output was based on the following:
3224 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
3225 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
3226 --runtime=2m --rw=rw
3228 Fio spits out a lot of output. While running, fio will display the status of the
3229 jobs created. An example of that would be::
3231 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]
3233 The characters inside the first set of square brackets denote the current status of
3234 each thread. The first character is the first job defined in the job file, and so
3235 forth. The possible values (in typical life cycle order) are:
3237 +------+-----+-----------------------------------------------------------+
3239 +======+=====+===========================================================+
3240 | P | | Thread setup, but not started. |
3241 +------+-----+-----------------------------------------------------------+
3242 | C | | Thread created. |
3243 +------+-----+-----------------------------------------------------------+
3244 | I | | Thread initialized, waiting or generating necessary data. |
3245 +------+-----+-----------------------------------------------------------+
3246 | | p | Thread running pre-reading file(s). |
3247 +------+-----+-----------------------------------------------------------+
3248 | | / | Thread is in ramp period. |
3249 +------+-----+-----------------------------------------------------------+
3250 | | R | Running, doing sequential reads. |
3251 +------+-----+-----------------------------------------------------------+
3252 | | r | Running, doing random reads. |
3253 +------+-----+-----------------------------------------------------------+
3254 | | W | Running, doing sequential writes. |
3255 +------+-----+-----------------------------------------------------------+
3256 | | w | Running, doing random writes. |
3257 +------+-----+-----------------------------------------------------------+
3258 | | M | Running, doing mixed sequential reads/writes. |
3259 +------+-----+-----------------------------------------------------------+
3260 | | m | Running, doing mixed random reads/writes. |
3261 +------+-----+-----------------------------------------------------------+
3262 | | D | Running, doing sequential trims. |
3263 +------+-----+-----------------------------------------------------------+
3264 | | d | Running, doing random trims. |
3265 +------+-----+-----------------------------------------------------------+
3266 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
3267 +------+-----+-----------------------------------------------------------+
3268 | | V | Running, doing verification of written data. |
3269 +------+-----+-----------------------------------------------------------+
3270 | f | | Thread finishing. |
3271 +------+-----+-----------------------------------------------------------+
3272 | E | | Thread exited, not reaped by main thread yet. |
3273 +------+-----+-----------------------------------------------------------+
3274 | _ | | Thread reaped. |
3275 +------+-----+-----------------------------------------------------------+
3276 | X | | Thread reaped, exited with an error. |
3277 +------+-----+-----------------------------------------------------------+
3278 | K | | Thread reaped, exited due to signal. |
3279 +------+-----+-----------------------------------------------------------+
3282 Example output was based on the following:
3283 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3284 --time_based --rate=2512k --bs=256K --numjobs=10 \
3285 --name=readers --rw=read --name=writers --rw=write
3287 Fio will condense the thread string as not to take up more space on the command
3288 line than needed. For instance, if you have 10 readers and 10 writers running,
3289 the output would look like this::
3291 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]
3293 Note that the status string is displayed in order, so it's possible to tell which of
3294 the jobs are currently doing what. In the example above this means that jobs 1--10
3295 are readers and 11--20 are writers.
3297 The other values are fairly self explanatory -- number of threads currently
3298 running and doing I/O, the number of currently open files (f=), the estimated
3299 completion percentage, the rate of I/O since last check (read speed listed first,
3300 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
3301 and time to completion for the current running group. It's impossible to estimate
3302 runtime of the following groups (if any).
3305 Example output was based on the following:
3306 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3307 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3308 --bs=7K --name=Client1 --rw=write
3310 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3311 each thread, group of threads, and disks in that order. For each overall thread (or
3312 group) the output looks like::
3314 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3315 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3316 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3317 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3318 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3319 clat percentiles (usec):
3320 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3321 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3322 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3323 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3325 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3326 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
3327 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3328 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3329 lat (msec) : 100=0.65%
3330 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3331 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3332 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3333 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3334 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3335 latency : target=0, window=0, percentile=100.00%, depth=8
3337 The job name (or first job's name when using :option:`group_reporting`) is printed,
3338 along with the group id, count of jobs being aggregated, last error id seen (which
3339 is 0 when there are no errors), pid/tid of that thread and the time the job/group
3340 completed. Below are the I/O statistics for each data direction performed (showing
3341 writes in the example above). In the order listed, they denote:
