4 The first step in getting fio to simulate a desired I/O workload, is writing a
5 job file describing that specific setup. A job file may contain any number of
6 threads and/or files -- the typical contents of the job file is a *global*
7 section defining shared parameters, and one or more job sections describing the
8 jobs involved. When run, fio parses this file and sets everything up as
9 described. If we break down a job from top to bottom, it contains the following
14 Defines the I/O pattern issued to the file(s). We may only be reading
15 sequentially from this file(s), or we may be writing randomly. Or even
16 mixing reads and writes, sequentially or randomly.
17 Should we be doing buffered I/O, or direct/raw I/O?
21 In how large chunks are we issuing I/O? This may be a single value,
22 or it may describe a range of block sizes.
26 How much data are we going to be reading/writing.
30 How do we issue I/O? We could be memory mapping the file, we could be
31 using regular read/write, we could be using splice, async I/O, or even
36 If the I/O engine is async, how large a queuing depth do we want to
42 How many files are we spreading the workload over.
44 `Threads, processes and job synchronization`_
46 How many threads or processes should we spread this workload over.
48 The above are the basic parameters defined for a workload, in addition there's a
49 multitude of parameters that modify other aspects of how this job behaves.
55 .. option:: --debug=type
57 Enable verbose tracing `type` of various fio actions. May be ``all`` for all types
58 or individual types separated by a comma (e.g. ``--debug=file,mem`` will
59 enable file and memory debugging). Currently, additional logging is
63 Dump info related to processes.
65 Dump info related to file actions.
67 Dump info related to I/O queuing.
69 Dump info related to memory allocations.
71 Dump info related to blktrace setup.
73 Dump info related to I/O verification.
75 Enable all debug options.
77 Dump info related to random offset generation.
79 Dump info related to option matching and parsing.
81 Dump info related to disk utilization updates.
83 Dump info only related to job number x.
85 Dump info only related to mutex up/down ops.
87 Dump info related to profile extensions.
89 Dump info related to internal time keeping.
91 Dump info related to networking connections.
93 Dump info related to I/O rate switching.
95 Dump info related to log compress/decompress.
97 Dump info related to steadystate detection.
99 Dump info related to the helper thread.
101 Dump info related to support for zoned block devices.
103 Show available debug options.
105 .. option:: --parse-only
107 Parse options only, don't start any I/O.
109 .. option:: --merge-blktrace-only
111 Merge blktraces only, don't start any I/O.
113 .. option:: --output=filename
115 Write output to file `filename`.
117 .. option:: --output-format=format
119 Set the reporting `format` to `normal`, `terse`, `json`, or `json+`. Multiple
120 formats can be selected, separated by a comma. `terse` is a CSV based
121 format. `json+` is like `json`, except it adds a full dump of the latency
124 .. option:: --bandwidth-log
126 Generate aggregate bandwidth logs.
128 .. option:: --minimal
130 Print statistics in a terse, semicolon-delimited format.
132 .. option:: --append-terse
134 Print statistics in selected mode AND terse, semicolon-delimited format.
135 **Deprecated**, use :option:`--output-format` instead to select multiple
138 .. option:: --terse-version=version
140 Set terse `version` output format (default 3, or 2 or 4 or 5).
142 .. option:: --version
144 Print version information and exit.
148 Print a summary of the command line options and exit.
150 .. option:: --cpuclock-test
152 Perform test and validation of internal CPU clock.
154 .. option:: --crctest=[test]
156 Test the speed of the built-in checksumming functions. If no argument is
157 given, all of them are tested. Alternatively, a comma separated list can
158 be passed, in which case the given ones are tested.
160 .. option:: --cmdhelp=command
162 Print help information for `command`. May be ``all`` for all commands.
164 .. option:: --enghelp=[ioengine[,command]]
166 List all commands defined by `ioengine`, or print help for `command`
167 defined by `ioengine`. If no `ioengine` is given, list all
170 .. option:: --showcmd
172 Convert given job files to a set of command-line options.
174 .. option:: --readonly
176 Turn on safety read-only checks, preventing writes and trims. The
177 ``--readonly`` option is an extra safety guard to prevent users from
178 accidentally starting a write or trim workload when that is not desired.
179 Fio will only modify the device under test if
180 `rw=write/randwrite/rw/randrw/trim/randtrim/trimwrite` is given. This
181 safety net can be used as an extra precaution.
183 .. option:: --eta=when
185 Specifies when real-time ETA estimate should be printed. `when` may be
186 `always`, `never` or `auto`. `auto` is the default, it prints ETA
187 when requested if the output is a TTY. `always` disregards the output
188 type, and prints ETA when requested. `never` never prints ETA.
190 .. option:: --eta-interval=time
192 By default, fio requests client ETA status roughly every second. With
193 this option, the interval is configurable. Fio imposes a minimum
194 allowed time to avoid flooding the console, less than 250 msec is
197 .. option:: --eta-newline=time
199 Force a new line for every `time` period passed. When the unit is omitted,
200 the value is interpreted in seconds.
202 .. option:: --status-interval=time
204 Force a full status dump of cumulative (from job start) values at `time`
205 intervals. This option does *not* provide per-period measurements. So
206 values such as bandwidth are running averages. When the time unit is omitted,
207 `time` is interpreted in seconds. Note that using this option with
208 ``--output-format=json`` will yield output that technically isn't valid
209 json, since the output will be collated sets of valid json. It will need
210 to be split into valid sets of json after the run.
212 .. option:: --section=name
214 Only run specified section `name` in job file. Multiple sections can be specified.
215 The ``--section`` option allows one to combine related jobs into one file.
216 E.g. one job file could define light, moderate, and heavy sections. Tell
217 fio to run only the "heavy" section by giving ``--section=heavy``
218 command line option. One can also specify the "write" operations in one
219 section and "verify" operation in another section. The ``--section`` option
220 only applies to job sections. The reserved *global* section is always
223 .. option:: --alloc-size=kb
225 Allocate additional internal smalloc pools of size `kb` in KiB. The
226 ``--alloc-size`` option increases shared memory set aside for use by fio.
227 If running large jobs with randommap enabled, fio can run out of memory.
228 Smalloc is an internal allocator for shared structures from a fixed size
229 memory pool and can grow to 16 pools. The pool size defaults to 16MiB.
231 NOTE: While running :file:`.fio_smalloc.*` backing store files are visible
234 .. option:: --warnings-fatal
236 All fio parser warnings are fatal, causing fio to exit with an
239 .. option:: --max-jobs=nr
241 Set the maximum number of threads/processes to support to `nr`.
242 NOTE: On Linux, it may be necessary to increase the shared-memory
243 limit (:file:`/proc/sys/kernel/shmmax`) if fio runs into errors while
246 .. option:: --server=args
248 Start a backend server, with `args` specifying what to listen to.
249 See `Client/Server`_ section.
251 .. option:: --daemonize=pidfile
253 Background a fio server, writing the pid to the given `pidfile` file.
255 .. option:: --client=hostname
257 Instead of running the jobs locally, send and run them on the given `hostname`
258 or set of `hostname`\s. See `Client/Server`_ section.
260 .. option:: --remote-config=file
262 Tell fio server to load this local `file`.
264 .. option:: --idle-prof=option
266 Report CPU idleness. `option` is one of the following:
269 Run unit work calibration only and exit.
272 Show aggregate system idleness and unit work.
275 As **system** but also show per CPU idleness.
277 .. option:: --inflate-log=log
279 Inflate and output compressed `log`.
281 .. option:: --trigger-file=file
283 Execute trigger command when `file` exists.
285 .. option:: --trigger-timeout=time
287 Execute trigger at this `time`.
289 .. option:: --trigger=command
291 Set this `command` as local trigger.
293 .. option:: --trigger-remote=command
295 Set this `command` as remote trigger.
297 .. option:: --aux-path=path
299 Use the directory specified by `path` for generated state files instead
300 of the current working directory.
302 Any parameters following the options will be assumed to be job files, unless
303 they match a job file parameter. Multiple job files can be listed and each job
304 file will be regarded as a separate group. Fio will :option:`stonewall`
305 execution between each group.
311 As previously described, fio accepts one or more job files describing what it is
312 supposed to do. The job file format is the classic ini file, where the names
313 enclosed in [] brackets define the job name. You are free to use any ASCII name
314 you want, except *global* which has special meaning. Following the job name is
315 a sequence of zero or more parameters, one per line, that define the behavior of
316 the job. If the first character in a line is a ';' or a '#', the entire line is
317 discarded as a comment.
319 A *global* section sets defaults for the jobs described in that file. A job may
320 override a *global* section parameter, and a job file may even have several
321 *global* sections if so desired. A job is only affected by a *global* section
324 The :option:`--cmdhelp` option also lists all options. If used with a `command`
325 argument, :option:`--cmdhelp` will detail the given `command`.
327 See the `examples/` directory for inspiration on how to write job files. Note
328 the copyright and license requirements currently apply to `examples/` files.
330 So let's look at a really simple job file that defines two processes, each
331 randomly reading from a 128MiB file:
335 ; -- start job file --
346 As you can see, the job file sections themselves are empty as all the described
347 parameters are shared. As no :option:`filename` option is given, fio makes up a
348 `filename` for each of the jobs as it sees fit. On the command line, this job
349 would look as follows::
351 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
354 Let's look at an example that has a number of processes writing randomly to
359 ; -- start job file --
370 Here we have no *global* section, as we only have one job defined anyway. We
371 want to use async I/O here, with a depth of 4 for each file. We also increased
372 the buffer size used to 32KiB and define numjobs to 4 to fork 4 identical
373 jobs. The result is 4 processes each randomly writing to their own 64MiB
374 file. Instead of using the above job file, you could have given the parameters
375 on the command line. For this case, you would specify::
377 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
379 When fio is utilized as a basis of any reasonably large test suite, it might be
380 desirable to share a set of standardized settings across multiple job files.
381 Instead of copy/pasting such settings, any section may pull in an external
382 :file:`filename.fio` file with *include filename* directive, as in the following
385 ; -- start job file including.fio --
389 include glob-include.fio
396 include test-include.fio
397 ; -- end job file including.fio --
401 ; -- start job file glob-include.fio --
404 ; -- end job file glob-include.fio --
408 ; -- start job file test-include.fio --
411 ; -- end job file test-include.fio --
413 Settings pulled into a section apply to that section only (except *global*
414 section). Include directives may be nested in that any included file may contain
415 further include directive(s). Include files may not contain [] sections.
418 Environment variables
419 ~~~~~~~~~~~~~~~~~~~~~
421 Fio also supports environment variable expansion in job files. Any sub-string of
422 the form ``${VARNAME}`` as part of an option value (in other words, on the right
423 of the '='), will be expanded to the value of the environment variable called
424 `VARNAME`. If no such environment variable is defined, or `VARNAME` is the
425 empty string, the empty string will be substituted.
427 As an example, let's look at a sample fio invocation and job file::
429 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
433 ; -- start job file --
440 This will expand to the following equivalent job file at runtime:
444 ; -- start job file --
451 Fio ships with a few example job files, you can also look there for inspiration.
456 Additionally, fio has a set of reserved keywords that will be replaced
457 internally with the appropriate value. Those keywords are:
461 The architecture page size of the running system.
465 Megabytes of total memory in the system.
469 Number of online available CPUs.
471 These can be used on the command line or in the job file, and will be
472 automatically substituted with the current system values when the job is
473 run. Simple math is also supported on these keywords, so you can perform actions
478 and get that properly expanded to 8 times the size of memory in the machine.
484 This section describes in details each parameter associated with a job. Some
485 parameters take an option of a given type, such as an integer or a
486 string. Anywhere a numeric value is required, an arithmetic expression may be
487 used, provided it is surrounded by parentheses. Supported operators are:
496 For time values in expressions, units are microseconds by default. This is
497 different than for time values not in expressions (not enclosed in
498 parentheses). The following types are used:
505 String: A sequence of alphanumeric characters.
508 Integer with possible time suffix. Without a unit value is interpreted as
509 seconds unless otherwise specified. Accepts a suffix of 'd' for days, 'h' for
510 hours, 'm' for minutes, 's' for seconds, 'ms' (or 'msec') for milliseconds and
511 'us' (or 'usec') for microseconds. For example, use 10m for 10 minutes.
516 Integer. A whole number value, which may contain an integer prefix
517 and an integer suffix:
519 [*integer prefix*] **number** [*integer suffix*]
521 The optional *integer prefix* specifies the number's base. The default
522 is decimal. *0x* specifies hexadecimal.
524 The optional *integer suffix* specifies the number's units, and includes an
525 optional unit prefix and an optional unit. For quantities of data, the
526 default unit is bytes. For quantities of time, the default unit is seconds
527 unless otherwise specified.
529 With :option:`kb_base`\=1000, fio follows international standards for unit
530 prefixes. To specify power-of-10 decimal values defined in the
531 International System of Units (SI):
533 * *K* -- means kilo (K) or 1000
534 * *M* -- means mega (M) or 1000**2
535 * *G* -- means giga (G) or 1000**3
536 * *T* -- means tera (T) or 1000**4
537 * *P* -- means peta (P) or 1000**5
539 To specify power-of-2 binary values defined in IEC 80000-13:
541 * *Ki* -- means kibi (Ki) or 1024
542 * *Mi* -- means mebi (Mi) or 1024**2
543 * *Gi* -- means gibi (Gi) or 1024**3
544 * *Ti* -- means tebi (Ti) or 1024**4
545 * *Pi* -- means pebi (Pi) or 1024**5
547 For Zone Block Device Mode:
550 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
551 from those specified in the SI and IEC 80000-13 standards to provide
552 compatibility with old scripts. For example, 4k means 4096.
554 For quantities of data, an optional unit of 'B' may be included
555 (e.g., 'kB' is the same as 'k').
557 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
558 not milli). 'b' and 'B' both mean byte, not bit.
560 Examples with :option:`kb_base`\=1000:
562 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
563 * *1 MiB*: 1048576, 1mi, 1024ki
564 * *1 MB*: 1000000, 1m, 1000k
565 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
566 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
568 Examples with :option:`kb_base`\=1024 (default):
570 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
571 * *1 MiB*: 1048576, 1m, 1024k
572 * *1 MB*: 1000000, 1mi, 1000ki
573 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
574 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
576 To specify times (units are not case sensitive):
580 * *M* -- means minutes
581 * *s* -- or sec means seconds (default)
582 * *ms* -- or *msec* means milliseconds
583 * *us* -- or *usec* means microseconds
585 If the option accepts an upper and lower range, use a colon ':' or
586 minus '-' to separate such values. See :ref:`irange <irange>`.
587 If the lower value specified happens to be larger than the upper value
588 the two values are swapped.
593 Boolean. Usually parsed as an integer, however only defined for
594 true and false (1 and 0).
599 Integer range with suffix. Allows value range to be given, such as
600 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
601 option allows two sets of ranges, they can be specified with a ',' or '/'
602 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
605 A list of floating point numbers, separated by a ':' character.
607 With the above in mind, here follows the complete list of fio job parameters.
613 .. option:: kb_base=int
615 Select the interpretation of unit prefixes in input parameters.
618 Inputs comply with IEC 80000-13 and the International
619 System of Units (SI). Use:
621 - power-of-2 values with IEC prefixes (e.g., KiB)
622 - power-of-10 values with SI prefixes (e.g., kB)
625 Compatibility mode (default). To avoid breaking old scripts:
627 - power-of-2 values with SI prefixes
628 - power-of-10 values with IEC prefixes
630 See :option:`bs` for more details on input parameters.
632 Outputs always use correct prefixes. Most outputs include both
635 bw=2383.3kB/s (2327.4KiB/s)
637 If only one value is reported, then kb_base selects the one to use:
639 **1000** -- SI prefixes
641 **1024** -- IEC prefixes
643 .. option:: unit_base=int
645 Base unit for reporting. Allowed values are:
648 Use auto-detection (default).
660 ASCII name of the job. This may be used to override the name printed by fio
661 for this job. Otherwise the job name is used. On the command line this
662 parameter has the special purpose of also signaling the start of a new job.
664 .. option:: description=str
666 Text description of the job. Doesn't do anything except dump this text
667 description when this job is run. It's not parsed.
669 .. option:: loops=int
671 Run the specified number of iterations of this job. Used to repeat the same
672 workload a given number of times. Defaults to 1.
674 .. option:: numjobs=int
676 Create the specified number of clones of this job. Each clone of job
677 is spawned as an independent thread or process. May be used to setup a
678 larger number of threads/processes doing the same thing. Each thread is
679 reported separately; to see statistics for all clones as a whole, use
680 :option:`group_reporting` in conjunction with :option:`new_group`.
681 See :option:`--max-jobs`. Default: 1.
684 Time related parameters
685 ~~~~~~~~~~~~~~~~~~~~~~~
687 .. option:: runtime=time
689 Limit runtime. The test will run until it completes the configured I/O
690 workload or until it has run for this specified amount of time, whichever
691 occurs first. It can be quite hard to determine for how long a specified
692 job will run, so this parameter is handy to cap the total runtime to a
693 given time. When the unit is omitted, the value is interpreted in
696 .. option:: time_based
698 If set, fio will run for the duration of the :option:`runtime` specified
699 even if the file(s) are completely read or written. It will simply loop over
700 the same workload as many times as the :option:`runtime` allows.
702 .. option:: startdelay=irange(time)
704 Delay the start of job for the specified amount of time. Can be a single
705 value or a range. When given as a range, each thread will choose a value
706 randomly from within the range. Value is in seconds if a unit is omitted.
708 .. option:: ramp_time=time
710 If set, fio will run the specified workload for this amount of time before
711 logging any performance numbers. Useful for letting performance settle
712 before logging results, thus minimizing the runtime required for stable
713 results. Note that the ``ramp_time`` is considered lead in time for a job,
714 thus it will increase the total runtime if a special timeout or
715 :option:`runtime` is specified. When the unit is omitted, the value is
718 .. option:: clocksource=str
720 Use the given clocksource as the base of timing. The supported options are:
723 :manpage:`gettimeofday(2)`
726 :manpage:`clock_gettime(2)`
729 Internal CPU clock source
731 cpu is the preferred clocksource if it is reliable, as it is very fast (and
732 fio is heavy on time calls). Fio will automatically use this clocksource if
733 it's supported and considered reliable on the system it is running on,
734 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
735 means supporting TSC Invariant.
737 .. option:: gtod_reduce=bool
739 Enable all of the :manpage:`gettimeofday(2)` reducing options
740 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
741 reduce precision of the timeout somewhat to really shrink the
742 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
743 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
744 time keeping was enabled.
746 .. option:: gtod_cpu=int
748 Sometimes it's cheaper to dedicate a single thread of execution to just
749 getting the current time. Fio (and databases, for instance) are very
750 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
751 one CPU aside for doing nothing but logging current time to a shared memory
752 location. Then the other threads/processes that run I/O workloads need only
753 copy that segment, instead of entering the kernel with a
754 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
755 calls will be excluded from other uses. Fio will manually clear it from the
756 CPU mask of other jobs.
762 .. option:: directory=str
764 Prefix filenames with this directory. Used to place files in a different
765 location than :file:`./`. You can specify a number of directories by
766 separating the names with a ':' character. These directories will be
767 assigned equally distributed to job clones created by :option:`numjobs` as
768 long as they are using generated filenames. If specific `filename(s)` are
769 set fio will use the first listed directory, and thereby matching the
770 `filename` semantic (which generates a file for each clone if not
771 specified, but lets all clones use the same file if set).
773 See the :option:`filename` option for information on how to escape "``:``"
774 characters within the directory path itself.
776 Note: To control the directory fio will use for internal state files
777 use :option:`--aux-path`.
779 .. option:: filename=str
781 Fio normally makes up a `filename` based on the job name, thread number, and
782 file number (see :option:`filename_format`). If you want to share files
783 between threads in a job or several
784 jobs with fixed file paths, specify a `filename` for each of them to override
785 the default. If the ioengine is file based, you can specify a number of files
786 by separating the names with a ':' colon. So if you wanted a job to open
787 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
788 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
789 specified, :option:`nrfiles` is ignored. The size of regular files specified
790 by this option will be :option:`size` divided by number of files unless an
791 explicit size is specified by :option:`filesize`.
