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`. This accepts only a
847 single directory and unlike related options, colons appearing in the
848 path must not be escaped.
850 .. option:: lockfile=str
852 Fio defaults to not locking any files before it does I/O to them. If a file
853 or file descriptor is shared, fio can serialize I/O to that file to make the
854 end result consistent. This is usual for emulating real workloads that share
855 files. The lock modes are:
858 No locking. The default.
860 Only one thread or process may do I/O at a time, excluding all
863 Read-write locking on the file. Many readers may
864 access the file at the same time, but writes get exclusive access.
866 .. option:: nrfiles=int
868 Number of files to use for this job. Defaults to 1. The size of files
869 will be :option:`size` divided by this unless explicit size is specified by
870 :option:`filesize`. Files are created for each thread separately, and each
871 file will have a file number within its name by default, as explained in
872 :option:`filename` section.
875 .. option:: openfiles=int
877 Number of files to keep open at the same time. Defaults to the same as
878 :option:`nrfiles`, can be set smaller to limit the number simultaneous
881 .. option:: file_service_type=str
883 Defines how fio decides which file from a job to service next. The following
887 Choose a file at random.
890 Round robin over opened files. This is the default.
893 Finish one file before moving on to the next. Multiple files can
894 still be open depending on :option:`openfiles`.
897 Use a *Zipf* distribution to decide what file to access.
900 Use a *Pareto* distribution to decide what file to access.
903 Use a *Gaussian* (normal) distribution to decide what file to
909 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
910 tell fio how many I/Os to issue before switching to a new file. For example,
911 specifying ``file_service_type=random:8`` would cause fio to issue
912 8 I/Os before selecting a new file at random. For the non-uniform
913 distributions, a floating point postfix can be given to influence how the
914 distribution is skewed. See :option:`random_distribution` for a description
915 of how that would work.
917 .. option:: ioscheduler=str
919 Attempt to switch the device hosting the file to the specified I/O scheduler
922 .. option:: create_serialize=bool
924 If true, serialize the file creation for the jobs. This may be handy to
925 avoid interleaving of data files, which may greatly depend on the filesystem
926 used and even the number of processors in the system. Default: true.
928 .. option:: create_fsync=bool
930 :manpage:`fsync(2)` the data file after creation. This is the default.
932 .. option:: create_on_open=bool
934 If true, don't pre-create files but allow the job's open() to create a file
935 when it's time to do I/O. Default: false -- pre-create all necessary files
938 .. option:: create_only=bool
940 If true, fio will only run the setup phase of the job. If files need to be
941 laid out or updated on disk, only that will be done -- the actual job contents
942 are not executed. Default: false.
944 .. option:: allow_file_create=bool
946 If true, fio is permitted to create files as part of its workload. If this
947 option is false, then fio will error out if
948 the files it needs to use don't already exist. Default: true.
950 .. option:: allow_mounted_write=bool
952 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
953 to what appears to be a mounted device or partition. This should help catch
954 creating inadvertently destructive tests, not realizing that the test will
955 destroy data on the mounted file system. Note that some platforms don't allow
956 writing against a mounted device regardless of this option. Default: false.
958 .. option:: pre_read=bool
960 If this is given, files will be pre-read into memory before starting the
961 given I/O operation. This will also clear the :option:`invalidate` flag,
962 since it is pointless to pre-read and then drop the cache. This will only
963 work for I/O engines that are seek-able, since they allow you to read the
964 same data multiple times. Thus it will not work on non-seekable I/O engines
965 (e.g. network, splice). Default: false.
967 .. option:: unlink=bool
969 Unlink the job files when done. Not the default, as repeated runs of that
970 job would then waste time recreating the file set again and again. Default:
973 .. option:: unlink_each_loop=bool
975 Unlink job files after each iteration or loop. Default: false.
977 .. option:: zonemode=str
982 The :option:`zonerange`, :option:`zonesize`,
983 :option `zonecapacity` and option:`zoneskip`
984 parameters are ignored.
986 I/O happens in a single zone until
987 :option:`zonesize` bytes have been transferred.
988 After that number of bytes has been
989 transferred processing of the next zone
990 starts. :option `zonecapacity` is ignored.
992 Zoned block device mode. I/O happens
993 sequentially in each zone, even if random I/O
994 has been selected. Random I/O happens across
995 all zones instead of being restricted to a
996 single zone. The :option:`zoneskip` parameter
997 is ignored. :option:`zonerange` and
998 :option:`zonesize` must be identical.
999 Trim is handled using a zone reset operation.
1000 Trim only considers non-empty sequential write
1001 required and sequential write preferred zones.