3344 The string before the colon shows the I/O direction the statistics
3345 are for. **IOPS** is the average I/Os performed per second. **BW**
3346 is the average bandwidth rate shown as: value in power of 2 format
3347 (value in power of 10 format). The last two values show: (**total
3348 I/O performed** in power of 2 format / **runtime** of that thread).
3351 Submission latency (**min** being the minimum, **max** being the
3352 maximum, **avg** being the average, **stdev** being the standard
3353 deviation). This is the time it took to submit the I/O. For
3354 sync I/O this row is not displayed as the slat is really the
3355 completion latency (since queue/complete is one operation there).
3356 This value can be in nanoseconds, microseconds or milliseconds ---
3357 fio will choose the most appropriate base and print that (in the
3358 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
3359 latencies are always expressed in microseconds.
3362 Completion latency. Same names as slat, this denotes the time from
3363 submission to completion of the I/O pieces. For sync I/O, clat will
3364 usually be equal (or very close) to 0, as the time from submit to
3365 complete is basically just CPU time (I/O has already been done, see slat
3369 Total latency. Same names as slat and clat, this denotes the time from
3370 when fio created the I/O unit to completion of the I/O operation.
3373 Bandwidth statistics based on samples. Same names as the xlat stats,
3374 but also includes the number of samples taken (**samples**) and an
3375 approximate percentage of total aggregate bandwidth this thread
3376 received in its group (**per**). This last value is only really
3377 useful if the threads in this group are on the same disk, since they
3378 are then competing for disk access.
3381 IOPS statistics based on samples. Same names as bw.
3383 **lat (nsec/usec/msec)**
3384 The distribution of I/O completion latencies. This is the time from when
3385 I/O leaves fio and when it gets completed. Unlike the separate
3386 read/write/trim sections above, the data here and in the remaining
3387 sections apply to all I/Os for the reporting group. 250=0.04% means that
3388 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
3389 of the I/Os required 250 to 499us for completion.
3392 CPU usage. User and system time, along with the number of context
3393 switches this thread went through, usage of system and user time, and
3394 finally the number of major and minor page faults. The CPU utilization
3395 numbers are averages for the jobs in that reporting group, while the
3396 context and fault counters are summed.
3399 The distribution of I/O depths over the job lifetime. The numbers are
3400 divided into powers of 2 and each entry covers depths from that value
3401 up to those that are lower than the next entry -- e.g., 16= covers
3402 depths from 16 to 31. Note that the range covered by a depth
3403 distribution entry can be different to the range covered by the
3404 equivalent submit/complete distribution entry.
3407 How many pieces of I/O were submitting in a single submit call. Each
3408 entry denotes that amount and below, until the previous entry -- e.g.,
3409 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3410 call. Note that the range covered by a submit distribution entry can
3411 be different to the range covered by the equivalent depth distribution
3415 Like the above submit number, but for completions instead.
3418 The number of read/write/trim requests issued, and how many of them were
3422 These values are for :option:`latency_target` and related options. When
3423 these options are engaged, this section describes the I/O depth required
3424 to meet the specified latency target.
3427 Example output was based on the following:
3428 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3429 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3430 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3432 After each client has been listed, the group statistics are printed. They
3433 will look like this::
3435 Run status group 0 (all jobs):
3436 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
3437 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
3439 For each data direction it prints:
3442 Aggregate bandwidth of threads in this group followed by the
3443 minimum and maximum bandwidth of all the threads in this group.
3444 Values outside of brackets are power-of-2 format and those
3445 within are the equivalent value in a power-of-10 format.
3447 Aggregate I/O performed of all threads in this group. The
3448 format is the same as bw.
3450 The smallest and longest runtimes of the threads in this group.
3452 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
3454 Disk stats (read/write):
3455 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3457 Each value is printed for both reads and writes, with reads first. The
3461 Number of I/Os performed by all groups.
3463 Number of merges performed by the I/O scheduler.
3465 Number of ticks we kept the disk busy.
3467 Total time spent in the disk queue.
3469 The disk utilization. A value of 100% means we kept the disk
3470 busy constantly, 50% would be a disk idling half of the time.