793 Each colon in the wanted path must be escaped with a ``\``
794 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
795 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
796 :file:`F:\\filename` then you would use ``filename=F\:\filename``.
798 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
799 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
800 Note: Windows and FreeBSD prevent write access to areas
801 of the disk containing in-use data (e.g. filesystems).
803 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
804 of the two depends on the read/write direction set.
806 .. option:: filename_format=str
808 If sharing multiple files between jobs, it is usually necessary to have fio
809 generate the exact names that you want. By default, fio will name a file
810 based on the default file format specification of
811 :file:`jobname.jobnumber.filenumber`. With this option, that can be
812 customized. Fio will recognize and replace the following keywords in this
816 The name of the worker thread or process.
818 IP of the fio process when using client/server mode.
820 The incremental number of the worker thread or process.
822 The incremental number of the file for that worker thread or
825 To have dependent jobs share a set of files, this option can be set to have
826 fio generate filenames that are shared between the two. For instance, if
827 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
828 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
829 will be used if no other format specifier is given.
831 If you specify a path then the directories will be created up to the
832 main directory for the file. So for example if you specify
833 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
834 created before the file setup part of the job. If you specify
835 :option:`directory` then the path will be relative that directory,
836 otherwise it is treated as the absolute path.
838 .. option:: unique_filename=bool
840 To avoid collisions between networked clients, fio defaults to prefixing any
841 generated filenames (with a directory specified) with the source of the
842 client connecting. To disable this behavior, set this option to 0.
844 .. option:: opendir=str
846 Recursively open any files below directory `str`.
848 .. option:: lockfile=str
850 Fio defaults to not locking any files before it does I/O to them. If a file
851 or file descriptor is shared, fio can serialize I/O to that file to make the
852 end result consistent. This is usual for emulating real workloads that share
853 files. The lock modes are:
856 No locking. The default.
858 Only one thread or process may do I/O at a time, excluding all
861 Read-write locking on the file. Many readers may
862 access the file at the same time, but writes get exclusive access.
864 .. option:: nrfiles=int
866 Number of files to use for this job. Defaults to 1. The size of files
867 will be :option:`size` divided by this unless explicit size is specified by
868 :option:`filesize`. Files are created for each thread separately, and each
869 file will have a file number within its name by default, as explained in
870 :option:`filename` section.
873 .. option:: openfiles=int
875 Number of files to keep open at the same time. Defaults to the same as
876 :option:`nrfiles`, can be set smaller to limit the number simultaneous
879 .. option:: file_service_type=str
881 Defines how fio decides which file from a job to service next. The following
885 Choose a file at random.
888 Round robin over opened files. This is the default.
891 Finish one file before moving on to the next. Multiple files can
892 still be open depending on :option:`openfiles`.
895 Use a *Zipf* distribution to decide what file to access.
898 Use a *Pareto* distribution to decide what file to access.
901 Use a *Gaussian* (normal) distribution to decide what file to
907 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
908 tell fio how many I/Os to issue before switching to a new file. For example,
909 specifying ``file_service_type=random:8`` would cause fio to issue
910 8 I/Os before selecting a new file at random. For the non-uniform
911 distributions, a floating point postfix can be given to influence how the
912 distribution is skewed. See :option:`random_distribution` for a description
913 of how that would work.
915 .. option:: ioscheduler=str
917 Attempt to switch the device hosting the file to the specified I/O scheduler
920 .. option:: create_serialize=bool
922 If true, serialize the file creation for the jobs. This may be handy to
923 avoid interleaving of data files, which may greatly depend on the filesystem
924 used and even the number of processors in the system. Default: true.
926 .. option:: create_fsync=bool
928 :manpage:`fsync(2)` the data file after creation. This is the default.
930 .. option:: create_on_open=bool
932 If true, don't pre-create files but allow the job's open() to create a file
933 when it's time to do I/O. Default: false -- pre-create all necessary files
936 .. option:: create_only=bool
938 If true, fio will only run the setup phase of the job. If files need to be
939 laid out or updated on disk, only that will be done -- the actual job contents
940 are not executed. Default: false.
942 .. option:: allow_file_create=bool
944 If true, fio is permitted to create files as part of its workload. If this
945 option is false, then fio will error out if
946 the files it needs to use don't already exist. Default: true.
948 .. option:: allow_mounted_write=bool
950 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
951 to what appears to be a mounted device or partition. This should help catch
952 creating inadvertently destructive tests, not realizing that the test will
953 destroy data on the mounted file system. Note that some platforms don't allow
954 writing against a mounted device regardless of this option. Default: false.
956 .. option:: pre_read=bool
958 If this is given, files will be pre-read into memory before starting the
959 given I/O operation. This will also clear the :option:`invalidate` flag,
960 since it is pointless to pre-read and then drop the cache. This will only
961 work for I/O engines that are seek-able, since they allow you to read the
962 same data multiple times. Thus it will not work on non-seekable I/O engines
963 (e.g. network, splice). Default: false.
965 .. option:: unlink=bool
967 Unlink the job files when done. Not the default, as repeated runs of that
968 job would then waste time recreating the file set again and again. Default:
971 .. option:: unlink_each_loop=bool
973 Unlink job files after each iteration or loop. Default: false.
975 .. option:: zonemode=str
980 The :option:`zonerange`, :option:`zonesize`,
981 :option `zonecapacity` and option:`zoneskip`
982 parameters are ignored.
984 I/O happens in a single zone until
985 :option:`zonesize` bytes have been transferred.
986 After that number of bytes has been
987 transferred processing of the next zone
988 starts. :option `zonecapacity` is ignored.
990 Zoned block device mode. I/O happens
991 sequentially in each zone, even if random I/O
992 has been selected. Random I/O happens across
993 all zones instead of being restricted to a
994 single zone. The :option:`zoneskip` parameter
995 is ignored. :option:`zonerange` and
996 :option:`zonesize` must be identical.
997 Trim is handled using a zone reset operation.
998 Trim only considers non-empty sequential write
999 required and sequential write preferred zones.
1001 .. option:: zonerange=int
1003 Size of a single zone. See also :option:`zonesize` and
1006 .. option:: zonesize=int
1008 For :option:`zonemode` =strided, this is the number of bytes to
1009 transfer before skipping :option:`zoneskip` bytes. If this parameter
1010 is smaller than :option:`zonerange` then only a fraction of each zone
1011 with :option:`zonerange` bytes will be accessed. If this parameter is
1012 larger than :option:`zonerange` then each zone will be accessed
1013 multiple times before skipping to the next zone.
1015 For :option:`zonemode` =zbd, this is the size of a single zone. The
1016 :option:`zonerange` parameter is ignored in this mode.
1019 .. option:: zonecapacity=int
1021 For :option:`zonemode` =zbd, this defines the capacity of a single zone,
1022 which is the accessible area starting from the zone start address.
1023 This parameter only applies when using :option:`zonemode` =zbd in
1024 combination with regular block devices. If not specified it defaults to
1025 the zone size. If the target device is a zoned block device, the zone
1026 capacity is obtained from the device information and this option is
1029 .. option:: zoneskip=int
1031 For :option:`zonemode` =strided, the number of bytes to skip after
1032 :option:`zonesize` bytes of data have been transferred. This parameter
1033 must be zero for :option:`zonemode` =zbd.
1035 .. option:: read_beyond_wp=bool
1037 This parameter applies to :option:`zonemode` =zbd only.
1039 Zoned block devices are block devices that consist of multiple zones.
1040 Each zone has a type, e.g. conventional or sequential. A conventional
1041 zone can be written at any offset that is a multiple of the block
1042 size. Sequential zones must be written sequentially. The position at
1043 which a write must occur is called the write pointer. A zoned block
1044 device can be either drive managed, host managed or host aware. For
1045 host managed devices the host must ensure that writes happen
1046 sequentially. Fio recognizes host managed devices and serializes
1047 writes to sequential zones for these devices.
1049 If a read occurs in a sequential zone beyond the write pointer then
1050 the zoned block device will complete the read without reading any data
1051 from the storage medium. Since such reads lead to unrealistically high
1052 bandwidth and IOPS numbers fio only reads beyond the write pointer if
1053 explicitly told to do so. Default: false.
1055 .. option:: max_open_zones=int
1057 A zone of a zoned block device is in the open state when it is partially
1058 written (i.e. not all sectors of the zone have been written). Zoned
1059 block devices may have a limit on the total number of zones that can
1060 be simultaneously in the open state, that is, the number of zones that
1061 can be written to simultaneously. The :option:`max_open_zones` parameter
1062 limits the number of zones to which write commands are issued by all fio
1063 jobs, that is, limits the number of zones that will be in the open
1064 state. This parameter is relevant only if the :option:`zonemode` =zbd is
1065 used. The default value is always equal to maximum number of open zones
1066 of the target zoned block device and a value higher than this limit
1067 cannot be specified by users unless the option
1068 :option:`ignore_zone_limits` is specified. When
1069 :option:`ignore_zone_limits` is specified or the target device has no
1070 limit on the number of zones that can be in an open state,
1071 :option:`max_open_zones` can specify 0 to disable any limit on the
1072 number of zones that can be simultaneously written to by all jobs.
1074 .. option:: job_max_open_zones=int
1076 In the same manner as :option:`max_open_zones`, limit the number of open
1077 zones per fio job, that is, the number of zones that a single job can
1078 simultaneously write to. A value of zero indicates no limit.
1081 .. option:: ignore_zone_limits=bool
1083 If this option is used, fio will ignore the maximum number of open
1084 zones limit of the zoned block device in use, thus allowing the
1085 option :option:`max_open_zones` value to be larger than the device
1086 reported limit. Default: false.
1088 .. option:: zone_reset_threshold=float
1090 A number between zero and one that indicates the ratio of written bytes
1091 in the zones with write pointers in the IO range to the size of the IO
1092 range. When current ratio is above this ratio, zones are reset
1093 periodically as :option:`zone_reset_frequency` specifies. If there are
1094 multiple jobs when using this option, the IO range for all write jobs
1097 .. option:: zone_reset_frequency=float
1099 A number between zero and one that indicates how often a zone reset
1100 should be issued if the zone reset threshold has been exceeded. A zone
1101 reset is submitted after each (1 / zone_reset_frequency) write
1102 requests. This and the previous parameter can be used to simulate
1103 garbage collection activity.
1109 .. option:: direct=bool
1111 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
1112 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
1113 ioengines don't support direct I/O. Default: false.
1115 .. option:: buffered=bool
1117 If value is true, use buffered I/O. This is the opposite of the
1118 :option:`direct` option. Defaults to true.
1120 .. option:: readwrite=str, rw=str
1122 Type of I/O pattern. Accepted values are:
1129 Sequential trims (Linux block devices and SCSI
1130 character devices only).
1136 Random trims (Linux block devices and SCSI
1137 character devices only).
1139 Sequential mixed reads and writes.
1141 Random mixed reads and writes.
1143 Sequential trim+write sequences. Blocks will be trimmed first,
1144 then the same blocks will be written to. So if ``io_size=64K``
1145 is specified, Fio will trim a total of 64K bytes and also
1146 write 64K bytes on the same trimmed blocks. This behaviour
1147 will be consistent with ``number_ios`` or other Fio options
1148 limiting the total bytes or number of I/O's.
1150 Like trimwrite, but uses random offsets rather
1151 than sequential writes.
1153 Fio defaults to read if the option is not specified. For the mixed I/O
1154 types, the default is to split them 50/50. For certain types of I/O the
1155 result may still be skewed a bit, since the speed may be different.
1157 It is possible to specify the number of I/Os to do before getting a new
1158 offset by appending ``:<nr>`` to the end of the string given. For a
1159 random read, it would look like ``rw=randread:8`` for passing in an offset
1160 modifier with a value of 8. If the suffix is used with a sequential I/O
1161 pattern, then the *<nr>* value specified will be **added** to the generated
1162 offset for each I/O turning sequential I/O into sequential I/O with holes.
1163 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1164 the :option:`rw_sequencer` option.
1166 .. option:: rw_sequencer=str
1168 If an offset modifier is given by appending a number to the ``rw=<str>``
1169 line, then this option controls how that number modifies the I/O offset
1170 being generated. Accepted values are:
1173 Generate sequential offset.
1175 Generate the same offset.
1177 ``sequential`` is only useful for random I/O, where fio would normally
1178 generate a new random offset for every I/O. If you append e.g. 8 to
1179 randread, i.e. ``rw=randread:8`` you would get a new random offset for
1180 every 8 I/Os. The result would be a sequence of 8 sequential offsets
1181 with a random starting point. However this behavior may change if a
1182 sequential I/O reaches end of the file. As sequential I/O is already
1183 sequential, setting ``sequential`` for that would not result in any
1184 difference. ``identical`` behaves in a similar fashion, except it sends
1185 the same offset 8 number of times before generating a new offset.
1190 rw_sequencer=sequential
1193 The generated sequence of offsets will look like this:
1194 4k, 8k, 12k, 16k, 20k, 24k, 28k, 32k, 92k, 96k, 100k, 104k, 108k,
1195 112k, 116k, 120k, 48k, 52k ...
1200 rw_sequencer=identical
1203 The generated sequence of offsets will look like this:
1204 4k, 4k, 4k, 4k, 4k, 4k, 4k, 4k, 92k, 92k, 92k, 92k, 92k, 92k, 92k, 92k,
1207 .. option:: unified_rw_reporting=str
1209 Fio normally reports statistics on a per data direction basis, meaning that
1210 reads, writes, and trims are accounted and reported separately. This option
1211 determines whether fio reports the results normally, summed together, or as
1213 Accepted values are:
1216 Normal statistics reporting.
1219 Statistics are summed per data direction and reported together.
1222 Statistics are reported normally, followed by the mixed statistics.
1225 Backward-compatible alias for **none**.
1228 Backward-compatible alias for **mixed**.
1233 .. option:: randrepeat=bool
1235 Seed the random number generator used for random I/O patterns in a
1236 predictable way so the pattern is repeatable across runs. Default: true.
1238 .. option:: allrandrepeat=bool
1240 Seed all random number generators in a predictable way so results are
1241 repeatable across runs. Default: false.
1243 .. option:: randseed=int
1245 Seed the random number generators based on this seed value, to be able to
1246 control what sequence of output is being generated. If not set, the random
1247 sequence depends on the :option:`randrepeat` setting.
1249 .. option:: fallocate=str
1251 Whether pre-allocation is performed when laying down files.
1252 Accepted values are:
1255 Do not pre-allocate space.
1258 Use a platform's native pre-allocation call but fall back to
1259 **none** behavior if it fails/is not implemented.
1262 Pre-allocate via :manpage:`posix_fallocate(3)`.
1265 Pre-allocate via :manpage:`fallocate(2)` with
1266 FALLOC_FL_KEEP_SIZE set.
1269 Extend file to final size via :manpage:`ftruncate(2)`
1270 instead of allocating.
1273 Backward-compatible alias for **none**.
1276 Backward-compatible alias for **posix**.
1278 May not be available on all supported platforms. **keep** is only available
1279 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1280 because ZFS doesn't support pre-allocation. Default: **native** if any
1281 pre-allocation methods except **truncate** are available, **none** if not.
1283 Note that using **truncate** on Windows will interact surprisingly
1284 with non-sequential write patterns. When writing to a file that has
1285 been extended by setting the end-of-file information, Windows will
1286 backfill the unwritten portion of the file up to that offset with
1287 zeroes before issuing the new write. This means that a single small
1288 write to the end of an extended file will stall until the entire
1289 file has been filled with zeroes.
1291 .. option:: fadvise_hint=str
1293 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1294 advise the kernel on what I/O patterns are likely to be issued.
1295 Accepted values are:
1298 Backwards-compatible hint for "no hint".
1301 Backwards compatible hint for "advise with fio workload type". This
1302 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1303 for a sequential workload.
1306 Advise using **FADV_SEQUENTIAL**.
1309 Advise using **FADV_RANDOM**.
1312 Advise using **FADV_NOREUSE**. This may be a no-op on older Linux
1313 kernels. Since Linux 6.3, it provides a hint to the LRU algorithm.
1314 See the :manpage:`posix_fadvise(2)` man page.
1316 .. option:: write_hint=str
1318 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1319 from a write. Only supported on Linux, as of version 4.13. Accepted
1323 No particular life time associated with this file.
1326 Data written to this file has a short life time.
1329 Data written to this file has a medium life time.
1332 Data written to this file has a long life time.
1335 Data written to this file has a very long life time.
1337 The values are all relative to each other, and no absolute meaning
1338 should be associated with them.
1340 .. option:: offset=int
1342 Start I/O at the provided offset in the file, given as either a fixed size in
1343 bytes, zones or a percentage. If a percentage is given, the generated offset will be
1344 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1345 provided. Data before the given offset will not be touched. This
1346 effectively caps the file size at `real_size - offset`. Can be combined with
1347 :option:`size` to constrain the start and end range of the I/O workload.
1348 A percentage can be specified by a number between 1 and 100 followed by '%',
1349 for example, ``offset=20%`` to specify 20%. In ZBD mode, value can be set as
1350 number of zones using 'z'.
1352 .. option:: offset_align=int
1354 If set to non-zero value, the byte offset generated by a percentage ``offset``
1355 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1356 offset is aligned to the minimum block size.
1358 .. option:: offset_increment=int
1360 If this is provided, then the real offset becomes `offset + offset_increment
1361 * thread_number`, where the thread number is a counter that starts at 0 and
1362 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1363 specified). This option is useful if there are several jobs which are
1364 intended to operate on a file in parallel disjoint segments, with even
1365 spacing between the starting points. Percentages can be used for this option.
1366 If a percentage is given, the generated offset will be aligned to the minimum
1367 ``blocksize`` or to the value of ``offset_align`` if provided. In ZBD mode, value can
1368 also be set as number of zones using 'z'.
1370 .. option:: number_ios=int
1372 Fio will normally perform I/Os until it has exhausted the size of the region
1373 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1374 condition). With this setting, the range/size can be set independently of
1375 the number of I/Os to perform. When fio reaches this number, it will exit
1376 normally and report status. Note that this does not extend the amount of I/O
1377 that will be done, it will only stop fio if this condition is met before
1378 other end-of-job criteria.
1380 .. option:: fsync=int
1382 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1383 the dirty data for every number of blocks given. For example, if you give 32
1384 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1385 using non-buffered I/O, we may not sync the file. The exception is the sg
1386 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1387 means fio does not periodically issue and wait for a sync to complete. Also
1388 see :option:`end_fsync` and :option:`fsync_on_close`.
1390 .. option:: fdatasync=int
1392 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1393 not metadata blocks. In Windows, DragonFlyBSD or OSX there is no
1394 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1395 Defaults to 0, which means fio does not periodically issue and wait for a
1396 data-only sync to complete.
1398 .. option:: write_barrier=int
1400 Make every `N-th` write a barrier write.
1402 .. option:: sync_file_range=str:int
1404 Use :manpage:`sync_file_range(2)` for every `int` number of write
1405 operations. Fio will track range of writes that have happened since the last
1406 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1409 SYNC_FILE_RANGE_WAIT_BEFORE
1411 SYNC_FILE_RANGE_WRITE
1413 SYNC_FILE_RANGE_WAIT_AFTER
1415 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1416 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1417 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1420 .. option:: overwrite=bool
1422 If true, writes to a file will always overwrite existing data. If the file
1423 doesn't already exist, it will be created before the write phase begins. If
1424 the file exists and is large enough for the specified write phase, nothing
1425 will be done. Default: false.
1427 .. option:: end_fsync=bool
1429 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1432 .. option:: fsync_on_close=bool
1434 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1435 from :option:`end_fsync` in that it will happen on every file close, not
1436 just at the end of the job. Default: false.
1438 .. option:: rwmixread=int
1440 Percentage of a mixed workload that should be reads. Default: 50.
1442 .. option:: rwmixwrite=int
1444 Percentage of a mixed workload that should be writes. If both
1445 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1446 add up to 100%, the latter of the two will be used to override the
1447 first. This may interfere with a given rate setting, if fio is asked to
1448 limit reads or writes to a certain rate. If that is the case, then the
1449 distribution may be skewed. Default: 50.
1451 .. option:: random_distribution=str:float[:float][,str:float][,str:float]
1453 By default, fio will use a completely uniform random distribution when asked
1454 to perform random I/O. Sometimes it is useful to skew the distribution in
1455 specific ways, ensuring that some parts of the data is more hot than others.
1456 fio includes the following distribution models:
1459 Uniform random distribution
1468 Normal (Gaussian) distribution
1471 Zoned random distribution
1474 Zone absolute random distribution
1476 When using a **zipf** or **pareto** distribution, an input value is also
1477 needed to define the access pattern. For **zipf**, this is the `Zipf
1478 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1479 program, :command:`fio-genzipf`, that can be used visualize what the given input
1480 values will yield in terms of hit rates. If you wanted to use **zipf** with
1481 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1482 option. If a non-uniform model is used, fio will disable use of the random
1483 map. For the **normal** distribution, a normal (Gaussian) deviation is
1484 supplied as a value between 0 and 100.
1486 The second, optional float is allowed for **pareto**, **zipf** and **normal** distributions.
1487 It allows one to set base of distribution in non-default place, giving more control
1488 over most probable outcome. This value is in range [0-1] which maps linearly to
1489 range of possible random values.
1490 Defaults are: random for **pareto** and **zipf**, and 0.5 for **normal**.
1491 If you wanted to use **zipf** with a `theta` of 1.2 centered on 1/4 of allowed value range,
1492 you would use ``random_distribution=zipf:1.2:0.25``.
1494 For a **zoned** distribution, fio supports specifying percentages of I/O
1495 access that should fall within what range of the file or device. For
1496 example, given a criteria of:
1498 * 60% of accesses should be to the first 10%
1499 * 30% of accesses should be to the next 20%
1500 * 8% of accesses should be to the next 30%
1501 * 2% of accesses should be to the next 40%
1503 we can define that through zoning of the random accesses. For the above
1504 example, the user would do::
1506 random_distribution=zoned:60/10:30/20:8/30:2/40
1508 A **zoned_abs** distribution works exactly like the **zoned**, except
1509 that it takes absolute sizes. For example, let's say you wanted to
1510 define access according to the following criteria:
1512 * 60% of accesses should be to the first 20G
1513 * 30% of accesses should be to the next 100G
1514 * 10% of accesses should be to the next 500G
1516 we can define an absolute zoning distribution with:
1518 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1520 For both **zoned** and **zoned_abs**, fio supports defining up to
1523 Similarly to how :option:`bssplit` works for setting ranges and
1524 percentages of block sizes. Like :option:`bssplit`, it's possible to
1525 specify separate zones for reads, writes, and trims. If just one set
1526 is given, it'll apply to all of them. This goes for both **zoned**
1527 **zoned_abs** distributions.
1529 .. option:: percentage_random=int[,int][,int]
1531 For a random workload, set how big a percentage should be random. This
1532 defaults to 100%, in which case the workload is fully random. It can be set
1533 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1534 sequential. Any setting in between will result in a random mix of sequential
1535 and random I/O, at the given percentages. Comma-separated values may be
1536 specified for reads, writes, and trims as described in :option:`blocksize`.
1538 .. option:: norandommap
1540 Normally fio will cover every block of the file when doing random I/O. If
1541 this option is given, fio will just get a new random offset without looking
1542 at past I/O history. This means that some blocks may not be read or written,
1543 and that some blocks may be read/written more than once. If this option is
1544 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1545 only intact blocks are verified, i.e., partially-overwritten blocks are
1546 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1547 the same block to be overwritten, which can cause verification errors. Either
1548 do not use norandommap in this case, or also use the lfsr random generator.
1550 .. option:: softrandommap=bool
1552 See :option:`norandommap`. If fio runs with the random block map enabled and
1553 it fails to allocate the map, if this option is set it will continue without
1554 a random block map. As coverage will not be as complete as with random maps,
1555 this option is disabled by default.
1557 .. option:: random_generator=str
1559 Fio supports the following engines for generating I/O offsets for random I/O:
1562 Strong 2^88 cycle random number generator.
1564 Linear feedback shift register generator.
1566 Strong 64-bit 2^258 cycle random number generator.
1568 **tausworthe** is a strong random number generator, but it requires tracking
1569 on the side if we want to ensure that blocks are only read or written
1570 once. **lfsr** guarantees that we never generate the same offset twice, and
1571 it's also less computationally expensive. It's not a true random generator,
1572 however, though for I/O purposes it's typically good enough. **lfsr** only
1573 works with single block sizes, not with workloads that use multiple block
1574 sizes. If used with such a workload, fio may read or write some blocks
1575 multiple times. The default value is **tausworthe**, unless the required
1576 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1577 selected automatically.
1583 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1585 The block size in bytes used for I/O units. Default: 4096. A single value
1586 applies to reads, writes, and trims. Comma-separated values may be
1587 specified for reads, writes, and trims. A value not terminated in a comma
1588 applies to subsequent types.
1593 means 256k for reads, writes and trims.
1596 means 8k for reads, 32k for writes and trims.
1599 means 8k for reads, 32k for writes, and default for trims.
1602 means default for reads, 8k for writes and trims.
1605 means default for reads, 8k for writes, and default for trims.
1607 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1609 A range of block sizes in bytes for I/O units. The issued I/O unit will
1610 always be a multiple of the minimum size, unless
1611 :option:`blocksize_unaligned` is set.
1613 Comma-separated ranges may be specified for reads, writes, and trims as
1614 described in :option:`blocksize`.
1616 Example: ``bsrange=1k-4k,2k-8k``.
1618 .. option:: bssplit=str[,str][,str]
1620 Sometimes you want even finer grained control of the block sizes
1621 issued, not just an even split between them. This option allows you to
1622 weight various block sizes, so that you are able to define a specific
1623 amount of block sizes issued. The format for this option is::
1625 bssplit=blocksize/percentage:blocksize/percentage
1627 for as many block sizes as needed. So if you want to define a workload
1628 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1631 bssplit=4k/10:64k/50:32k/40
1633 Ordering does not matter. If the percentage is left blank, fio will
1634 fill in the remaining values evenly. So a bssplit option like this one::
1636 bssplit=4k/50:1k/:32k/
1638 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1639 add up to 100, if bssplit is given a range that adds up to more, it
1642 Comma-separated values may be specified for reads, writes, and trims as
1643 described in :option:`blocksize`.
1645 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1646 having 90% 4k writes and 10% 8k writes, you would specify::
1648 bssplit=2k/50:4k/50,4k/90:8k/10
1650 Fio supports defining up to 64 different weights for each data
1653 .. option:: blocksize_unaligned, bs_unaligned
1655 If set, fio will issue I/O units with any size within
1656 :option:`blocksize_range`, not just multiples of the minimum size. This
1657 typically won't work with direct I/O, as that normally requires sector
1660 .. option:: bs_is_seq_rand=bool
1662 If this option is set, fio will use the normal read,write blocksize settings
1663 as sequential,random blocksize settings instead. Any random read or write
1664 will use the WRITE blocksize settings, and any sequential read or write will
1665 use the READ blocksize settings.
1667 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1669 Boundary to which fio will align random I/O units. Default:
1670 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1671 I/O, though it usually depends on the hardware block size. This option is
1672 mutually exclusive with using a random map for files, so it will turn off
1673 that option. Comma-separated values may be specified for reads, writes, and
1674 trims as described in :option:`blocksize`.
1680 .. option:: zero_buffers
1682 Initialize buffers with all zeros. Default: fill buffers with random data.
1684 .. option:: refill_buffers
1686 If this option is given, fio will refill the I/O buffers on every
1687 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1688 naturally. Defaults to being unset i.e., the buffer is only filled at
1689 init time and the data in it is reused when possible but if any of
1690 :option:`verify`, :option:`buffer_compress_percentage` or
1691 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1692 automatically enabled.
1694 .. option:: scramble_buffers=bool
1696 If :option:`refill_buffers` is too costly and the target is using data
1697 deduplication, then setting this option will slightly modify the I/O buffer
1698 contents to defeat normal de-dupe attempts. This is not enough to defeat
1699 more clever block compression attempts, but it will stop naive dedupe of
1700 blocks. Default: true.
1702 .. option:: buffer_compress_percentage=int
1704 If this is set, then fio will attempt to provide I/O buffer content
1705 (on WRITEs) that compresses to the specified level. Fio does this by
1706 providing a mix of random data followed by fixed pattern data. The
1707 fixed pattern is either zeros, or the pattern specified by
1708 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1709 might skew the compression ratio slightly. Setting
1710 `buffer_compress_percentage` to a value other than 100 will also
1711 enable :option:`refill_buffers` in order to reduce the likelihood that
1712 adjacent blocks are so similar that they over compress when seen
1713 together. See :option:`buffer_compress_chunk` for how to set a finer or
1714 coarser granularity for the random/fixed data region. Defaults to unset
1715 i.e., buffer data will not adhere to any compression level.
1717 .. option:: buffer_compress_chunk=int
1719 This setting allows fio to manage how big the random/fixed data region
1720 is when using :option:`buffer_compress_percentage`. When
1721 `buffer_compress_chunk` is set to some non-zero value smaller than the
1722 block size, fio can repeat the random/fixed region throughout the I/O
1723 buffer at the specified interval (which particularly useful when
1724 bigger block sizes are used for a job). When set to 0, fio will use a
1725 chunk size that matches the block size resulting in a single
1726 random/fixed region within the I/O buffer. Defaults to 512. When the
1727 unit is omitted, the value is interpreted in bytes.
1729 .. option:: buffer_pattern=str
1731 If set, fio will fill the I/O buffers with this pattern or with the contents
1732 of a file. If not set, the contents of I/O buffers are defined by the other
1733 options related to buffer contents. The setting can be any pattern of bytes,
1734 and can be prefixed with 0x for hex values. It may also be a string, where
1735 the string must then be wrapped with ``""``. Or it may also be a filename,
1736 where the filename must be wrapped with ``''`` in which case the file is
1737 opened and read. Note that not all the file contents will be read if that
1738 would cause the buffers to overflow. So, for example::
1740 buffer_pattern='filename'
1744 buffer_pattern="abcd"
1752 buffer_pattern=0xdeadface
1754 Also you can combine everything together in any order::
1756 buffer_pattern=0xdeadface"abcd"-12'filename'
1758 .. option:: dedupe_percentage=int
1760 If set, fio will generate this percentage of identical buffers when
1761 writing. These buffers will be naturally dedupable. The contents of the
1762 buffers depend on what other buffer compression settings have been set. It's
1763 possible to have the individual buffers either fully compressible, or not at
1764 all -- this option only controls the distribution of unique buffers. Setting
1765 this option will also enable :option:`refill_buffers` to prevent every buffer
1768 .. option:: dedupe_mode=str
1770 If ``dedupe_percentage=<int>`` is given, then this option controls how fio
1771 generates the dedupe buffers.
1774 Generate dedupe buffers by repeating previous writes
1776 Generate dedupe buffers from working set
1778 ``repeat`` is the default option for fio. Dedupe buffers are generated
1779 by repeating previous unique write.
1781 ``working_set`` is a more realistic workload.
1782 With ``working_set``, ``dedupe_working_set_percentage=<int>`` should be provided.
1783 Given that, fio will use the initial unique write buffers as its working set.
1784 Upon deciding to dedupe, fio will randomly choose a buffer from the working set.
1785 Note that by using ``working_set`` the dedupe percentage will converge
1786 to the desired over time while ``repeat`` maintains the desired percentage
1789 .. option:: dedupe_working_set_percentage=int
1791 If ``dedupe_mode=<str>`` is set to ``working_set``, then this controls
1792 the percentage of size of the file or device used as the buffers
1793 fio will choose to generate the dedupe buffers from
1795 Note that size needs to be explicitly provided and only 1 file per
1798 .. option:: dedupe_global=bool
1800 This controls whether the deduplication buffers will be shared amongst
1801 all jobs that have this option set. The buffers are spread evenly between
1804 .. option:: invalidate=bool
1806 Invalidate the buffer/page cache parts of the files to be used prior to
1807 starting I/O if the platform and file type support it. Defaults to true.
1808 This will be ignored if :option:`pre_read` is also specified for the
1811 .. option:: sync=str
1813 Whether, and what type, of synchronous I/O to use for writes. The allowed
1817 Do not use synchronous IO, the default.
1823 Use synchronous file IO. For the majority of I/O engines,
1824 this means using O_SYNC.
1830 Use synchronous data IO. For the majority of I/O engines,
1831 this means using O_DSYNC.
1834 .. option:: iomem=str, mem=str
1836 Fio can use various types of memory as the I/O unit buffer. The allowed
1840 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1844 Use shared memory as the buffers. Allocated through
1845 :manpage:`shmget(2)`.
1848 Same as shm, but use huge pages as backing.
1851 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1852 be file backed if a filename is given after the option. The format
1853 is `mem=mmap:/path/to/file`.
1856 Use a memory mapped huge file as the buffer backing. Append filename
1857 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1860 Same as mmap, but use a MMAP_SHARED mapping.
1863 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1864 The :option:`ioengine` must be `rdma`.
1866 The area allocated is a function of the maximum allowed bs size for the job,
1867 multiplied by the I/O depth given. Note that for **shmhuge** and
1868 **mmaphuge** to work, the system must have free huge pages allocated. This
1869 can normally be checked and set by reading/writing
1870 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1871 is 2 or 4MiB in size depending on the platform. So to calculate the
1872 number of huge pages you need for a given job file, add up the I/O
1873 depth of all jobs (normally one unless :option:`iodepth` is used) and
1874 multiply by the maximum bs set. Then divide that number by the huge
1875 page size. You can see the size of the huge pages in
1876 :file:`/proc/meminfo`. If no huge pages are allocated by having a
1877 non-zero number in `nr_hugepages`, using **mmaphuge** or **shmhuge**
1878 will fail. Also see :option:`hugepage-size`.
1880 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1881 should point there. So if it's mounted in :file:`/huge`, you would use
1882 `mem=mmaphuge:/huge/somefile`.
1884 .. option:: iomem_align=int, mem_align=int
1886 This indicates the memory alignment of the I/O memory buffers. Note that
1887 the given alignment is applied to the first I/O unit buffer, if using
1888 :option:`iodepth` the alignment of the following buffers are given by the
1889 :option:`bs` used. In other words, if using a :option:`bs` that is a
1890 multiple of the page sized in the system, all buffers will be aligned to
1891 this value. If using a :option:`bs` that is not page aligned, the alignment
1892 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1895 .. option:: hugepage-size=int
1897 Defines the size of a huge page. Must at least be equal to the system
1898 setting, see :file:`/proc/meminfo` and
1899 :file:`/sys/kernel/mm/hugepages/`. Defaults to 2 or 4MiB depending on
1900 the platform. Should probably always be a multiple of megabytes, so
1901 using ``hugepage-size=Xm`` is the preferred way to set this to avoid
1902 setting a non-pow-2 bad value.
1904 .. option:: lockmem=int
1906 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1907 simulate a smaller amount of memory. The amount specified is per worker.
1913 .. option:: size=int
1915 The total size of file I/O for each thread of this job. Fio will run until
1916 this many bytes has been transferred, unless runtime is altered by other means
1917 such as (1) :option:`runtime`, (2) :option:`io_size` (3) :option:`number_ios`,
1918 (4) gaps/holes while doing I/O's such as ``rw=read:16K``, or (5) sequential
1919 I/O reaching end of the file which is possible when :option:`percentage_random`
1921 Fio will divide this size between the available files determined by options
1922 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1923 specified by the job. If the result of division happens to be 0, the size is
1924 set to the physical size of the given files or devices if they exist.
1925 If this option is not specified, fio will use the full size of the given
1926 files or devices. If the files do not exist, size must be given. It is also
1927 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1928 given, fio will use 20% of the full size of the given files or devices.
1929 In ZBD mode, value can also be set as number of zones using 'z'.
1930 Can be combined with :option:`offset` to constrain the start and end range
1931 that I/O will be done within.
1933 .. option:: io_size=int, io_limit=int
1935 Normally fio operates within the region set by :option:`size`, which means
1936 that the :option:`size` option sets both the region and size of I/O to be
1937 performed. Sometimes that is not what you want. With this option, it is
1938 possible to define just the amount of I/O that fio should do. For instance,
1939 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1940 will perform I/O within the first 20GiB but exit when 5GiB have been
1941 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1942 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1943 the 0..20GiB region.
1945 .. option:: filesize=irange(int)
1947 Individual file sizes. May be a range, in which case fio will select sizes for
1948 files at random within the given range. If not given, each created file is the
1949 same size. This option overrides :option:`size` in terms of file size, i.e. if
1950 :option:`filesize` is specified then :option:`size` becomes merely the default
1951 for :option:`io_size` and has no effect at all if :option:`io_size` is set
1954 .. option:: file_append=bool
1956 Perform I/O after the end of the file. Normally fio will operate within the
1957 size of a file. If this option is set, then fio will append to the file
1958 instead. This has identical behavior to setting :option:`offset` to the size
1959 of a file. This option is ignored on non-regular files.
1961 .. option:: fill_device=bool, fill_fs=bool
1963 Sets size to something really large and waits for ENOSPC (no space left on
1964 device) or EDQUOT (disk quota exceeded)
1965 as the terminating condition. Only makes sense with sequential
1966 write. For a read workload, the mount point will be filled first then I/O
1967 started on the result. This option doesn't make sense if operating on a raw
1968 device node, since the size of that is already known by the file system.
1969 Additionally, writing beyond end-of-device will not return ENOSPC there.
1975 .. option:: ioengine=str
1977 Defines how the job issues I/O to the file. The following types are defined:
1980 Basic :manpage:`read(2)` or :manpage:`write(2)`
1981 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1982 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1985 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1986 all supported operating systems except for Windows.
1989 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1990 queuing by coalescing adjacent I/Os into a single submission.
1993 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1996 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1999 Fast Linux native asynchronous I/O. Supports async IO
2000 for both direct and buffered IO.
2001 This engine defines engine specific options.
2004 Fast Linux native asynchronous I/O for pass through commands.
2005 This engine defines engine specific options.
2008 Linux native asynchronous I/O. Note that Linux may only support
2009 queued behavior with non-buffered I/O (set ``direct=1`` or
2011 This engine defines engine specific options.
2014 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
2015 :manpage:`aio_write(3)`.
2018 Solaris native asynchronous I/O.
2021 Windows native asynchronous I/O. Default on Windows.
2024 File is memory mapped with :manpage:`mmap(2)` and data copied
2025 to/from using :manpage:`memcpy(3)`.
2028 :manpage:`splice(2)` is used to transfer the data and
2029 :manpage:`vmsplice(2)` to transfer data from user space to the
2033 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
2034 ioctl, or if the target is an sg character device we use
2035 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
2036 I/O. Requires :option:`filename` option to specify either block or
2037 character devices. This engine supports trim operations.
2038 The sg engine includes engine specific options.
2041 Read, write, trim and ZBC/ZAC operations to a zoned
2042 block device using libzbc library. The target can be
2043 either an SG character device or a block device file.
2046 Doesn't transfer any data, just pretends to. This is mainly used to
2047 exercise fio itself and for debugging/testing purposes.
2050 Transfer over the network to given ``host:port``. Depending on the
2051 :option:`protocol` used, the :option:`hostname`, :option:`port`,
2052 :option:`listen` and :option:`filename` options are used to specify
2053 what sort of connection to make, while the :option:`protocol` option
2054 determines which protocol will be used. This engine defines engine
2058 Like **net**, but uses :manpage:`splice(2)` and
2059 :manpage:`vmsplice(2)` to map data and send/receive.