3472 It is also possible to get fio to dump the current output while it is running,
3473 without terminating the job. To do that, send fio the **USR1** signal. You can
3474 also get regularly timed dumps by using the :option:`--status-interval`
3475 parameter, or by creating a file in :file:`/tmp` named
3476 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3477 current output status.
3483 For scripted usage where you typically want to generate tables or graphs of the
3484 results, fio can output the results in a semicolon separated format. The format
3485 is one long line of values, such as::
3487 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%
3488 A description of this job goes here.
3490 The job description (if provided) follows on a second line.
3492 To enable terse output, use the :option:`--minimal` or
3493 :option:`--output-format`\=terse command line options. The
3494 first value is the version of the terse output format. If the output has to be
3495 changed for some reason, this number will be incremented by 1 to signify that
3498 Split up, the format is as follows (comments in brackets denote when a
3499 field was introduced or whether it's specific to some terse version):
3503 terse version, fio version [v3], jobname, groupid, error
3507 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3508 Submission latency: min, max, mean, stdev (usec)
3509 Completion latency: min, max, mean, stdev (usec)
3510 Completion latency percentiles: 20 fields (see below)
3511 Total latency: min, max, mean, stdev (usec)
3512 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3513 IOPS [v5]: min, max, mean, stdev, number of samples
3519 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3520 Submission latency: min, max, mean, stdev (usec)
3521 Completion latency: min, max, mean, stdev (usec)
3522 Completion latency percentiles: 20 fields (see below)
3523 Total latency: min, max, mean, stdev (usec)
3524 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3525 IOPS [v5]: min, max, mean, stdev, number of samples
3527 TRIM status [all but version 3]:
3529 Fields are similar to READ/WRITE status.
3533 user, system, context switches, major faults, minor faults
3537 <=1, 2, 4, 8, 16, 32, >=64
3539 I/O latencies microseconds::
3541 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3543 I/O latencies milliseconds::
3545 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3547 Disk utilization [v3]::
3549 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
3550 time spent in queue, disk utilization percentage
3552 Additional Info (dependent on continue_on_error, default off)::
3554 total # errors, first error code
3556 Additional Info (dependent on description being set)::
3560 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3561 terse output fio writes all of them. Each field will look like this::
3565 which is the Xth percentile, and the `usec` latency associated with it.
3567 For `Disk utilization`, all disks used by fio are shown. So for each disk there
3568 will be a disk utilization section.
3570 Below is a single line containing short names for each of the fields in the
3571 minimal output v3, separated by semicolons::
3573 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth;read_iops;read_runtime_ms;read_slat_min;read_slat_max;read_slat_mean;read_slat_dev;read_clat_min;read_clat_max;read_clat_mean;read_clat_dev;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min;read_lat_max;read_lat_mean;read_lat_dev;read_bw_min;read_bw_max;read_bw_agg_pct;read_bw_mean;read_bw_dev;write_kb;write_bandwidth;write_iops;write_runtime_ms;write_slat_min;write_slat_max;write_slat_mean;write_slat_dev;write_clat_min;write_clat_max;write_clat_mean;write_clat_dev;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min;write_lat_max;write_lat_mean;write_lat_dev;write_bw_min;write_bw_max;write_bw_agg_pct;write_bw_mean;write_bw_dev;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
3579 The `json` output format is intended to be both human readable and convenient
3580 for automated parsing. For the most part its sections mirror those of the
3581 `normal` output. The `runtime` value is reported in msec and the `bw` value is
3582 reported in 1024 bytes per second units.
3588 The `json+` output format is identical to the `json` output format except that it
3589 adds a full dump of the completion latency bins. Each `bins` object contains a
3590 set of (key, value) pairs where keys are latency durations and values count how
3591 many I/Os had completion latencies of the corresponding duration. For example,
3594 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
3596 This data indicates that one I/O required 87,552ns to complete, two I/Os required
3597 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
3599 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
3600 json+ output and generates CSV-formatted latency data suitable for plotting.
3602 The latency durations actually represent the midpoints of latency intervals.
3603 For details refer to :file:`stat.h`.
3609 There are two trace file format that you can encounter. The older (v1) format is
3610 unsupported since version 1.20-rc3 (March 2008). It will still be described
3611 below in case that you get an old trace and want to understand it.