2060 This engine defines engine specific options.
2063 Doesn't transfer any data, but burns CPU cycles according to the
2064 :option:`cpuload`, :option:`cpuchunks` and :option:`cpumode` options.
2065 Setting :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
2066 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
2067 to get desired CPU usage, as the cpuload only loads a
2068 single CPU at the desired rate. A job never finishes unless there is
2069 at least one non-cpuio job.
2070 Setting :option:`cpumode`\=qsort replace the default noop instructions loop
2071 by a qsort algorithm to consume more energy.
2074 The RDMA I/O engine supports both RDMA memory semantics
2075 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
2076 InfiniBand, RoCE and iWARP protocols. This engine defines engine
2080 I/O engine that does regular fallocate to simulate data transfer as
2084 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
2087 does fallocate(,mode = 0).
2090 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
2093 I/O engine that sends :manpage:`ftruncate(2)` operations in response
2094 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
2095 size to the current block offset. :option:`blocksize` is ignored.
2098 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
2099 defragment activity in request to DDIR_WRITE event.
2102 I/O engine supporting direct access to Ceph Reliable Autonomic
2103 Distributed Object Store (RADOS) via librados. This ioengine
2104 defines engine specific options.
2107 I/O engine supporting direct access to Ceph Rados Block Devices
2108 (RBD) via librbd without the need to use the kernel rbd driver. This
2109 ioengine defines engine specific options.
2112 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
2113 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
2115 This engine only supports direct IO of iodepth=1; you need to scale this
2116 via numjobs. blocksize defines the size of the objects to be created.
2118 TRIM is translated to object deletion.
2121 Using GlusterFS libgfapi sync interface to direct access to
2122 GlusterFS volumes without having to go through FUSE. This ioengine
2123 defines engine specific options.
2126 Using GlusterFS libgfapi async interface to direct access to
2127 GlusterFS volumes without having to go through FUSE. This ioengine
2128 defines engine specific options.
2131 Read and write through Hadoop (HDFS). The :option:`filename` option
2132 is used to specify host,port of the hdfs name-node to connect. This
2133 engine interprets offsets a little differently. In HDFS, files once
2134 created cannot be modified so random writes are not possible. To
2135 imitate this the libhdfs engine expects a bunch of small files to be
2136 created over HDFS and will randomly pick a file from them
2137 based on the offset generated by fio backend (see the example
2138 job file to create such files, use ``rw=write`` option). Please
2139 note, it may be necessary to set environment variables to work
2140 with HDFS/libhdfs properly. Each job uses its own connection to
2144 Read, write and erase an MTD character device (e.g.,
2145 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
2146 underlying device type, the I/O may have to go in a certain pattern,
2147 e.g., on NAND, writing sequentially to erase blocks and discarding
2148 before overwriting. The `trimwrite` mode works well for this
2152 Read and write using device DAX to a persistent memory device (e.g.,
2153 /dev/dax0.0) through the PMDK libpmem library.
2156 Prefix to specify loading an external I/O engine object file. Append
2157 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
2158 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
2159 absolute or relative. See :file:`engines/skeleton_external.c` for
2160 details of writing an external I/O engine.
2163 Simply create the files and do no I/O to them. You still need to
2164 set `filesize` so that all the accounting still occurs, but no
2165 actual I/O will be done other than creating the file.
2168 Simply do stat() and do no I/O to the file. You need to set 'filesize'
2169 and 'nrfiles', so that files will be created.
2170 This engine is to measure file lookup and meta data access.
2173 Simply delete the files by unlink() and do no I/O to them. You need to set 'filesize'
2174 and 'nrfiles', so that the files will be created.
2175 This engine is to measure file delete.
2178 Read and write using mmap I/O to a file on a filesystem
2179 mounted with DAX on a persistent memory device through the PMDK
2183 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2184 This engine is very basic and issues calls to IME whenever an IO is
2188 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2189 This engine uses iovecs and will try to stack as much IOs as possible
2190 (if the IOs are "contiguous" and the IO depth is not exceeded)
2191 before issuing a call to IME.
2194 Asynchronous read and write using DDN's Infinite Memory Engine (IME).
2195 This engine will try to stack as much IOs as possible by creating
2196 requests for IME. FIO will then decide when to commit these requests.
2199 Read and write iscsi lun with libiscsi.
2202 Read and write a Network Block Device (NBD).
2205 I/O engine supporting libcufile synchronous access to nvidia-fs and a
2206 GPUDirect Storage-supported filesystem. This engine performs
2207 I/O without transferring buffers between user-space and the kernel,
2208 unless :option:`verify` is set or :option:`cuda_io` is `posix`.
2209 :option:`iomem` must not be `cudamalloc`. This ioengine defines
2210 engine specific options.
2213 I/O engine supporting asynchronous read and write operations to the
2214 DAOS File System (DFS) via libdfs.
2217 I/O engine supporting asynchronous read and write operations to
2218 NFS filesystems from userspace via libnfs. This is useful for
2219 achieving higher concurrency and thus throughput than is possible
2223 Execute 3rd party tools. Could be used to perform monitoring during jobs runtime.
2226 I/O engine using the xNVMe C API, for NVMe devices. The xnvme engine provides
2227 flexibility to access GNU/Linux Kernel NVMe driver via libaio, IOCTLs, io_uring,
2228 the SPDK NVMe driver, or your own custom NVMe driver. The xnvme engine includes
2229 engine specific options. (See https://xnvme.io).
2232 Use the libblkio library
2233 (https://gitlab.com/libblkio/libblkio). The specific
2234 *driver* to use must be set using
2235 :option:`libblkio_driver`. If
2236 :option:`mem`/:option:`iomem` is not specified, memory
2237 allocation is delegated to libblkio (and so is
2238 guaranteed to work with the selected *driver*). One
2239 libblkio instance is used per process, so all jobs
2240 setting option :option:`thread` will share a single
2241 instance (with one queue per thread) and must specify
2242 compatible options. Note that some drivers don't allow
2243 several instances to access the same device or file
2244 simultaneously, but allow it for threads.
2246 I/O engine specific parameters
2247 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2249 In addition, there are some parameters which are only valid when a specific
2250 :option:`ioengine` is in use. These are used identically to normal parameters,
2251 with the caveat that when used on the command line, they must come after the
2252 :option:`ioengine` that defines them is selected.
2254 .. option:: cmdprio_percentage=int[,int] : [io_uring] [libaio]
2256 Set the percentage of I/O that will be issued with the highest priority.
2257 Default: 0. A single value applies to reads and writes. Comma-separated
2258 values may be specified for reads and writes. For this option to be
2259 effective, NCQ priority must be supported and enabled, and the :option:`direct`
2260 option must be set. fio must also be run as the root user. Unlike
2261 slat/clat/lat stats, which can be tracked and reported independently, per
2262 priority stats only track and report a single type of latency. By default,
2263 completion latency (clat) will be reported, if :option:`lat_percentiles` is
2264 set, total latency (lat) will be reported.
2266 .. option:: cmdprio_class=int[,int] : [io_uring] [libaio]
2268 Set the I/O priority class to use for I/Os that must be issued with
2269 a priority when :option:`cmdprio_percentage` or
2270 :option:`cmdprio_bssplit` is set. If not specified when
2271 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2272 this defaults to the highest priority class. A single value applies
2273 to reads and writes. Comma-separated values may be specified for
2274 reads and writes. See :manpage:`ionice(1)`. See also the
2275 :option:`prioclass` option.
2277 .. option:: cmdprio=int[,int] : [io_uring] [libaio]
2279 Set the I/O priority value to use for I/Os that must be issued with
2280 a priority when :option:`cmdprio_percentage` or
2281 :option:`cmdprio_bssplit` is set. If not specified when
2282 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2284 Linux limits us to a positive value between 0 and 7, with 0 being the
2285 highest. A single value applies to reads and writes. Comma-separated
2286 values may be specified for reads and writes. See :manpage:`ionice(1)`.
2287 Refer to an appropriate manpage for other operating systems since
2288 meaning of priority may differ. See also the :option:`prio` option.
2290 .. option:: cmdprio_bssplit=str[,str] : [io_uring] [libaio]
2292 To get a finer control over I/O priority, this option allows
2293 specifying the percentage of IOs that must have a priority set
2294 depending on the block size of the IO. This option is useful only
2295 when used together with the :option:`bssplit` option, that is,
2296 multiple different block sizes are used for reads and writes.
2298 The first accepted format for this option is the same as the format of
2299 the :option:`bssplit` option:
2301 cmdprio_bssplit=blocksize/percentage:blocksize/percentage
2303 In this case, each entry will use the priority class and priority
2304 level defined by the options :option:`cmdprio_class` and
2305 :option:`cmdprio` respectively.
2307 The second accepted format for this option is:
2309 cmdprio_bssplit=blocksize/percentage/class/level:blocksize/percentage/class/level
2311 In this case, the priority class and priority level is defined inside
2312 each entry. In comparison with the first accepted format, the second
2313 accepted format does not restrict all entries to have the same priority
2314 class and priority level.
2316 For both formats, only the read and write data directions are supported,
2317 values for trim IOs are ignored. This option is mutually exclusive with
2318 the :option:`cmdprio_percentage` option.
2320 .. option:: fixedbufs : [io_uring] [io_uring_cmd]
2322 If fio is asked to do direct IO, then Linux will map pages for each
2323 IO call, and release them when IO is done. If this option is set, the
2324 pages are pre-mapped before IO is started. This eliminates the need to
2325 map and release for each IO. This is more efficient, and reduces the
2328 .. option:: nonvectored=int : [io_uring] [io_uring_cmd]
2330 With this option, fio will use non-vectored read/write commands, where
2331 address must contain the address directly. Default is -1.
2333 .. option:: force_async=int : [io_uring] [io_uring_cmd]
2335 Normal operation for io_uring is to try and issue an sqe as
2336 non-blocking first, and if that fails, execute it in an async manner.
2337 With this option set to N, then every N request fio will ask sqe to
2338 be issued in an async manner. Default is 0.
2340 .. option:: registerfiles : [io_uring] [io_uring_cmd]
2342 With this option, fio registers the set of files being used with the
2343 kernel. This avoids the overhead of managing file counts in the kernel,
2344 making the submission and completion part more lightweight. Required
2345 for the below :option:`sqthread_poll` option.
2347 .. option:: sqthread_poll : [io_uring] [io_uring_cmd] [xnvme]
2349 Normally fio will submit IO by issuing a system call to notify the
2350 kernel of available items in the SQ ring. If this option is set, the
2351 act of submitting IO will be done by a polling thread in the kernel.
2352 This frees up cycles for fio, at the cost of using more CPU in the
2353 system. As submission is just the time it takes to fill in the sqe
2354 entries and any syscall required to wake up the idle kernel thread,
2355 fio will not report submission latencies.
2357 .. option:: sqthread_poll_cpu=int : [io_uring] [io_uring_cmd]
2359 When :option:`sqthread_poll` is set, this option provides a way to
2360 define which CPU should be used for the polling thread.
2362 .. option:: cmd_type=str : [io_uring_cmd]
2364 Specifies the type of uring passthrough command to be used. Supported
2365 value is nvme. Default is nvme.
2369 [io_uring] [io_uring_cmd] [xnvme]
2371 If this option is set, fio will attempt to use polled IO completions.
2372 Normal IO completions generate interrupts to signal the completion of
2373 IO, polled completions do not. Hence they are require active reaping
2374 by the application. The benefits are more efficient IO for high IOPS
2375 scenarios, and lower latencies for low queue depth IO.
2379 Use poll queues. This is incompatible with
2380 :option:`libblkio_wait_mode=eventfd <libblkio_wait_mode>` and
2381 :option:`libblkio_force_enable_completion_eventfd`.
2385 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2390 If this option is set, fio will attempt to use polled IO completions.
2391 This will have a similar effect as (io_uring)hipri. Only SCSI READ and
2392 WRITE commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor
2393 VERIFY). Older versions of the Linux sg driver that do not support
2394 hipri will simply ignore this flag and do normal IO. The Linux SCSI
2395 Low Level Driver (LLD) that "owns" the device also needs to support
2396 hipri (also known as iopoll and mq_poll). The MegaRAID driver is an
2397 example of a SCSI LLD. Default: clear (0) which does normal
2398 (interrupted based) IO.
2400 .. option:: userspace_reap : [libaio]
2402 Normally, with the libaio engine in use, fio will use the
2403 :manpage:`io_getevents(2)` system call to reap newly returned events. With
2404 this flag turned on, the AIO ring will be read directly from user-space to
2405 reap events. The reaping mode is only enabled when polling for a minimum of
2406 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
2408 .. option:: hipri_percentage : [pvsync2]
2410 When hipri is set this determines the probability of a pvsync2 I/O being high
2411 priority. The default is 100%.
2413 .. option:: nowait=bool : [pvsync2] [libaio] [io_uring] [io_uring_cmd]
2415 By default if a request cannot be executed immediately (e.g. resource starvation,
2416 waiting on locks) it is queued and the initiating process will be blocked until
2417 the required resource becomes free.
2419 This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and
2420 the call will return instantly with EAGAIN or a partial result rather than waiting.
2422 It is useful to also use ignore_error=EAGAIN when using this option.
2424 Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2.
2425 They return EOPNOTSUP instead of EAGAIN.
2427 For cached I/O, using this option usually means a request operates only with
2428 cached data. Currently the RWF_NOWAIT flag does not supported for cached write.
2430 For direct I/O, requests will only succeed if cache invalidation isn't required,
2431 file blocks are fully allocated and the disk request could be issued immediately.
2433 .. option:: fdp=bool : [io_uring_cmd]
2435 Enable Flexible Data Placement mode for write commands.
2437 .. option:: fdp_pli=str : [io_uring_cmd]
2439 Select which Placement ID Index/Indicies this job is allowed to use for
2440 writes. By default, the job will cycle through all available Placement
2441 IDs, so use this to isolate these identifiers to specific jobs. If you
2442 want fio to use placement identifier only at indices 0, 2 and 5 specify
2445 .. option:: cpuload=int : [cpuio]
2447 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2448 option when using cpuio I/O engine.
2450 .. option:: cpuchunks=int : [cpuio]
2452 Split the load into cycles of the given time. In microseconds.
2454 .. option:: cpumode=str : [cpuio]
2456 Specify how to stress the CPU. It can take these two values:
2459 This is the default where the CPU executes noop instructions.
2461 Replace the default noop instructions loop with a qsort algorithm to
2462 consume more energy.
2464 .. option:: exit_on_io_done=bool : [cpuio]
2466 Detect when I/O threads are done, then exit.
2468 .. option:: namenode=str : [libhdfs]
2470 The hostname or IP address of a HDFS cluster namenode to contact.
2472 .. option:: port=int
2476 The listening port of the HFDS cluster namenode.
2480 The TCP or UDP port to bind to or connect to. If this is used with
2481 :option:`numjobs` to spawn multiple instances of the same job type, then
2482 this will be the starting port number since fio will use a range of
2487 The port to use for RDMA-CM communication. This should be the same value
2488 on the client and the server side.
2490 .. option:: hostname=str : [netsplice] [net] [rdma]
2492 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
2493 is a TCP listener or UDP reader, the hostname is not used and must be omitted
2494 unless it is a valid UDP multicast address.
2496 .. option:: serverip=str : [librpma_*]
2498 The IP address to be used for RDMA-CM based I/O.
2500 .. option:: direct_write_to_pmem=bool : [librpma_*]
2502 Set to 1 only when Direct Write to PMem from the remote host is possible.
2503 Otherwise, set to 0.
2505 .. option:: busy_wait_polling=bool : [librpma_*_server]
2507 Set to 0 to wait for completion instead of busy-wait polling completion.
2510 .. option:: interface=str : [netsplice] [net]
2512 The IP address of the network interface used to send or receive UDP
2515 .. option:: ttl=int : [netsplice] [net]
2517 Time-to-live value for outgoing UDP multicast packets. Default: 1.
2519 .. option:: nodelay=bool : [netsplice] [net]
2521 Set TCP_NODELAY on TCP connections.
2523 .. option:: protocol=str, proto=str : [netsplice] [net]
2525 The network protocol to use. Accepted values are:
2528 Transmission control protocol.
2530 Transmission control protocol V6.
2532 User datagram protocol.
2534 User datagram protocol V6.
2538 When the protocol is TCP or UDP, the port must also be given, as well as the
2539 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
2540 normal :option:`filename` option should be used and the port is invalid.
2542 .. option:: listen : [netsplice] [net]
2544 For TCP network connections, tell fio to listen for incoming connections
2545 rather than initiating an outgoing connection. The :option:`hostname` must
2546 be omitted if this option is used.
2548 .. option:: pingpong : [netsplice] [net]
2550 Normally a network writer will just continue writing data, and a network
2551 reader will just consume packages. If ``pingpong=1`` is set, a writer will
2552 send its normal payload to the reader, then wait for the reader to send the
2553 same payload back. This allows fio to measure network latencies. The
2554 submission and completion latencies then measure local time spent sending or
2555 receiving, and the completion latency measures how long it took for the
2556 other end to receive and send back. For UDP multicast traffic
2557 ``pingpong=1`` should only be set for a single reader when multiple readers
2558 are listening to the same address.
2560 .. option:: window_size : [netsplice] [net]
2562 Set the desired socket buffer size for the connection.
2564 .. option:: mss : [netsplice] [net]
2566 Set the TCP maximum segment size (TCP_MAXSEG).
2568 .. option:: donorname=str : [e4defrag]
2570 File will be used as a block donor (swap extents between files).
2572 .. option:: inplace=int : [e4defrag]
2574 Configure donor file blocks allocation strategy:
2577 Default. Preallocate donor's file on init.
2579 Allocate space immediately inside defragment event, and free right
2582 .. option:: clustername=str : [rbd,rados]
2584 Specifies the name of the Ceph cluster.
2586 .. option:: rbdname=str : [rbd]
2588 Specifies the name of the RBD.
2590 .. option:: clientname=str : [rbd,rados]
2592 Specifies the username (without the 'client.' prefix) used to access the
2593 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2594 the full *type.id* string. If no type. prefix is given, fio will add
2595 'client.' by default.
2597 .. option:: conf=str : [rados]
2599 Specifies the configuration path of ceph cluster, so conf file does not
2600 have to be /etc/ceph/ceph.conf.
2602 .. option:: busy_poll=bool : [rbd,rados]
2604 Poll store instead of waiting for completion. Usually this provides better
2605 throughput at cost of higher(up to 100%) CPU utilization.
2607 .. option:: touch_objects=bool : [rados]
2609 During initialization, touch (create if do not exist) all objects (files).
2610 Touching all objects affects ceph caches and likely impacts test results.
2613 .. option:: pool=str :
2617 Specifies the name of the Ceph pool containing RBD or RADOS data.
2621 Specify the label or UUID of the DAOS pool to connect to.
2623 .. option:: cont=str : [dfs]
2625 Specify the label or UUID of the DAOS container to open.
2627 .. option:: chunk_size=int
2631 Specify a different chunk size (in bytes) for the dfs file.
2632 Use DAOS container's chunk size by default.
2636 The size of the chunk to use for each file.
2638 .. option:: object_class=str : [dfs]
2640 Specify a different object class for the dfs file.
2641 Use DAOS container's object class by default.
2643 .. option:: skip_bad=bool : [mtd]
2645 Skip operations against known bad blocks.
2647 .. option:: hdfsdirectory : [libhdfs]
2649 libhdfs will create chunk in this HDFS directory.
2651 .. option:: verb=str : [rdma]
2653 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2654 values are write, read, send and recv. These correspond to the equivalent
2655 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2656 specified on the client side of the connection. See the examples folder.