3613 In any case the trace is a simple text file with a single action per line.
3616 Trace file format v1
3617 ~~~~~~~~~~~~~~~~~~~~
3619 Each line represents a single I/O action in the following format::
3623 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
3625 This format is not supported in fio versions >= 1.20-rc3.
3628 Trace file format v2
3629 ~~~~~~~~~~~~~~~~~~~~
3631 The second version of the trace file format was added in fio version 1.17. It
3632 allows to access more then one file per trace and has a bigger set of possible
3635 The first line of the trace file has to be::
3639 Following this can be lines in two different formats, which are described below.
3641 The file management format::
3645 The `filename` is given as an absolute path. The `action` can be one of these:
3648 Add the given `filename` to the trace.
3650 Open the file with the given `filename`. The `filename` has to have
3651 been added with the **add** action before.
3653 Close the file with the given `filename`. The file has to have been
3657 The file I/O action format::
3659 filename action offset length
3661 The `filename` is given as an absolute path, and has to have been added and
3662 opened before it can be used with this format. The `offset` and `length` are
3663 given in bytes. The `action` can be one of these:
3666 Wait for `offset` microseconds. Everything below 100 is discarded.
3667 The time is relative to the previous `wait` statement.
3669 Read `length` bytes beginning from `offset`.
3671 Write `length` bytes beginning from `offset`.
3673 :manpage:`fsync(2)` the file.
3675 :manpage:`fdatasync(2)` the file.
3677 Trim the given file from the given `offset` for `length` bytes.
3679 CPU idleness profiling
3680 ----------------------
3682 In some cases, we want to understand CPU overhead in a test. For example, we
3683 test patches for the specific goodness of whether they reduce CPU usage.
3684 Fio implements a balloon approach to create a thread per CPU that runs at idle
3685 priority, meaning that it only runs when nobody else needs the cpu.
3686 By measuring the amount of work completed by the thread, idleness of each CPU
3687 can be derived accordingly.
3689 An unit work is defined as touching a full page of unsigned characters. Mean and
3690 standard deviation of time to complete an unit work is reported in "unit work"
3691 section. Options can be chosen to report detailed percpu idleness or overall
3692 system idleness by aggregating percpu stats.
3695 Verification and triggers
3696 -------------------------
3698 Fio is usually run in one of two ways, when data verification is done. The first
3699 is a normal write job of some sort with verify enabled. When the write phase has
3700 completed, fio switches to reads and verifies everything it wrote. The second
3701 model is running just the write phase, and then later on running the same job
3702 (but with reads instead of writes) to repeat the same I/O patterns and verify
3703 the contents. Both of these methods depend on the write phase being completed,
3704 as fio otherwise has no idea how much data was written.
3706 With verification triggers, fio supports dumping the current write state to
3707 local files. Then a subsequent read verify workload can load this state and know
3708 exactly where to stop. This is useful for testing cases where power is cut to a
3709 server in a managed fashion, for instance.
3711 A verification trigger consists of two things:
3713 1) Storing the write state of each job.
3714 2) Executing a trigger command.
3716 The write state is relatively small, on the order of hundreds of bytes to single
3717 kilobytes. It contains information on the number of completions done, the last X
3720 A trigger is invoked either through creation ('touch') of a specified file in
3721 the system, or through a timeout setting. If fio is run with
3722 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
3723 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
3724 will fire off the trigger (thus saving state, and executing the trigger
3727 For client/server runs, there's both a local and remote trigger. If fio is
3728 running as a server backend, it will send the job states back to the client for
3729 safe storage, then execute the remote trigger, if specified. If a local trigger
3730 is specified, the server will still send back the write state, but the client
3731 will then execute the trigger.
3733 Verification trigger example
3734 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3736 Let's say we want to run a powercut test on the remote Linux machine 'server'.
3737 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
3738 some point during the run, and we'll run this test from the safety or our local
3739 machine, 'localbox'. On the server, we'll start the fio backend normally::
3741 server# fio --server
3743 and on the client, we'll fire off the workload::
3745 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
3747 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
3749 echo b > /proc/sysrq-trigger
3751 on the server once it has received the trigger and sent us the write state. This
3752 will work, but it's not **really** cutting power to the server, it's merely
3753 abruptly rebooting it. If we have a remote way of cutting power to the server
3754 through IPMI or similar, we could do that through a local trigger command
3755 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
3756 ipmi-reboot. On localbox, we could then have run fio with a local trigger
3759 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
3761 For this case, fio would wait for the server to send us the write state, then
3762 execute ``ipmi-reboot server`` when that happened.