2658 .. option:: bindname=str : [rdma]
2660 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2661 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2662 will be passed into the rdma_bind_addr() function and on the client site it
2663 will be used in the rdma_resolve_add() function. This can be useful when
2664 multiple paths exist between the client and the server or in certain loopback
2667 .. option:: stat_type=str : [filestat]
2669 Specify stat system call type to measure lookup/getattr performance.
2670 Default is **stat** for :manpage:`stat(2)`.
2672 .. option:: readfua=bool : [sg]
2674 With readfua option set to 1, read operations include
2675 the force unit access (fua) flag. Default is 0.
2677 .. option:: writefua=bool : [sg]
2679 With writefua option set to 1, write operations include
2680 the force unit access (fua) flag. Default is 0.
2682 .. option:: sg_write_mode=str : [sg]
2684 Specify the type of write commands to issue. This option can take three values:
2687 This is the default where write opcodes are issued as usual.
2688 **write_and_verify**
2689 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2690 directs the device to carry out a medium verification with no data
2691 comparison. The writefua option is ignored with this selection.
2693 This option is deprecated. Use write_and_verify instead.
2695 Issue WRITE SAME commands. This transfers a single block to the device
2696 and writes this same block of data to a contiguous sequence of LBAs
2697 beginning at the specified offset. fio's block size parameter specifies
2698 the amount of data written with each command. However, the amount of data
2699 actually transferred to the device is equal to the device's block
2700 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2701 write 16 sectors with each command. fio will still generate 8k of data
2702 for each command but only the first 512 bytes will be used and
2703 transferred to the device. The writefua option is ignored with this
2706 This option is deprecated. Use write_same instead.
2708 Issue WRITE SAME(16) commands as above but with the No Data Output
2709 Buffer (NDOB) bit set. No data will be transferred to the device with
2710 this bit set. Data written will be a pre-determined pattern such as
2713 Issue WRITE STREAM(16) commands. Use the **stream_id** option to specify
2714 the stream identifier.
2715 **verify_bytchk_00**
2716 Issue VERIFY commands with BYTCHK set to 00. This directs the
2717 device to carry out a medium verification with no data comparison.
2718 **verify_bytchk_01**
2719 Issue VERIFY commands with BYTCHK set to 01. This directs the device to
2720 compare the data on the device with the data transferred to the device.
2721 **verify_bytchk_11**
2722 Issue VERIFY commands with BYTCHK set to 11. This transfers a
2723 single block to the device and compares the contents of this block with the
2724 data on the device beginning at the specified offset. fio's block size
2725 parameter specifies the total amount of data compared with this command.
2726 However, only one block (sector) worth of data is transferred to the device.
2727 This is similar to the WRITE SAME command except that data is compared instead
2730 .. option:: stream_id=int : [sg]
2732 Set the stream identifier for WRITE STREAM commands. If this is set to 0 (which is not
2733 a valid stream identifier) fio will open a stream and then close it when done. Default
2736 .. option:: http_host=str : [http]
2738 Hostname to connect to. For S3, this could be the bucket hostname.
2739 Default is **localhost**
2741 .. option:: http_user=str : [http]
2743 Username for HTTP authentication.
2745 .. option:: http_pass=str : [http]
2747 Password for HTTP authentication.
2749 .. option:: https=str : [http]
2751 Enable HTTPS instead of http. *on* enables HTTPS; *insecure*
2752 will enable HTTPS, but disable SSL peer verification (use with
2753 caution!). Default is **off**
2755 .. option:: http_mode=str : [http]
2757 Which HTTP access mode to use: *webdav*, *swift*, or *s3*.
2758 Default is **webdav**
2760 .. option:: http_s3_region=str : [http]
2762 The S3 region/zone string.
2763 Default is **us-east-1**
2765 .. option:: http_s3_key=str : [http]
2769 .. option:: http_s3_keyid=str : [http]
2771 The S3 key/access id.
2773 .. option:: http_s3_sse_customer_key=str : [http]
2775 The encryption customer key in SSE server side.
2777 .. option:: http_s3_sse_customer_algorithm=str : [http]
2779 The encryption customer algorithm in SSE server side.
2780 Default is **AES256**
2782 .. option:: http_s3_storage_class=str : [http]
2784 Which storage class to access. User-customizable settings.
2785 Default is **STANDARD**
2787 .. option:: http_swift_auth_token=str : [http]
2789 The Swift auth token. See the example configuration file on how
2792 .. option:: http_verbose=int : [http]
2794 Enable verbose requests from libcurl. Useful for debugging. 1
2795 turns on verbose logging from libcurl, 2 additionally enables
2796 HTTP IO tracing. Default is **0**
2798 .. option:: uri=str : [nbd]
2800 Specify the NBD URI of the server to test. The string
2801 is a standard NBD URI
2802 (see https://github.com/NetworkBlockDevice/nbd/tree/master/doc).
2803 Example URIs: nbd://localhost:10809
2804 nbd+unix:///?socket=/tmp/socket
2805 nbds://tlshost/exportname
2807 .. option:: gpu_dev_ids=str : [libcufile]
2809 Specify the GPU IDs to use with CUDA. This is a colon-separated list of
2810 int. GPUs are assigned to workers roundrobin. Default is 0.
2812 .. option:: cuda_io=str : [libcufile]
2814 Specify the type of I/O to use with CUDA. Default is **cufile**.
2817 Use libcufile and nvidia-fs. This option performs I/O directly
2818 between a GPUDirect Storage filesystem and GPU buffers,
2819 avoiding use of a bounce buffer. If :option:`verify` is set,
2820 cudaMemcpy is used to copy verificaton data between RAM and GPU.
2821 Verification data is copied from RAM to GPU before a write
2822 and from GPU to RAM after a read. :option:`direct` must be 1.
2824 Use POSIX to perform I/O with a RAM buffer, and use cudaMemcpy
2825 to transfer data between RAM and the GPUs. Data is copied from
2826 GPU to RAM before a write and copied from RAM to GPU after a
2827 read. :option:`verify` does not affect use of cudaMemcpy.
2829 .. option:: nfs_url=str : [nfs]
2831 URL in libnfs format, eg nfs://<server|ipv4|ipv6>/path[?arg=val[&arg=val]*]
2832 Refer to the libnfs README for more details.
2834 .. option:: program=str : [exec]
2836 Specify the program to execute.
2838 .. option:: arguments=str : [exec]
2840 Specify arguments to pass to program.
2841 Some special variables can be expanded to pass fio's job details to the program.
2844 Replaced by the duration of the job in seconds.
2846 Replaced by the name of the job.
2848 .. option:: grace_time=int : [exec]
2850 Specify the time between the SIGTERM and SIGKILL signals. Default is 1 second.
2852 .. option:: std_redirect=bool : [exec]
2854 If set, stdout and stderr streams are redirected to files named from the job name. Default is true.
2856 .. option:: xnvme_async=str : [xnvme]
2858 Select the xnvme async command interface. This can take these values.
2861 This is default and use to emulate asynchronous I/O by using a
2862 single thread to create a queue pair on top of a synchronous
2863 I/O interface using the NVMe driver IOCTL.
2865 Emulate an asynchronous I/O interface with a pool of userspace
2866 threads on top of a synchronous I/O interface using the NVMe
2867 driver IOCTL. By default four threads are used.
2869 Linux native asynchronous I/O interface which supports both
2870 direct and buffered I/O.
2872 Fast Linux native asynchronous I/O interface for NVMe pass
2873 through commands. This only works with NVMe character device
2876 Use Linux aio for Asynchronous I/O.
2878 Use the posix asynchronous I/O interface to perform one or
2879 more I/O operations asynchronously.
2881 Use the user-space VFIO-based backend, implemented using
2882 libvfn instead of SPDK.
2884 Do not transfer any data; just pretend to. This is mainly used
2885 for introspective performance evaluation.
2887 .. option:: xnvme_sync=str : [xnvme]
2889 Select the xnvme synchronous command interface. This can take these values.
2892 This is default and uses Linux NVMe Driver ioctl() for
2895 This supports regular as well as vectored pread() and pwrite()
2898 This is the same as psync except that it also supports zone
2899 management commands using Linux block layer IOCTLs.
2901 .. option:: xnvme_admin=str : [xnvme]
2903 Select the xnvme admin command interface. This can take these values.
2906 This is default and uses linux NVMe Driver ioctl() for admin
2909 Use Linux Block Layer ioctl() and sysfs for admin commands.
2911 .. option:: xnvme_dev_nsid=int : [xnvme]
2913 xnvme namespace identifier for userspace NVMe driver, SPDK or vfio.
2915 .. option:: xnvme_dev_subnqn=str : [xnvme]
2917 Sets the subsystem NQN for fabrics. This is for xNVMe to utilize a
2918 fabrics target with multiple systems.
2920 .. option:: xnvme_mem=str : [xnvme]
2922 Select the xnvme memory backend. This can take these values.
2925 This is the default posix memory backend for linux NVMe driver.
2927 Use hugepages, instead of existing posix memory backend. The
2928 memory backend uses hugetlbfs. This require users to allocate
2929 hugepages, mount hugetlbfs and set an enviornment variable for
2932 Uses SPDK's memory allocator.
2934 Uses libvfn's memory allocator. This also specifies the use
2935 of libvfn backend instead of SPDK.
2937 .. option:: xnvme_iovec=int : [xnvme]
2939 If this option is set. xnvme will use vectored read/write commands.
2941 .. option:: libblkio_driver=str : [libblkio]
2943 The libblkio *driver* to use. Different drivers access devices through
2944 different underlying interfaces. Available drivers depend on the
2945 libblkio version in use and are listed at
2946 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
2948 .. option:: libblkio_path=str : [libblkio]
2950 Sets the value of the driver-specific "path" property before connecting
2951 the libblkio instance, which identifies the target device or file on
2952 which to perform I/O. Its exact semantics are driver-dependent and not
2953 all drivers may support it; see
2954 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
2956 .. option:: libblkio_pre_connect_props=str : [libblkio]
2958 A colon-separated list of additional libblkio properties to be set after
2959 creating but before connecting the libblkio instance. Each property must
2960 have the format ``<name>=<value>``. Colons can be escaped as ``\:``.
2961 These are set after the engine sets any other properties, so those can
2962 be overriden. Available properties depend on the libblkio version in use
2964 https://libblkio.gitlab.io/libblkio/blkio.html#properties
2966 .. option:: libblkio_num_entries=int : [libblkio]
2968 Sets the value of the driver-specific "num-entries" property before
2969 starting the libblkio instance. Its exact semantics are driver-dependent
2970 and not all drivers may support it; see
2971 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
2973 .. option:: libblkio_queue_size=int : [libblkio]
2975 Sets the value of the driver-specific "queue-size" property before
2976 starting the libblkio instance. Its exact semantics are driver-dependent
2977 and not all drivers may support it; see
2978 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
2980 .. option:: libblkio_pre_start_props=str : [libblkio]
2982 A colon-separated list of additional libblkio properties to be set after
2983 connecting but before starting the libblkio instance. Each property must
2984 have the format ``<name>=<value>``. Colons can be escaped as ``\:``.
2985 These are set after the engine sets any other properties, so those can
2986 be overriden. Available properties depend on the libblkio version in use
2988 https://libblkio.gitlab.io/libblkio/blkio.html#properties
2990 .. option:: libblkio_vectored : [libblkio]
2992 Submit vectored read and write requests.
2994 .. option:: libblkio_write_zeroes_on_trim : [libblkio]
2996 Submit trims as "write zeroes" requests instead of discard requests.
2998 .. option:: libblkio_wait_mode=str : [libblkio]
3000 How to wait for completions:
3003 Use a blocking call to ``blkioq_do_io()``.
3005 Use a blocking call to ``read()`` on the completion eventfd.
3007 Use a busy loop with a non-blocking call to ``blkioq_do_io()``.
3009 .. option:: libblkio_force_enable_completion_eventfd : [libblkio]
3011 Enable the queue's completion eventfd even when unused. This may impact
3012 performance. The default is to enable it only if
3013 :option:`libblkio_wait_mode=eventfd <libblkio_wait_mode>`.
3018 .. option:: iodepth=int
3020 Number of I/O units to keep in flight against the file. Note that
3021 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
3022 for small degrees when :option:`verify_async` is in use). Even async
3023 engines may impose OS restrictions causing the desired depth not to be
3024 achieved. This may happen on Linux when using libaio and not setting
3025 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
3026 eye on the I/O depth distribution in the fio output to verify that the
3027 achieved depth is as expected. Default: 1.
3029 .. option:: iodepth_batch_submit=int, iodepth_batch=int
3031 This defines how many pieces of I/O to submit at once. It defaults to 1
3032 which means that we submit each I/O as soon as it is available, but can be
3033 raised to submit bigger batches of I/O at the time. If it is set to 0 the
3034 :option:`iodepth` value will be used.
3036 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
3038 This defines how many pieces of I/O to retrieve at once. It defaults to 1
3039 which means that we'll ask for a minimum of 1 I/O in the retrieval process
3040 from the kernel. The I/O retrieval will go on until we hit the limit set by
3041 :option:`iodepth_low`. If this variable is set to 0, then fio will always
3042 check for completed events before queuing more I/O. This helps reduce I/O
3043 latency, at the cost of more retrieval system calls.
3045 .. option:: iodepth_batch_complete_max=int
3047 This defines maximum pieces of I/O to retrieve at once. This variable should
3048 be used along with :option:`iodepth_batch_complete_min`\=int variable,
3049 specifying the range of min and max amount of I/O which should be
3050 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
3055 iodepth_batch_complete_min=1
3056 iodepth_batch_complete_max=<iodepth>
3058 which means that we will retrieve at least 1 I/O and up to the whole
3059 submitted queue depth. If none of I/O has been completed yet, we will wait.
3063 iodepth_batch_complete_min=0
3064 iodepth_batch_complete_max=<iodepth>
3066 which means that we can retrieve up to the whole submitted queue depth, but
3067 if none of I/O has been completed yet, we will NOT wait and immediately exit
3068 the system call. In this example we simply do polling.
3070 .. option:: iodepth_low=int
3072 The low water mark indicating when to start filling the queue
3073 again. Defaults to the same as :option:`iodepth`, meaning that fio will
3074 attempt to keep the queue full at all times. If :option:`iodepth` is set to
3075 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
3076 16 requests, it will let the depth drain down to 4 before starting to fill
3079 .. option:: serialize_overlap=bool
3081 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
3082 When two or more I/Os are submitted simultaneously, there is no guarantee that
3083 the I/Os will be processed or completed in the submitted order. Further, if
3084 two or more of those I/Os are writes, any overlapping region between them can
3085 become indeterminate/undefined on certain storage. These issues can cause
3086 verification to fail erratically when at least one of the racing I/Os is
3087 changing data and the overlapping region has a non-zero size. Setting
3088 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
3089 serializing in-flight I/Os that have a non-zero overlap. Note that setting
3090 this option can reduce both performance and the :option:`iodepth` achieved.
3092 This option only applies to I/Os issued for a single job except when it is
3093 enabled along with :option:`io_submit_mode`\=offload. In offload mode, fio
3094 will check for overlap among all I/Os submitted by offload jobs with :option:`serialize_overlap`
3099 .. option:: io_submit_mode=str
3101 This option controls how fio submits the I/O to the I/O engine. The default
3102 is `inline`, which means that the fio job threads submit and reap I/O
3103 directly. If set to `offload`, the job threads will offload I/O submission
3104 to a dedicated pool of I/O threads. This requires some coordination and thus
3105 has a bit of extra overhead, especially for lower queue depth I/O where it
3106 can increase latencies. The benefit is that fio can manage submission rates
3107 independently of the device completion rates. This avoids skewed latency
3108 reporting if I/O gets backed up on the device side (the coordinated omission
3109 problem). Note that this option cannot reliably be used with async IO
3116 .. option:: thinktime=time
3118 Stall the job for the specified period of time after an I/O has completed before issuing the
3119 next. May be used to simulate processing being done by an application.
3120 When the unit is omitted, the value is interpreted in microseconds. See
3121 :option:`thinktime_blocks`, :option:`thinktime_iotime` and :option:`thinktime_spin`.
3123 .. option:: thinktime_spin=time
3125 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
3126 something with the data received, before falling back to sleeping for the
3127 rest of the period specified by :option:`thinktime`. When the unit is
3128 omitted, the value is interpreted in microseconds.
3130 .. option:: thinktime_blocks=int
3132 Only valid if :option:`thinktime` is set - control how many blocks to issue,
3133 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
3134 fio wait :option:`thinktime` usecs after every block. This effectively makes any
3135 queue depth setting redundant, since no more than 1 I/O will be queued
3136 before we have to complete it and do our :option:`thinktime`. In other words, this
3137 setting effectively caps the queue depth if the latter is larger.
3139 .. option:: thinktime_blocks_type=str
3141 Only valid if :option:`thinktime` is set - control how :option:`thinktime_blocks`
3142 triggers. The default is `complete`, which triggers thinktime when fio completes
3143 :option:`thinktime_blocks` blocks. If this is set to `issue`, then the trigger happens
3146 .. option:: thinktime_iotime=time
3148 Only valid if :option:`thinktime` is set - control :option:`thinktime`
3149 interval by time. The :option:`thinktime` stall is repeated after IOs
3150 are executed for :option:`thinktime_iotime`. For example,
3151 ``--thinktime_iotime=9s --thinktime=1s`` repeat 10-second cycle with IOs
3152 for 9 seconds and stall for 1 second. When the unit is omitted,
3153 :option:`thinktime_iotime` is interpreted as a number of seconds. If
3154 this option is used together with :option:`thinktime_blocks`, the
3155 :option:`thinktime` stall is repeated after :option:`thinktime_iotime`
3156 or after :option:`thinktime_blocks` IOs, whichever happens first.
3158 .. option:: rate=int[,int][,int]
3160 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
3161 suffix rules apply. Comma-separated values may be specified for reads,
3162 writes, and trims as described in :option:`blocksize`.
3164 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
3165 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
3166 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
3167 latter will only limit reads.
3169 .. option:: rate_min=int[,int][,int]
3171 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
3172 to meet this requirement will cause the job to exit. Comma-separated values
3173 may be specified for reads, writes, and trims as described in
3174 :option:`blocksize`.
3176 .. option:: rate_iops=int[,int][,int]
3178 Cap the bandwidth to this number of IOPS. Basically the same as
3179 :option:`rate`, just specified independently of bandwidth. If the job is
3180 given a block size range instead of a fixed value, the smallest block size
3181 is used as the metric. Comma-separated values may be specified for reads,
3182 writes, and trims as described in :option:`blocksize`.
3184 .. option:: rate_iops_min=int[,int][,int]
3186 If fio doesn't meet this rate of I/O, it will cause the job to exit.
3187 Comma-separated values may be specified for reads, writes, and trims as
3188 described in :option:`blocksize`.
3190 .. option:: rate_process=str
3192 This option controls how fio manages rated I/O submissions. The default is
3193 `linear`, which submits I/O in a linear fashion with fixed delays between
3194 I/Os that gets adjusted based on I/O completion rates. If this is set to
3195 `poisson`, fio will submit I/O based on a more real world random request
3196 flow, known as the Poisson process
3197 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
3198 10^6 / IOPS for the given workload.
3200 .. option:: rate_ignore_thinktime=bool
3202 By default, fio will attempt to catch up to the specified rate setting,
3203 if any kind of thinktime setting was used. If this option is set, then
3204 fio will ignore the thinktime and continue doing IO at the specified
3205 rate, instead of entering a catch-up mode after thinktime is done.
3211 .. option:: latency_target=time
3213 If set, fio will attempt to find the max performance point that the given
3214 workload will run at while maintaining a latency below this target. When
3215 the unit is omitted, the value is interpreted in microseconds. See
3216 :option:`latency_window` and :option:`latency_percentile`.
3218 .. option:: latency_window=time
3220 Used with :option:`latency_target` to specify the sample window that the job
3221 is run at varying queue depths to test the performance. When the unit is
3222 omitted, the value is interpreted in microseconds.
3224 .. option:: latency_percentile=float
3226 The percentage of I/Os that must fall within the criteria specified by
3227 :option:`latency_target` and :option:`latency_window`. If not set, this
3228 defaults to 100.0, meaning that all I/Os must be equal or below to the value
3229 set by :option:`latency_target`.
3231 .. option:: latency_run=bool
3233 Used with :option:`latency_target`. If false (default), fio will find
3234 the highest queue depth that meets :option:`latency_target` and exit. If
3235 true, fio will continue running and try to meet :option:`latency_target`
3236 by adjusting queue depth.
3238 .. option:: max_latency=time[,time][,time]
3240 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
3241 maximum latency. When the unit is omitted, the value is interpreted in
3242 microseconds. Comma-separated values may be specified for reads, writes,
3243 and trims as described in :option:`blocksize`.
3245 .. option:: rate_cycle=int
3247 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
3248 of milliseconds. Defaults to 1000.
3254 .. option:: write_iolog=str
3256 Write the issued I/O patterns to the specified file. See
3257 :option:`read_iolog`. Specify a separate file for each job, otherwise the
3258 iologs will be interspersed and the file may be corrupt. This file will
3259 be opened in append mode.
3261 .. option:: read_iolog=str
3263 Open an iolog with the specified filename and replay the I/O patterns it
3264 contains. This can be used to store a workload and replay it sometime
3265 later. The iolog given may also be a blktrace binary file, which allows fio
3266 to replay a workload captured by :command:`blktrace`. See
3267 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
3268 replay, the file needs to be turned into a blkparse binary data file first
3269 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
3270 You can specify a number of files by separating the names with a ':'
3271 character. See the :option:`filename` option for information on how to
3272 escape ':' characters within the file names. These files will
3273 be sequentially assigned to job clones created by :option:`numjobs`.
3274 '-' is a reserved name, meaning read from stdin, notably if
3275 :option:`filename` is set to '-' which means stdin as well, then
3276 this flag can't be set to '-'.
3278 .. option:: read_iolog_chunked=bool
3280 Determines how iolog is read. If false(default) entire :option:`read_iolog`
3281 will be read at once. If selected true, input from iolog will be read
3282 gradually. Useful when iolog is very large, or it is generated.
3284 .. option:: merge_blktrace_file=str
3286 When specified, rather than replaying the logs passed to :option:`read_iolog`,
3287 the logs go through a merge phase which aggregates them into a single
3288 blktrace. The resulting file is then passed on as the :option:`read_iolog`
3289 parameter. The intention here is to make the order of events consistent.
3290 This limits the influence of the scheduler compared to replaying multiple
3291 blktraces via concurrent jobs.
3293 .. option:: merge_blktrace_scalars=float_list
3295 This is a percentage based option that is index paired with the list of
3296 files passed to :option:`read_iolog`. When merging is performed, scale
3297 the time of each event by the corresponding amount. For example,
3298 ``--merge_blktrace_scalars="50:100"`` runs the first trace in halftime
3299 and the second trace in realtime. This knob is separately tunable from
3300 :option:`replay_time_scale` which scales the trace during runtime and
3301 does not change the output of the merge unlike this option.
3303 .. option:: merge_blktrace_iters=float_list
3305 This is a whole number option that is index paired with the list of files
3306 passed to :option:`read_iolog`. When merging is performed, run each trace
3307 for the specified number of iterations. For example,
3308 ``--merge_blktrace_iters="2:1"`` runs the first trace for two iterations
3309 and the second trace for one iteration.
3311 .. option:: replay_no_stall=bool
3313 When replaying I/O with :option:`read_iolog` the default behavior is to
3314 attempt to respect the timestamps within the log and replay them with the
3315 appropriate delay between IOPS. By setting this variable fio will not
3316 respect the timestamps and attempt to replay them as fast as possible while
3317 still respecting ordering. The result is the same I/O pattern to a given
3318 device, but different timings.
3320 .. option:: replay_time_scale=int
3322 When replaying I/O with :option:`read_iolog`, fio will honor the
3323 original timing in the trace. With this option, it's possible to scale
3324 the time. It's a percentage option, if set to 50 it means run at 50%
3325 the original IO rate in the trace. If set to 200, run at twice the
3326 original IO rate. Defaults to 100.
3328 .. option:: replay_redirect=str
3330 While replaying I/O patterns using :option:`read_iolog` the default behavior
3331 is to replay the IOPS onto the major/minor device that each IOP was recorded
3332 from. This is sometimes undesirable because on a different machine those
3333 major/minor numbers can map to a different device. Changing hardware on the
3334 same system can also result in a different major/minor mapping.
3335 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
3336 device regardless of the device it was recorded
3337 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
3338 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
3339 multiple devices will be replayed onto a single device, if the trace
3340 contains multiple devices. If you want multiple devices to be replayed
3341 concurrently to multiple redirected devices you must blkparse your trace
3342 into separate traces and replay them with independent fio invocations.
3343 Unfortunately this also breaks the strict time ordering between multiple
3346 .. option:: replay_align=int
3348 Force alignment of the byte offsets in a trace to this value. The value
3349 must be a power of 2.
3351 .. option:: replay_scale=int
3353 Scale byte offsets down by this factor when replaying traces. Should most
3354 likely use :option:`replay_align` as well.
3356 .. option:: replay_skip=str
3358 Sometimes it's useful to skip certain IO types in a replay trace.
3359 This could be, for instance, eliminating the writes in the trace.
3360 Or not replaying the trims/discards, if you are redirecting to
3361 a device that doesn't support them. This option takes a comma
3362 separated list of read, write, trim, sync.
3365 Threads, processes and job synchronization
3366 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3370 Fio defaults to creating jobs by using fork, however if this option is
3371 given, fio will create jobs by using POSIX Threads' function
3372 :manpage:`pthread_create(3)` to create threads instead.
3374 .. option:: wait_for=str
3376 If set, the current job won't be started until all workers of the specified
3377 waitee job are done.
3379 ``wait_for`` operates on the job name basis, so there are a few
3380 limitations. First, the waitee must be defined prior to the waiter job
3381 (meaning no forward references). Second, if a job is being referenced as a
3382 waitee, it must have a unique name (no duplicate waitees).
3384 .. option:: nice=int
3386 Run the job with the given nice value. See man :manpage:`nice(2)`.
3388 On Windows, values less than -15 set the process class to "High"; -1 through
3389 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
3392 .. option:: prio=int
3394 Set the I/O priority value of this job. Linux limits us to a positive value
3395 between 0 and 7, with 0 being the highest. See man
3396 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
3397 systems since meaning of priority may differ. For per-command priority
3398 setting, see I/O engine specific :option:`cmdprio_percentage` and
3399 :option:`cmdprio` options.
3401 .. option:: prioclass=int
3403 Set the I/O priority class. See man :manpage:`ionice(1)`. For per-command
3404 priority setting, see I/O engine specific :option:`cmdprio_percentage`
3405 and :option:`cmdprio_class` options.
3407 .. option:: cpus_allowed=str
3409 Controls the same options as :option:`cpumask`, but accepts a textual
3410 specification of the permitted CPUs instead and CPUs are indexed from 0. So
3411 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
3412 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
3413 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
3415 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
3416 processor group will be used and affinity settings are inherited from the
3417 system. An fio build configured to target Windows 7 makes options that set
3418 CPUs processor group aware and values will set both the processor group
3419 and a CPU from within that group. For example, on a system where processor
3420 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
3421 values between 0 and 39 will bind CPUs from processor group 0 and
3422 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
3423 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
3424 single ``cpus_allowed`` option must be from the same processor group. For
3425 Windows fio builds not built for Windows 7, CPUs will only be selected from
3426 (and be relative to) whatever processor group fio happens to be running in
3427 and CPUs from other processor groups cannot be used.
3429 .. option:: cpus_allowed_policy=str
3431 Set the policy of how fio distributes the CPUs specified by
3432 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
3435 All jobs will share the CPU set specified.
3437 Each job will get a unique CPU from the CPU set.
3439 **shared** is the default behavior, if the option isn't specified. If
3440 **split** is specified, then fio will assign one cpu per job. If not
3441 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
3444 .. option:: cpumask=int
3446 Set the CPU affinity of this job. The parameter given is a bit mask of
3447 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
3448 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
3449 :manpage:`sched_setaffinity(2)`. This may not work on all supported
3450 operating systems or kernel versions. This option doesn't work well for a
3451 higher CPU count than what you can store in an integer mask, so it can only
3452 control cpus 1-32. For boxes with larger CPU counts, use
3453 :option:`cpus_allowed`.
3455 .. option:: numa_cpu_nodes=str
3457 Set this job running on specified NUMA nodes' CPUs. The arguments allow
3458 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
3459 NUMA options support, fio must be built on a system with libnuma-dev(el)
3462 .. option:: numa_mem_policy=str
3464 Set this job's memory policy and corresponding NUMA nodes. Format of the
3469 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
3470 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
3471 policies, no node needs to be specified. For ``prefer``, only one node is
3472 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
3473 follows: a comma delimited list of numbers, A-B ranges, or `all`.
3475 .. option:: cgroup=str
3477 Add job to this control group. If it doesn't exist, it will be created. The
3478 system must have a mounted cgroup blkio mount point for this to work. If
3479 your system doesn't have it mounted, you can do so with::
3481 # mount -t cgroup -o blkio none /cgroup
3483 .. option:: cgroup_weight=int
3485 Set the weight of the cgroup to this value. See the documentation that comes
3486 with the kernel, allowed values are in the range of 100..1000.
3488 .. option:: cgroup_nodelete=bool
3490 Normally fio will delete the cgroups it has created after the job
3491 completion. To override this behavior and to leave cgroups around after the
3492 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
3493 to inspect various cgroup files after job completion. Default: false.
3495 .. option:: flow_id=int
3497 The ID of the flow. If not specified, it defaults to being a global
3498 flow. See :option:`flow`.
3500 .. option:: flow=int
3502 Weight in token-based flow control. If this value is used, then fio
3503 regulates the activity between two or more jobs sharing the same
3504 flow_id. Fio attempts to keep each job activity proportional to other
3505 jobs' activities in the same flow_id group, with respect to requested
3506 weight per job. That is, if one job has `flow=3', another job has
3507 `flow=2' and another with `flow=1`, then there will be a roughly 3:2:1
3508 ratio in how much one runs vs the others.
3510 .. option:: flow_sleep=int
3512 The period of time, in microseconds, to wait after the flow counter
3513 has exceeded its proportion before retrying operations.
3515 .. option:: stonewall, wait_for_previous
3517 Wait for preceding jobs in the job file to exit, before starting this
3518 one. Can be used to insert serialization points in the job file. A stone
3519 wall also implies starting a new reporting group, see
3520 :option:`group_reporting`.
3524 By default, fio will continue running all other jobs when one job finishes.
3525 Sometimes this is not the desired action. Setting ``exitall`` will instead
3526 make fio terminate all jobs in the same group, as soon as one job of that
3529 .. option:: exit_what=str
3531 By default, fio will continue running all other jobs when one job finishes.
3532 Sometimes this is not the desired action. Setting ``exitall`` will
3533 instead make fio terminate all jobs in the same group. The option
3534 ``exit_what`` allows to control which jobs get terminated when ``exitall`` is
3535 enabled. The default is ``group`` and does not change the behaviour of
3536 ``exitall``. The setting ``all`` terminates all jobs. The setting ``stonewall``
3537 terminates all currently running jobs across all groups and continues execution
3538 with the next stonewalled group.
3540 .. option:: exec_prerun=str
3542 Before running this job, issue the command specified through
3543 :manpage:`system(3)`. Output is redirected in a file called
3544 :file:`jobname.prerun.txt`.
3546 .. option:: exec_postrun=str
3548 After the job completes, issue the command specified though
3549 :manpage:`system(3)`. Output is redirected in a file called
3550 :file:`jobname.postrun.txt`.
3554 Instead of running as the invoking user, set the user ID to this value
3555 before the thread/process does any work.
3559 Set group ID, see :option:`uid`.
3565 .. option:: verify_only
3567 Do not perform specified workload, only verify data still matches previous
3568 invocation of this workload. This option allows one to check data multiple
3569 times at a later date without overwriting it. This option makes sense only
3570 for workloads that write data, and does not support workloads with the
3571 :option:`time_based` option set.
3573 .. option:: do_verify=bool
3575 Run the verify phase after a write phase. Only valid if :option:`verify` is
3578 .. option:: verify=str
3580 If writing to a file, fio can verify the file contents after each iteration
3581 of the job. Each verification method also implies verification of special
3582 header, which is written to the beginning of each block. This header also
3583 includes meta information, like offset of the block, block number, timestamp
3584 when block was written, etc. :option:`verify` can be combined with
3585 :option:`verify_pattern` option. The allowed values are:
3588 Use an md5 sum of the data area and store it in the header of
3592 Use an experimental crc64 sum of the data area and store it in the
3593 header of each block.
3596 Use a crc32c sum of the data area and store it in the header of
3597 each block. This will automatically use hardware acceleration
3598 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
3599 fall back to software crc32c if none is found. Generally the
3600 fastest checksum fio supports when hardware accelerated.
3606 Use a crc32 sum of the data area and store it in the header of each
3610 Use a crc16 sum of the data area and store it in the header of each
3614 Use a crc7 sum of the data area and store it in the header of each
3618 Use xxhash as the checksum function. Generally the fastest software
3619 checksum that fio supports.
3622 Use sha512 as the checksum function.
3625 Use sha256 as the checksum function.
3628 Use optimized sha1 as the checksum function.
3631 Use optimized sha3-224 as the checksum function.
3634 Use optimized sha3-256 as the checksum function.
3637 Use optimized sha3-384 as the checksum function.
3640 Use optimized sha3-512 as the checksum function.
3643 This option is deprecated, since now meta information is included in
3644 generic verification header and meta verification happens by
3645 default. For detailed information see the description of the
3646 :option:`verify` setting. This option is kept because of
3647 compatibility's sake with old configurations. Do not use it.
3650 Verify a strict pattern. Normally fio includes a header with some
3651 basic information and checksumming, but if this option is set, only
3652 the specific pattern set with :option:`verify_pattern` is verified.
3655 Only pretend to verify. Useful for testing internals with
3656 :option:`ioengine`\=null, not for much else.
3658 This option can be used for repeated burn-in tests of a system to make sure
3659 that the written data is also correctly read back. If the data direction
3660 given is a read or random read, fio will assume that it should verify a
3661 previously written file. If the data direction includes any form of write,
3662 the verify will be of the newly written data.
3664 To avoid false verification errors, do not use the norandommap option when
3665 verifying data with async I/O engines and I/O depths > 1. Or use the
3666 norandommap and the lfsr random generator together to avoid writing to the
3667 same offset with multiple outstanding I/Os.
3669 .. option:: verify_offset=int
3671 Swap the verification header with data somewhere else in the block before
3672 writing. It is swapped back before verifying.
3674 .. option:: verify_interval=int
3676 Write the verification header at a finer granularity than the
3677 :option:`blocksize`. It will be written for chunks the size of
3678 ``verify_interval``. :option:`blocksize` should divide this evenly.
3680 .. option:: verify_pattern=str
3682 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
3683 filling with totally random bytes, but sometimes it's interesting to fill
3684 with a known pattern for I/O verification purposes. Depending on the width
3685 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
3686 be either a decimal or a hex number). The ``verify_pattern`` if larger than
3687 a 32-bit quantity has to be a hex number that starts with either "0x" or
3688 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
3689 format, which means that for each block offset will be written and then
3690 verified back, e.g.::
3694 Or use combination of everything::
3696 verify_pattern=0xff%o"abcd"-12
3698 .. option:: verify_fatal=bool
3700 Normally fio will keep checking the entire contents before quitting on a
3701 block verification failure. If this option is set, fio will exit the job on
3702 the first observed failure. Default: false.
3704 .. option:: verify_dump=bool
3706 If set, dump the contents of both the original data block and the data block
3707 we read off disk to files. This allows later analysis to inspect just what
3708 kind of data corruption occurred. Off by default.
3710 .. option:: verify_async=int
3712 Fio will normally verify I/O inline from the submitting thread. This option
3713 takes an integer describing how many async offload threads to create for I/O
3714 verification instead, causing fio to offload the duty of verifying I/O
3715 contents to one or more separate threads. If using this offload option, even
3716 sync I/O engines can benefit from using an :option:`iodepth` setting higher
3717 than 1, as it allows them to have I/O in flight while verifies are running.
3718 Defaults to 0 async threads, i.e. verification is not asynchronous.
3720 .. option:: verify_async_cpus=str
3722 Tell fio to set the given CPU affinity on the async I/O verification
3723 threads. See :option:`cpus_allowed` for the format used.
3725 .. option:: verify_backlog=int
3727 Fio will normally verify the written contents of a job that utilizes verify
3728 once that job has completed. In other words, everything is written then
3729 everything is read back and verified. You may want to verify continually
3730 instead for a variety of reasons. Fio stores the meta data associated with
3731 an I/O block in memory, so for large verify workloads, quite a bit of memory
3732 would be used up holding this meta data. If this option is enabled, fio will
3733 write only N blocks before verifying these blocks.
3735 .. option:: verify_backlog_batch=int
3737 Control how many blocks fio will verify if :option:`verify_backlog` is
3738 set. If not set, will default to the value of :option:`verify_backlog`
3739 (meaning the entire queue is read back and verified). If
3740 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
3741 blocks will be verified, if ``verify_backlog_batch`` is larger than
3742 :option:`verify_backlog`, some blocks will be verified more than once.
3744 .. option:: verify_state_save=bool
3746 When a job exits during the write phase of a verify workload, save its
3747 current state. This allows fio to replay up until that point, if the verify
3748 state is loaded for the verify read phase. The format of the filename is,
3751 <type>-<jobname>-<jobindex>-verify.state.
3753 <type> is "local" for a local run, "sock" for a client/server socket
3754 connection, and "ip" (192.168.0.1, for instance) for a networked
3755 client/server connection. Defaults to true.
3757 .. option:: verify_state_load=bool
3759 If a verify termination trigger was used, fio stores the current write state
3760 of each thread. This can be used at verification time so that fio knows how
3761 far it should verify. Without this information, fio will run a full
3762 verification pass, according to the settings in the job file used. Default
3765 .. option:: trim_percentage=int
3767 Number of verify blocks to discard/trim.
3769 .. option:: trim_verify_zero=bool
3771 Verify that trim/discarded blocks are returned as zeros.
3773 .. option:: trim_backlog=int
3775 Trim after this number of blocks are written.
3777 .. option:: trim_backlog_batch=int
3779 Trim this number of I/O blocks.
3781 .. option:: experimental_verify=bool
3783 Enable experimental verification. Standard verify records I/O metadata
3784 for later use during the verification phase. Experimental verify
3785 instead resets the file after the write phase and then replays I/Os for
3786 the verification phase.
3791 .. option:: steadystate=str:float, ss=str:float
3793 Define the criterion and limit for assessing steady state performance. The
3794 first parameter designates the criterion whereas the second parameter sets
3795 the threshold. When the criterion falls below the threshold for the
3796 specified duration, the job will stop. For example, `iops_slope:0.1%` will
3797 direct fio to terminate the job when the least squares regression slope
3798 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
3799 this will apply to all jobs in the group. Below is the list of available
3800 steady state assessment criteria. All assessments are carried out using only
3801 data from the rolling collection window. Threshold limits can be expressed
3802 as a fixed value or as a percentage of the mean in the collection window.
3804 When using this feature, most jobs should include the :option:`time_based`
3805 and :option:`runtime` options or the :option:`loops` option so that fio does not
3806 stop running after it has covered the full size of the specified file(s) or device(s).
3809 Collect IOPS data. Stop the job if all individual IOPS measurements
3810 are within the specified limit of the mean IOPS (e.g., ``iops:2``
3811 means that all individual IOPS values must be within 2 of the mean,
3812 whereas ``iops:0.2%`` means that all individual IOPS values must be
3813 within 0.2% of the mean IOPS to terminate the job).
3816 Collect IOPS data and calculate the least squares regression
3817 slope. Stop the job if the slope falls below the specified limit.