3764 Loading verify state
3765 ~~~~~~~~~~~~~~~~~~~~
3767 To load stored write state, a read verification job file must contain the
3768 :option:`verify_state_load` option. If that is set, fio will load the previously
3769 stored state. For a local fio run this is done by loading the files directly,
3770 and on a client/server run, the server backend will ask the client to send the
3771 files over and load them from there.
3777 Fio supports a variety of log file formats, for logging latencies, bandwidth,
3778 and IOPS. The logs share a common format, which looks like this:
3780 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
3783 *Time* for the log entry is always in milliseconds. The *value* logged depends
3784 on the type of log, it will be one of the following:
3787 Value is latency in nsecs
3793 *Data direction* is one of the following:
3802 The entry's *block size* is always in bytes. The *offset* is the offset, in bytes,
3803 from the start of the file, for that particular I/O. The logging of the offset can be
3804 toggled with :option:`log_offset`.
3806 Fio defaults to logging every individual I/O. When IOPS are logged for individual
3807 I/Os the *value* entry will always be 1. If windowed logging is enabled through
3808 :option:`log_avg_msec`, fio logs the average values over the specified period of time.
3809 If windowed logging is enabled and :option:`log_max_value` is set, then fio logs
3810 maximum values in that window instead of averages. Since *data direction*, *block
3811 size* and *offset* are per-I/O values, if windowed logging is enabled they
3812 aren't applicable and will be 0.
3817 Normally fio is invoked as a stand-alone application on the machine where the
3818 I/O workload should be generated. However, the backend and frontend of fio can
3819 be run separately i.e., the fio server can generate an I/O workload on the "Device
3820 Under Test" while being controlled by a client on another machine.
3822 Start the server on the machine which has access to the storage DUT::
3826 where `args` defines what fio listens to. The arguments are of the form
3827 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
3828 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
3829 *hostname* is either a hostname or IP address, and *port* is the port to listen
3830 to (only valid for TCP/IP, not a local socket). Some examples:
3834 Start a fio server, listening on all interfaces on the default port (8765).
3836 2) ``fio --server=ip:hostname,4444``
3838 Start a fio server, listening on IP belonging to hostname and on port 4444.
3840 3) ``fio --server=ip6:::1,4444``
3842 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
3844 4) ``fio --server=,4444``
3846 Start a fio server, listening on all interfaces on port 4444.
3848 5) ``fio --server=1.2.3.4``
3850 Start a fio server, listening on IP 1.2.3.4 on the default port.
3852 6) ``fio --server=sock:/tmp/fio.sock``
3854 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
3856 Once a server is running, a "client" can connect to the fio server with::
3858 fio <local-args> --client=<server> <remote-args> <job file(s)>
3860 where `local-args` are arguments for the client where it is running, `server`
3861 is the connect string, and `remote-args` and `job file(s)` are sent to the
3862 server. The `server` string follows the same format as it does on the server
3863 side, to allow IP/hostname/socket and port strings.
3865 Fio can connect to multiple servers this way::
3867 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
3869 If the job file is located on the fio server, then you can tell the server to
3870 load a local file as well. This is done by using :option:`--remote-config` ::
3872 fio --client=server --remote-config /path/to/file.fio
3874 Then fio will open this local (to the server) job file instead of being passed
3875 one from the client.
3877 If you have many servers (example: 100 VMs/containers), you can input a pathname
3878 of a file containing host IPs/names as the parameter value for the
3879 :option:`--client` option. For example, here is an example :file:`host.list`
3880 file containing 2 hostnames::
3882 host1.your.dns.domain
3883 host2.your.dns.domain
3885 The fio command would then be::
3887 fio --client=host.list <job file(s)>
3889 In this mode, you cannot input server-specific parameters or job files -- all
3890 servers receive the same job file.
3892 In order to let ``fio --client`` runs use a shared filesystem from multiple
3893 hosts, ``fio --client`` now prepends the IP address of the server to the
3894 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
3895 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
3896 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
3897 192.168.10.121, then fio will create two files::
3899 /mnt/nfs/fio/192.168.10.120.fileio.tmp
3900 /mnt/nfs/fio/192.168.10.121.fileio.tmp