3820 Collect bandwidth data. Stop the job if all individual bandwidth
3821 measurements are within the specified limit of the mean bandwidth.
3824 Collect bandwidth data and calculate the least squares regression
3825 slope. Stop the job if the slope falls below the specified limit.
3827 .. option:: steadystate_duration=time, ss_dur=time
3829 A rolling window of this duration will be used to judge whether steady
3830 state has been reached. Data will be collected every
3831 :option:`ss_interval`. The default is 0 which disables steady state
3832 detection. When the unit is omitted, the value is interpreted in
3835 .. option:: steadystate_ramp_time=time, ss_ramp=time
3837 Allow the job to run for the specified duration before beginning data
3838 collection for checking the steady state job termination criterion. The
3839 default is 0. When the unit is omitted, the value is interpreted in seconds.
3841 .. option:: steadystate_check_interval=time, ss_interval=time
3843 The values during the rolling window will be collected with a period of
3844 this value. If :option:`ss_interval` is 30s and :option:`ss_dur` is
3845 300s, 10 measurements will be taken. Default is 1s but that might not
3846 converge, especially for slower devices, so set this accordingly. When
3847 the unit is omitted, the value is interpreted in seconds.
3850 Measurements and reporting
3851 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3853 .. option:: per_job_logs=bool
3855 If set, this generates bw/clat/iops log with per file private filenames. If
3856 not set, jobs with identical names will share the log filename. Default:
3859 .. option:: group_reporting
3861 It may sometimes be interesting to display statistics for groups of jobs as
3862 a whole instead of for each individual job. This is especially true if
3863 :option:`numjobs` is used; looking at individual thread/process output
3864 quickly becomes unwieldy. To see the final report per-group instead of
3865 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
3866 same reporting group, unless if separated by a :option:`stonewall`, or by
3867 using :option:`new_group`.
3869 .. option:: new_group
3871 Start a new reporting group. See: :option:`group_reporting`. If not given,
3872 all jobs in a file will be part of the same reporting group, unless
3873 separated by a :option:`stonewall`.
3875 .. option:: stats=bool
3877 By default, fio collects and shows final output results for all jobs
3878 that run. If this option is set to 0, then fio will ignore it in
3879 the final stat output.
3881 .. option:: write_bw_log=str
3883 If given, write a bandwidth log for this job. Can be used to store data of
3884 the bandwidth of the jobs in their lifetime.
3886 If no str argument is given, the default filename of
3887 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
3888 will still append the type of log. So if one specifies::
3892 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
3893 of the job (`1..N`, where `N` is the number of jobs). If
3894 :option:`per_job_logs` is false, then the filename will not include the
3897 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
3898 text files into nice graphs. See `Log File Formats`_ for how data is
3899 structured within the file.
3901 .. option:: write_lat_log=str
3903 Same as :option:`write_bw_log`, except this option creates I/O
3904 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
3905 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
3906 latency files instead. See :option:`write_bw_log` for details about
3907 the filename format and `Log File Formats`_ for how data is structured
3910 .. option:: write_hist_log=str
3912 Same as :option:`write_bw_log` but writes an I/O completion latency
3913 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
3914 file will be empty unless :option:`log_hist_msec` has also been set.
3915 See :option:`write_bw_log` for details about the filename format and
3916 `Log File Formats`_ for how data is structured within the file.
3918 .. option:: write_iops_log=str
3920 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
3921 :file:`name_iops.x.log`) instead. Because fio defaults to individual
3922 I/O logging, the value entry in the IOPS log will be 1 unless windowed
3923 logging (see :option:`log_avg_msec`) has been enabled. See
3924 :option:`write_bw_log` for details about the filename format and `Log
3925 File Formats`_ for how data is structured within the file.
3927 .. option:: log_entries=int
3929 By default, fio will log an entry in the iops, latency, or bw log for
3930 every I/O that completes. The initial number of I/O log entries is 1024.
3931 When the log entries are all used, new log entries are dynamically
3932 allocated. This dynamic log entry allocation may negatively impact
3933 time-related statistics such as I/O tail latencies (e.g. 99.9th percentile
3934 completion latency). This option allows specifying a larger initial
3935 number of log entries to avoid run-time allocations of new log entries,
3936 resulting in more precise time-related I/O statistics.
3937 Also see :option:`log_avg_msec`. Defaults to 1024.
3939 .. option:: log_avg_msec=int
3941 By default, fio will log an entry in the iops, latency, or bw log for every
3942 I/O that completes. When writing to the disk log, that can quickly grow to a
3943 very large size. Setting this option makes fio average the each log entry
3944 over the specified period of time, reducing the resolution of the log. See
3945 :option:`log_max_value` as well. Defaults to 0, logging all entries.
3946 Also see `Log File Formats`_.
3948 .. option:: log_hist_msec=int
3950 Same as :option:`log_avg_msec`, but logs entries for completion latency
3951 histograms. Computing latency percentiles from averages of intervals using
3952 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
3953 histogram entries over the specified period of time, reducing log sizes for
3954 high IOPS devices while retaining percentile accuracy. See
3955 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
3956 Defaults to 0, meaning histogram logging is disabled.
3958 .. option:: log_hist_coarseness=int
3960 Integer ranging from 0 to 6, defining the coarseness of the resolution of
3961 the histogram logs enabled with :option:`log_hist_msec`. For each increment
3962 in coarseness, fio outputs half as many bins. Defaults to 0, for which
3963 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
3964 and `Log File Formats`_.
3966 .. option:: log_max_value=bool
3968 If :option:`log_avg_msec` is set, fio logs the average over that window. If
3969 you instead want to log the maximum value, set this option to 1. Defaults to
3970 0, meaning that averaged values are logged.
3972 .. option:: log_offset=bool
3974 If this is set, the iolog options will include the byte offset for the I/O
3975 entry as well as the other data values. Defaults to 0 meaning that
3976 offsets are not present in logs. Also see `Log File Formats`_.
3978 .. option:: log_compression=int
3980 If this is set, fio will compress the I/O logs as it goes, to keep the
3981 memory footprint lower. When a log reaches the specified size, that chunk is
3982 removed and compressed in the background. Given that I/O logs are fairly
3983 highly compressible, this yields a nice memory savings for longer runs. The
3984 downside is that the compression will consume some background CPU cycles, so
3985 it may impact the run. This, however, is also true if the logging ends up
3986 consuming most of the system memory. So pick your poison. The I/O logs are
3987 saved normally at the end of a run, by decompressing the chunks and storing
3988 them in the specified log file. This feature depends on the availability of
3991 .. option:: log_compression_cpus=str
3993 Define the set of CPUs that are allowed to handle online log compression for
3994 the I/O jobs. This can provide better isolation between performance
3995 sensitive jobs, and background compression work. See
3996 :option:`cpus_allowed` for the format used.
3998 .. option:: log_store_compressed=bool
4000 If set, fio will store the log files in a compressed format. They can be
4001 decompressed with fio, using the :option:`--inflate-log` command line
4002 parameter. The files will be stored with a :file:`.fz` suffix.
4004 .. option:: log_unix_epoch=bool
4006 If set, fio will log Unix timestamps to the log files produced by enabling
4007 write_type_log for each log type, instead of the default zero-based
4010 .. option:: log_alternate_epoch=bool
4012 If set, fio will log timestamps based on the epoch used by the clock specified
4013 in the log_alternate_epoch_clock_id option, to the log files produced by
4014 enabling write_type_log for each log type, instead of the default zero-based
4017 .. option:: log_alternate_epoch_clock_id=int
4019 Specifies the clock_id to be used by clock_gettime to obtain the alternate epoch
4020 if either log_unix_epoch or log_alternate_epoch are true. Otherwise has no
4021 effect. Default value is 0, or CLOCK_REALTIME.
4023 .. option:: block_error_percentiles=bool
4025 If set, record errors in trim block-sized units from writes and trims and
4026 output a histogram of how many trims it took to get to errors, and what kind
4027 of error was encountered.
4029 .. option:: bwavgtime=int
4031 Average the calculated bandwidth over the given time. Value is specified in
4032 milliseconds. If the job also does bandwidth logging through
4033 :option:`write_bw_log`, then the minimum of this option and
4034 :option:`log_avg_msec` will be used. Default: 500ms.
4036 .. option:: iopsavgtime=int
4038 Average the calculated IOPS over the given time. Value is specified in
4039 milliseconds. If the job also does IOPS logging through
4040 :option:`write_iops_log`, then the minimum of this option and
4041 :option:`log_avg_msec` will be used. Default: 500ms.
4043 .. option:: disk_util=bool
4045 Generate disk utilization statistics, if the platform supports it.
4048 .. option:: disable_lat=bool
4050 Disable measurements of total latency numbers. Useful only for cutting back
4051 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
4052 performance at really high IOPS rates. Note that to really get rid of a
4053 large amount of these calls, this option must be used with
4054 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
4056 .. option:: disable_clat=bool
4058 Disable measurements of completion latency numbers. See
4059 :option:`disable_lat`.
4061 .. option:: disable_slat=bool
4063 Disable measurements of submission latency numbers. See
4064 :option:`disable_lat`.
4066 .. option:: disable_bw_measurement=bool, disable_bw=bool
4068 Disable measurements of throughput/bandwidth numbers. See
4069 :option:`disable_lat`.
4071 .. option:: slat_percentiles=bool
4073 Report submission latency percentiles. Submission latency is not recorded
4074 for synchronous ioengines.
4076 .. option:: clat_percentiles=bool
4078 Report completion latency percentiles.
4080 .. option:: lat_percentiles=bool
4082 Report total latency percentiles. Total latency is the sum of submission
4083 latency and completion latency.
4085 .. option:: percentile_list=float_list
4087 Overwrite the default list of percentiles for latencies and the block error
4088 histogram. Each number is a floating point number in the range (0,100], and
4089 the maximum length of the list is 20. Use ``:`` to separate the numbers. For
4090 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
4091 latency durations below which 99.5% and 99.9% of the observed latencies fell,
4094 .. option:: significant_figures=int
4096 If using :option:`--output-format` of `normal`, set the significant
4097 figures to this value. Higher values will yield more precise IOPS and
4098 throughput units, while lower values will round. Requires a minimum
4099 value of 1 and a maximum value of 10. Defaults to 4.
4105 .. option:: exitall_on_error
4107 When one job finishes in error, terminate the rest. The default is to wait
4108 for each job to finish.
4110 .. option:: continue_on_error=str
4112 Normally fio will exit the job on the first observed failure. If this option
4113 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
4114 EILSEQ) until the runtime is exceeded or the I/O size specified is
4115 completed. If this option is used, there are two more stats that are
4116 appended, the total error count and the first error. The error field given
4117 in the stats is the first error that was hit during the run.
4119 Note: a write error from the device may go unnoticed by fio when using
4120 buffered IO, as the write() (or similar) system call merely dirties the
4121 kernel pages, unless :option:`sync` or :option:`direct` is used. Device IO
4122 errors occur when the dirty data is actually written out to disk. If fully
4123 sync writes aren't desirable, :option:`fsync` or :option:`fdatasync` can be
4124 used as well. This is specific to writes, as reads are always synchronous.
4126 The allowed values are:
4129 Exit on any I/O or verify errors.
4132 Continue on read errors, exit on all others.
4135 Continue on write errors, exit on all others.
4138 Continue on any I/O error, exit on all others.
4141 Continue on verify errors, exit on all others.
4144 Continue on all errors.
4147 Backward-compatible alias for 'none'.
4150 Backward-compatible alias for 'all'.
4152 .. option:: ignore_error=str
4154 Sometimes you want to ignore some errors during test in that case you can
4155 specify error list for each error type, instead of only being able to
4156 ignore the default 'non-fatal error' using :option:`continue_on_error`.
4157 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
4158 given error type is separated with ':'. Error may be symbol ('ENOSPC',
4159 'ENOMEM') or integer. Example::
4161 ignore_error=EAGAIN,ENOSPC:122
4163 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
4164 WRITE. This option works by overriding :option:`continue_on_error` with
4165 the list of errors for each error type if any.
4167 .. option:: error_dump=bool
4169 If set dump every error even if it is non fatal, true by default. If
4170 disabled only fatal error will be dumped.
4172 Running predefined workloads
4173 ----------------------------
4175 Fio includes predefined profiles that mimic the I/O workloads generated by
4178 .. option:: profile=str
4180 The predefined workload to run. Current profiles are:
4183 Threaded I/O bench (tiotest/tiobench) like workload.
4186 Aerospike Certification Tool (ACT) like workload.
4188 To view a profile's additional options use :option:`--cmdhelp` after specifying
4189 the profile. For example::
4191 $ fio --profile=act --cmdhelp
4196 .. option:: device-names=str
4201 .. option:: load=int
4204 ACT load multiplier. Default: 1.
4206 .. option:: test-duration=time
4209 How long the entire test takes to run. When the unit is omitted, the value
4210 is given in seconds. Default: 24h.
4212 .. option:: threads-per-queue=int
4215 Number of read I/O threads per device. Default: 8.
4217 .. option:: read-req-num-512-blocks=int
4220 Number of 512B blocks to read at the time. Default: 3.
4222 .. option:: large-block-op-kbytes=int
4225 Size of large block ops in KiB (writes). Default: 131072.
4230 Set to run ACT prep phase.
4232 Tiobench profile options
4233 ~~~~~~~~~~~~~~~~~~~~~~~~
4235 .. option:: size=str
4240 .. option:: block=int
4243 Block size in bytes. Default: 4096.
4245 .. option:: numruns=int
4255 .. option:: threads=int
4260 Interpreting the output
4261 -----------------------
4264 Example output was based on the following:
4265 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
4266 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
4267 --runtime=2m --rw=rw
4269 Fio spits out a lot of output. While running, fio will display the status of the
4270 jobs created. An example of that would be::
4272 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]
4274 The characters inside the first set of square brackets denote the current status of
4275 each thread. The first character is the first job defined in the job file, and so
4276 forth. The possible values (in typical life cycle order) are:
4278 +------+-----+-----------------------------------------------------------+
4280 +======+=====+===========================================================+
4281 | P | | Thread setup, but not started. |
4282 +------+-----+-----------------------------------------------------------+
4283 | C | | Thread created. |
4284 +------+-----+-----------------------------------------------------------+
4285 | I | | Thread initialized, waiting or generating necessary data. |
4286 +------+-----+-----------------------------------------------------------+
4287 | | p | Thread running pre-reading file(s). |
4288 +------+-----+-----------------------------------------------------------+
4289 | | / | Thread is in ramp period. |
4290 +------+-----+-----------------------------------------------------------+
4291 | | R | Running, doing sequential reads. |
4292 +------+-----+-----------------------------------------------------------+
4293 | | r | Running, doing random reads. |
4294 +------+-----+-----------------------------------------------------------+
4295 | | W | Running, doing sequential writes. |
4296 +------+-----+-----------------------------------------------------------+
4297 | | w | Running, doing random writes. |
4298 +------+-----+-----------------------------------------------------------+
4299 | | M | Running, doing mixed sequential reads/writes. |
4300 +------+-----+-----------------------------------------------------------+
4301 | | m | Running, doing mixed random reads/writes. |
4302 +------+-----+-----------------------------------------------------------+
4303 | | D | Running, doing sequential trims. |
4304 +------+-----+-----------------------------------------------------------+
4305 | | d | Running, doing random trims. |
4306 +------+-----+-----------------------------------------------------------+
4307 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
4308 +------+-----+-----------------------------------------------------------+
4309 | | V | Running, doing verification of written data. |
4310 +------+-----+-----------------------------------------------------------+
4311 | f | | Thread finishing. |
4312 +------+-----+-----------------------------------------------------------+
4313 | E | | Thread exited, not reaped by main thread yet. |
4314 +------+-----+-----------------------------------------------------------+
4315 | _ | | Thread reaped. |
4316 +------+-----+-----------------------------------------------------------+
4317 | X | | Thread reaped, exited with an error. |
4318 +------+-----+-----------------------------------------------------------+
4319 | K | | Thread reaped, exited due to signal. |
4320 +------+-----+-----------------------------------------------------------+
4323 Example output was based on the following:
4324 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
4325 --time_based --rate=2512k --bs=256K --numjobs=10 \
4326 --name=readers --rw=read --name=writers --rw=write
4328 Fio will condense the thread string as not to take up more space on the command
4329 line than needed. For instance, if you have 10 readers and 10 writers running,
4330 the output would look like this::
4332 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]
4334 Note that the status string is displayed in order, so it's possible to tell which of
4335 the jobs are currently doing what. In the example above this means that jobs 1--10
4336 are readers and 11--20 are writers.
4338 The other values are fairly self explanatory -- number of threads currently
4339 running and doing I/O, the number of currently open files (f=), the estimated
4340 completion percentage, the rate of I/O since last check (read speed listed first,
4341 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
4342 and time to completion for the current running group. It's impossible to estimate
4343 runtime of the following groups (if any).
4346 Example output was based on the following:
4347 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
4348 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
4349 --bs=7K --name=Client1 --rw=write
4351 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
4352 each thread, group of threads, and disks in that order. For each overall thread (or
4353 group) the output looks like::
4355 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
4356 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
4357 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
4358 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
4359 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
4360 clat percentiles (usec):
4361 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
4362 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
4363 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
4364 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
4366 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
4367 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
4368 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
4369 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
4370 lat (msec) : 100=0.65%
4371 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
4372 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
4373 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4374 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4375 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
4376 latency : target=0, window=0, percentile=100.00%, depth=8
4378 The job name (or first job's name when using :option:`group_reporting`) is printed,
4379 along with the group id, count of jobs being aggregated, last error id seen (which
4380 is 0 when there are no errors), pid/tid of that thread and the time the job/group
4381 completed. Below are the I/O statistics for each data direction performed (showing
4382 writes in the example above). In the order listed, they denote:
4385 The string before the colon shows the I/O direction the statistics
4386 are for. **IOPS** is the average I/Os performed per second. **BW**
4387 is the average bandwidth rate shown as: value in power of 2 format
4388 (value in power of 10 format). The last two values show: (**total
4389 I/O performed** in power of 2 format / **runtime** of that thread).
4392 Submission latency (**min** being the minimum, **max** being the
4393 maximum, **avg** being the average, **stdev** being the standard
4394 deviation). This is the time from when fio initialized the I/O
4395 to submission. For synchronous ioengines this includes the time
4396 up until just before the ioengine's queue function is called.
4397 For asynchronous ioengines this includes the time up through the
4398 completion of the ioengine's queue function (and commit function
4399 if it is defined). For sync I/O this row is not displayed as the
4400 slat is negligible. This value can be in nanoseconds,
4401 microseconds or milliseconds --- fio will choose the most
4402 appropriate base and print that (in the example above
4403 nanoseconds was the best scale). Note: in :option:`--minimal`
4404 mode latencies are always expressed in microseconds.
4407 Completion latency. Same names as slat, this denotes the time from
4408 submission to completion of the I/O pieces. For sync I/O, this
4409 represents the time from when the I/O was submitted to the
4410 operating system to when it was completed. For asynchronous
4411 ioengines this is the time from when the ioengine's queue (and
4412 commit if available) functions were completed to when the I/O's
4413 completion was reaped by fio.
4416 Total latency. Same names as slat and clat, this denotes the time from
4417 when fio created the I/O unit to completion of the I/O operation.
4418 It is the sum of submission and completion latency.
4421 Bandwidth statistics based on samples. Same names as the xlat stats,
4422 but also includes the number of samples taken (**samples**) and an
4423 approximate percentage of total aggregate bandwidth this thread
4424 received in its group (**per**). This last value is only really
4425 useful if the threads in this group are on the same disk, since they
4426 are then competing for disk access.
4429 IOPS statistics based on samples. Same names as bw.
4431 **lat (nsec/usec/msec)**
4432 The distribution of I/O completion latencies. This is the time from when
4433 I/O leaves fio and when it gets completed. Unlike the separate
4434 read/write/trim sections above, the data here and in the remaining
4435 sections apply to all I/Os for the reporting group. 250=0.04% means that
4436 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
4437 of the I/Os required 250 to 499us for completion.
4440 CPU usage. User and system time, along with the number of context
4441 switches this thread went through, usage of system and user time, and
4442 finally the number of major and minor page faults. The CPU utilization
4443 numbers are averages for the jobs in that reporting group, while the
4444 context and fault counters are summed.
4447 The distribution of I/O depths over the job lifetime. The numbers are
4448 divided into powers of 2 and each entry covers depths from that value
4449 up to those that are lower than the next entry -- e.g., 16= covers
4450 depths from 16 to 31. Note that the range covered by a depth
4451 distribution entry can be different to the range covered by the
4452 equivalent submit/complete distribution entry.
4455 How many pieces of I/O were submitting in a single submit call. Each
4456 entry denotes that amount and below, until the previous entry -- e.g.,
4457 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
4458 call. Note that the range covered by a submit distribution entry can
4459 be different to the range covered by the equivalent depth distribution
4463 Like the above submit number, but for completions instead.
4466 The number of read/write/trim requests issued, and how many of them were
4470 These values are for :option:`latency_target` and related options. When
4471 these options are engaged, this section describes the I/O depth required
4472 to meet the specified latency target.
4475 Example output was based on the following:
4476 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
4477 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
4478 --rate=11M --name=write --rw=write --bs=2k --rate=700k
4480 After each client has been listed, the group statistics are printed. They
4481 will look like this::
4483 Run status group 0 (all jobs):
4484 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
4485 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
4487 For each data direction it prints:
4490 Aggregate bandwidth of threads in this group followed by the
4491 minimum and maximum bandwidth of all the threads in this group.
4492 Values outside of brackets are power-of-2 format and those
4493 within are the equivalent value in a power-of-10 format.
4495 Aggregate I/O performed of all threads in this group. The
4496 format is the same as bw.
4498 The smallest and longest runtimes of the threads in this group.
4500 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
4502 Disk stats (read/write):
4503 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
4505 Each value is printed for both reads and writes, with reads first. The
4509 Number of I/Os performed by all groups.
4511 Number of merges performed by the I/O scheduler.
4513 Number of ticks we kept the disk busy.
4515 Total time spent in the disk queue.
4517 The disk utilization. A value of 100% means we kept the disk
4518 busy constantly, 50% would be a disk idling half of the time.
4520 It is also possible to get fio to dump the current output while it is running,
4521 without terminating the job. To do that, send fio the **USR1** signal. You can
4522 also get regularly timed dumps by using the :option:`--status-interval`
4523 parameter, or by creating a file in :file:`/tmp` named
4524 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
4525 current output status.
4531 For scripted usage where you typically want to generate tables or graphs of the
4532 results, fio can output the results in a semicolon separated format. The format
4533 is one long line of values, such as::
4535 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%
4536 A description of this job goes here.
4538 The job description (if provided) follows on a second line for terse v2.
4539 It appears on the same line for other terse versions.
4541 To enable terse output, use the :option:`--minimal` or
4542 :option:`--output-format`\=terse command line options. The
4543 first value is the version of the terse output format. If the output has to be
4544 changed for some reason, this number will be incremented by 1 to signify that
4547 Split up, the format is as follows (comments in brackets denote when a
4548 field was introduced or whether it's specific to some terse version):
4552 terse version, fio version [v3], jobname, groupid, error
4556 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4557 Submission latency: min, max, mean, stdev (usec)
4558 Completion latency: min, max, mean, stdev (usec)
4559 Completion latency percentiles: 20 fields (see below)
4560 Total latency: min, max, mean, stdev (usec)
4561 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4562 IOPS [v5]: min, max, mean, stdev, number of samples
4568 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4569 Submission latency: min, max, mean, stdev (usec)
4570 Completion latency: min, max, mean, stdev (usec)
4571 Completion latency percentiles: 20 fields (see below)
4572 Total latency: min, max, mean, stdev (usec)
4573 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4574 IOPS [v5]: min, max, mean, stdev, number of samples
4576 TRIM status [all but version 3]:
4578 Fields are similar to READ/WRITE status.
4582 user, system, context switches, major faults, minor faults
4586 <=1, 2, 4, 8, 16, 32, >=64
4588 I/O latencies microseconds::
4590 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
4592 I/O latencies milliseconds::
4594 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
4596 Disk utilization [v3]::
4598 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
4599 time spent in queue, disk utilization percentage
4601 Additional Info (dependent on continue_on_error, default off)::
4603 total # errors, first error code
4605 Additional Info (dependent on description being set)::
4609 Completion latency percentiles can be a grouping of up to 20 sets, so for the
4610 terse output fio writes all of them. Each field will look like this::
4614 which is the Xth percentile, and the `usec` latency associated with it.
4616 For `Disk utilization`, all disks used by fio are shown. So for each disk there
4617 will be a disk utilization section.
4619 Below is a single line containing short names for each of the fields in the
4620 minimal output v3, separated by semicolons::
4622 terse_version_3;fio_version;jobname;groupid;error;read_kb;read_bandwidth_kb;read_iops;read_runtime_ms;read_slat_min_us;read_slat_max_us;read_slat_mean_us;read_slat_dev_us;read_clat_min_us;read_clat_max_us;read_clat_mean_us;read_clat_dev_us;read_clat_pct01;read_clat_pct02;read_clat_pct03;read_clat_pct04;read_clat_pct05;read_clat_pct06;read_clat_pct07;read_clat_pct08;read_clat_pct09;read_clat_pct10;read_clat_pct11;read_clat_pct12;read_clat_pct13;read_clat_pct14;read_clat_pct15;read_clat_pct16;read_clat_pct17;read_clat_pct18;read_clat_pct19;read_clat_pct20;read_tlat_min_us;read_lat_max_us;read_lat_mean_us;read_lat_dev_us;read_bw_min_kb;read_bw_max_kb;read_bw_agg_pct;read_bw_mean_kb;read_bw_dev_kb;write_kb;write_bandwidth_kb;write_iops;write_runtime_ms;write_slat_min_us;write_slat_max_us;write_slat_mean_us;write_slat_dev_us;write_clat_min_us;write_clat_max_us;write_clat_mean_us;write_clat_dev_us;write_clat_pct01;write_clat_pct02;write_clat_pct03;write_clat_pct04;write_clat_pct05;write_clat_pct06;write_clat_pct07;write_clat_pct08;write_clat_pct09;write_clat_pct10;write_clat_pct11;write_clat_pct12;write_clat_pct13;write_clat_pct14;write_clat_pct15;write_clat_pct16;write_clat_pct17;write_clat_pct18;write_clat_pct19;write_clat_pct20;write_tlat_min_us;write_lat_max_us;write_lat_mean_us;write_lat_dev_us;write_bw_min_kb;write_bw_max_kb;write_bw_agg_pct;write_bw_mean_kb;write_bw_dev_kb;cpu_user;cpu_sys;cpu_csw;cpu_mjf;cpu_minf;iodepth_1;iodepth_2;iodepth_4;iodepth_8;iodepth_16;iodepth_32;iodepth_64;lat_2us;lat_4us;lat_10us;lat_20us;lat_50us;lat_100us;lat_250us;lat_500us;lat_750us;lat_1000us;lat_2ms;lat_4ms;lat_10ms;lat_20ms;lat_50ms;lat_100ms;lat_250ms;lat_500ms;lat_750ms;lat_1000ms;lat_2000ms;lat_over_2000ms;disk_name;disk_read_iops;disk_write_iops;disk_read_merges;disk_write_merges;disk_read_ticks;write_ticks;disk_queue_time;disk_util
4624 In client/server mode terse output differs from what appears when jobs are run
4625 locally. Disk utilization data is omitted from the standard terse output and
4626 for v3 and later appears on its own separate line at the end of each terse
4633 The `json` output format is intended to be both human readable and convenient
4634 for automated parsing. For the most part its sections mirror those of the
4635 `normal` output. The `runtime` value is reported in msec and the `bw` value is
4636 reported in 1024 bytes per second units.
4642 The `json+` output format is identical to the `json` output format except that it
4643 adds a full dump of the completion latency bins. Each `bins` object contains a
4644 set of (key, value) pairs where keys are latency durations and values count how
4645 many I/Os had completion latencies of the corresponding duration. For example,
4648 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
4650 This data indicates that one I/O required 87,552ns to complete, two I/Os required
4651 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
4653 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
4654 json+ output and generates CSV-formatted latency data suitable for plotting.
4656 The latency durations actually represent the midpoints of latency intervals.
4657 For details refer to :file:`stat.h`.
4663 There are two trace file format that you can encounter. The older (v1) format is
4664 unsupported since version 1.20-rc3 (March 2008). It will still be described
4665 below in case that you get an old trace and want to understand it.
4667 In any case the trace is a simple text file with a single action per line.
4670 Trace file format v1
4671 ~~~~~~~~~~~~~~~~~~~~
4673 Each line represents a single I/O action in the following format::
4677 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
4679 This format is not supported in fio versions >= 1.20-rc3.
4682 Trace file format v2
4683 ~~~~~~~~~~~~~~~~~~~~
4685 The second version of the trace file format was added in fio version 1.17. It
4686 allows one to access more than one file per trace and has a bigger set of possible
4689 The first line of the trace file has to be::
4693 Following this can be lines in two different formats, which are described below.
4695 The file management format::
4699 The `filename` is given as an absolute path. The `action` can be one of these:
4702 Add the given `filename` to the trace.
4704 Open the file with the given `filename`. The `filename` has to have
4705 been added with the **add** action before.
4707 Close the file with the given `filename`. The file has to have been
4711 The file I/O action format::
4713 filename action offset length
4715 The `filename` is given as an absolute path, and has to have been added and
4716 opened before it can be used with this format. The `offset` and `length` are
4717 given in bytes. The `action` can be one of these:
4720 Wait for `offset` microseconds. Everything below 100 is discarded.
4721 The time is relative to the previous `wait` statement. Note that
4722 action `wait` is not allowed as of version 3, as the same behavior
4723 can be achieved using timestamps.
4725 Read `length` bytes beginning from `offset`.
4727 Write `length` bytes beginning from `offset`.
4729 :manpage:`fsync(2)` the file.
4731 :manpage:`fdatasync(2)` the file.
4733 Trim the given file from the given `offset` for `length` bytes.
4736 Trace file format v3
4737 ~~~~~~~~~~~~~~~~~~~~
4739 The third version of the trace file format was added in fio version 3.31. It
4740 forces each action to have a timestamp associated with it.
4742 The first line of the trace file has to be::
4746 Following this can be lines in two different formats, which are described below.
4748 The file management format::
4750 timestamp filename action
4752 The file I/O action format::
4754 timestamp filename action offset length
4756 The `timestamp` is relative to the beginning of the run (ie starts at 0). The
4757 `filename`, `action`, `offset` and `length` are identical to version 2, except
4758 that version 3 does not allow the `wait` action.
4761 I/O Replay - Merging Traces
4762 ---------------------------
4764 Colocation is a common practice used to get the most out of a machine.
4765 Knowing which workloads play nicely with each other and which ones don't is
4766 a much harder task. While fio can replay workloads concurrently via multiple
4767 jobs, it leaves some variability up to the scheduler making results harder to
4768 reproduce. Merging is a way to make the order of events consistent.
4770 Merging is integrated into I/O replay and done when a
4771 :option:`merge_blktrace_file` is specified. The list of files passed to
4772 :option:`read_iolog` go through the merge process and output a single file
4773 stored to the specified file. The output file is passed on as if it were the
4774 only file passed to :option:`read_iolog`. An example would look like::
4776 $ fio --read_iolog="<file1>:<file2>" --merge_blktrace_file="<output_file>"
4778 Creating only the merged file can be done by passing the command line argument
4779 :option:`--merge-blktrace-only`.
4781 Scaling traces can be done to see the relative impact of any particular trace
4782 being slowed down or sped up. :option:`merge_blktrace_scalars` takes in a colon
4783 separated list of percentage scalars. It is index paired with the files passed
4784 to :option:`read_iolog`.
4786 With scaling, it may be desirable to match the running time of all traces.
4787 This can be done with :option:`merge_blktrace_iters`. It is index paired with
4788 :option:`read_iolog` just like :option:`merge_blktrace_scalars`.
4790 In an example, given two traces, A and B, each 60s long. If we want to see
4791 the impact of trace A issuing IOs twice as fast and repeat trace A over the
4792 runtime of trace B, the following can be done::
4794 $ fio --read_iolog="<trace_a>:"<trace_b>" --merge_blktrace_file"<output_file>" --merge_blktrace_scalars="50:100" --merge_blktrace_iters="2:1"
4796 This runs trace A at 2x the speed twice for approximately the same runtime as
4797 a single run of trace B.
4800 CPU idleness profiling
4801 ----------------------
4803 In some cases, we want to understand CPU overhead in a test. For example, we
4804 test patches for the specific goodness of whether they reduce CPU usage.
4805 Fio implements a balloon approach to create a thread per CPU that runs at idle
4806 priority, meaning that it only runs when nobody else needs the cpu.
4807 By measuring the amount of work completed by the thread, idleness of each CPU
4808 can be derived accordingly.
4810 An unit work is defined as touching a full page of unsigned characters. Mean and
4811 standard deviation of time to complete an unit work is reported in "unit work"
4812 section. Options can be chosen to report detailed percpu idleness or overall
4813 system idleness by aggregating percpu stats.
4816 Verification and triggers
4817 -------------------------
4819 Fio is usually run in one of two ways, when data verification is done. The first
4820 is a normal write job of some sort with verify enabled. When the write phase has
4821 completed, fio switches to reads and verifies everything it wrote. The second
4822 model is running just the write phase, and then later on running the same job
4823 (but with reads instead of writes) to repeat the same I/O patterns and verify
4824 the contents. Both of these methods depend on the write phase being completed,
4825 as fio otherwise has no idea how much data was written.
4827 With verification triggers, fio supports dumping the current write state to
4828 local files. Then a subsequent read verify workload can load this state and know
4829 exactly where to stop. This is useful for testing cases where power is cut to a
4830 server in a managed fashion, for instance.
4832 A verification trigger consists of two things:
4834 1) Storing the write state of each job.
4835 2) Executing a trigger command.
4837 The write state is relatively small, on the order of hundreds of bytes to single
4838 kilobytes. It contains information on the number of completions done, the last X
4841 A trigger is invoked either through creation ('touch') of a specified file in
4842 the system, or through a timeout setting. If fio is run with
4843 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
4844 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
4845 will fire off the trigger (thus saving state, and executing the trigger
4848 For client/server runs, there's both a local and remote trigger. If fio is
4849 running as a server backend, it will send the job states back to the client for
4850 safe storage, then execute the remote trigger, if specified. If a local trigger
4851 is specified, the server will still send back the write state, but the client
4852 will then execute the trigger.
4854 Verification trigger example
4855 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4857 Let's say we want to run a powercut test on the remote Linux machine 'server'.
4858 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
4859 some point during the run, and we'll run this test from the safety or our local
4860 machine, 'localbox'. On the server, we'll start the fio backend normally::
4862 server# fio --server
4864 and on the client, we'll fire off the workload::
4866 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
4868 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
4870 echo b > /proc/sysrq-trigger
4872 on the server once it has received the trigger and sent us the write state. This
4873 will work, but it's not **really** cutting power to the server, it's merely
4874 abruptly rebooting it. If we have a remote way of cutting power to the server
4875 through IPMI or similar, we could do that through a local trigger command
4876 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
4877 ipmi-reboot. On localbox, we could then have run fio with a local trigger
4880 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
4882 For this case, fio would wait for the server to send us the write state, then
4883 execute ``ipmi-reboot server`` when that happened.
4885 Loading verify state
4886 ~~~~~~~~~~~~~~~~~~~~
4888 To load stored write state, a read verification job file must contain the
4889 :option:`verify_state_load` option. If that is set, fio will load the previously
4890 stored state. For a local fio run this is done by loading the files directly,
4891 and on a client/server run, the server backend will ask the client to send the
4892 files over and load them from there.
4898 Fio supports a variety of log file formats, for logging latencies, bandwidth,
4899 and IOPS. The logs share a common format, which looks like this:
4901 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
4902 *offset* (`bytes`), *command priority*
4904 *Time* for the log entry is always in milliseconds. The *value* logged depends
4905 on the type of log, it will be one of the following:
4908 Value is latency in nsecs
4914 *Data direction* is one of the following:
4923 The entry's *block size* is always in bytes. The *offset* is the position in bytes
4924 from the start of the file for that particular I/O. The logging of the offset can be
4925 toggled with :option:`log_offset`.
4927 *Command priority* is 0 for normal priority and 1 for high priority. This is controlled
4928 by the ioengine specific :option:`cmdprio_percentage`.
4930 Fio defaults to logging every individual I/O but when windowed logging is set
4931 through :option:`log_avg_msec`, either the average (by default) or the maximum
4932 (:option:`log_max_value` is set) *value* seen over the specified period of time
4933 is recorded. Each *data direction* seen within the window period will aggregate
4934 its values in a separate row. Further, when using windowed logging the *block
4935 size* and *offset* entries will always contain 0.
4941 Normally fio is invoked as a stand-alone application on the machine where the
4942 I/O workload should be generated. However, the backend and frontend of fio can
4943 be run separately i.e., the fio server can generate an I/O workload on the "Device
4944 Under Test" while being controlled by a client on another machine.
4946 Start the server on the machine which has access to the storage DUT::
4950 where `args` defines what fio listens to. The arguments are of the form
4951 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
4952 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
4953 *hostname* is either a hostname or IP address, and *port* is the port to listen
4954 to (only valid for TCP/IP, not a local socket). Some examples:
4958 Start a fio server, listening on all interfaces on the default port (8765).
4960 2) ``fio --server=ip:hostname,4444``
4962 Start a fio server, listening on IP belonging to hostname and on port 4444.
4964 3) ``fio --server=ip6:::1,4444``
4966 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
4968 4) ``fio --server=,4444``
4970 Start a fio server, listening on all interfaces on port 4444.
4972 5) ``fio --server=1.2.3.4``
4974 Start a fio server, listening on IP 1.2.3.4 on the default port.
4976 6) ``fio --server=sock:/tmp/fio.sock``
4978 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
4980 Once a server is running, a "client" can connect to the fio server with::
4982 fio <local-args> --client=<server> <remote-args> <job file(s)>
4984 where `local-args` are arguments for the client where it is running, `server`
4985 is the connect string, and `remote-args` and `job file(s)` are sent to the
4986 server. The `server` string follows the same format as it does on the server
4987 side, to allow IP/hostname/socket and port strings.
4989 Fio can connect to multiple servers this way::
4991 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
4993 If the job file is located on the fio server, then you can tell the server to
4994 load a local file as well. This is done by using :option:`--remote-config` ::
4996 fio --client=server --remote-config /path/to/file.fio
4998 Then fio will open this local (to the server) job file instead of being passed
4999 one from the client.
5001 If you have many servers (example: 100 VMs/containers), you can input a pathname
5002 of a file containing host IPs/names as the parameter value for the
5003 :option:`--client` option. For example, here is an example :file:`host.list`
5004 file containing 2 hostnames::
5006 host1.your.dns.domain
5007 host2.your.dns.domain
5009 The fio command would then be::
5011 fio --client=host.list <job file(s)>
5013 In this mode, you cannot input server-specific parameters or job files -- all
5014 servers receive the same job file.
5016 In order to let ``fio --client`` runs use a shared filesystem from multiple
5017 hosts, ``fio --client`` now prepends the IP address of the server to the
5018 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
5019 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
5020 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
5021 192.168.10.121, then fio will create two files::
5023 /mnt/nfs/fio/192.168.10.120.fileio.tmp
5024 /mnt/nfs/fio/192.168.10.121.fileio.tmp
5026 Terse output in client/server mode will differ slightly from what is produced
5027 when fio is run in stand-alone mode. See the terse output section for details.