1003 .. option:: zonerange=int
1005 Size of a single zone. See also :option:`zonesize` and
1008 .. option:: zonesize=int
1010 For :option:`zonemode` =strided, this is the number of bytes to
1011 transfer before skipping :option:`zoneskip` bytes. If this parameter
1012 is smaller than :option:`zonerange` then only a fraction of each zone
1013 with :option:`zonerange` bytes will be accessed. If this parameter is
1014 larger than :option:`zonerange` then each zone will be accessed
1015 multiple times before skipping to the next zone.
1017 For :option:`zonemode` =zbd, this is the size of a single zone. The
1018 :option:`zonerange` parameter is ignored in this mode.
1021 .. option:: zonecapacity=int
1023 For :option:`zonemode` =zbd, this defines the capacity of a single zone,
1024 which is the accessible area starting from the zone start address.
1025 This parameter only applies when using :option:`zonemode` =zbd in
1026 combination with regular block devices. If not specified it defaults to
1027 the zone size. If the target device is a zoned block device, the zone
1028 capacity is obtained from the device information and this option is
1031 .. option:: zoneskip=int
1033 For :option:`zonemode` =strided, the number of bytes to skip after
1034 :option:`zonesize` bytes of data have been transferred. This parameter
1035 must be zero for :option:`zonemode` =zbd.
1037 .. option:: read_beyond_wp=bool
1039 This parameter applies to :option:`zonemode` =zbd only.
1041 Zoned block devices are block devices that consist of multiple zones.
1042 Each zone has a type, e.g. conventional or sequential. A conventional
1043 zone can be written at any offset that is a multiple of the block
1044 size. Sequential zones must be written sequentially. The position at
1045 which a write must occur is called the write pointer. A zoned block
1046 device can be either drive managed, host managed or host aware. For
1047 host managed devices the host must ensure that writes happen
1048 sequentially. Fio recognizes host managed devices and serializes
1049 writes to sequential zones for these devices.
1051 If a read occurs in a sequential zone beyond the write pointer then
1052 the zoned block device will complete the read without reading any data
1053 from the storage medium. Since such reads lead to unrealistically high
1054 bandwidth and IOPS numbers fio only reads beyond the write pointer if
1055 explicitly told to do so. Default: false.
1057 .. option:: max_open_zones=int
1059 When a zone of a zoned block device is partially written (i.e. not all
1060 sectors of the zone have been written), the zone is in one of three
1061 conditions: 'implicit open', 'explicit open' or 'closed'. Zoned block
1062 devices may have a limit called 'max_open_zones' (same name as the
1063 parameter) on the total number of zones that can simultaneously be in
1064 the 'implicit open' or 'explicit open' conditions. Zoned block devices
1065 may have another limit called 'max_active_zones', on the total number of
1066 zones that can simultaneously be in the three conditions. The
1067 :option:`max_open_zones` parameter limits the number of zones to which
1068 write commands are issued by all fio jobs, that is, limits the number of
1069 zones that will be in the conditions. When the device has the
1070 max_open_zones limit and does not have the max_active_zones limit, the
1071 :option:`max_open_zones` parameter limits the number of zones in the two
1072 open conditions up to the limit. In this case, fio includes zones in the
1073 two open conditions to the write target zones at fio start. When the
1074 device has both the max_open_zones and the max_active_zones limits, the
1075 :option:`max_open_zones` parameter limits the number of zones in the
1076 three conditions up to the limit. In this case, fio includes zones in
1077 the three conditions to the write target zones at fio start.
1079 This parameter is relevant only if the :option:`zonemode` =zbd is used.
1080 The default value is always equal to the max_open_zones limit of the
1081 target zoned block device and a value higher than this limit cannot be
1082 specified by users unless the option :option:`ignore_zone_limits` is
1083 specified. When :option:`ignore_zone_limits` is specified or the target
1084 device does not have the max_open_zones limit, :option:`max_open_zones`
1085 can specify 0 to disable any limit on the number of zones that can be
1086 simultaneously written to by all jobs.
1088 .. option:: job_max_open_zones=int
1090 In the same manner as :option:`max_open_zones`, limit the number of open
1091 zones per fio job, that is, the number of zones that a single job can
1092 simultaneously write to. A value of zero indicates no limit.
1095 .. option:: ignore_zone_limits=bool
1097 If this option is used, fio will ignore the maximum number of open
1098 zones limit of the zoned block device in use, thus allowing the
1099 option :option:`max_open_zones` value to be larger than the device
1100 reported limit. Default: false.
1102 .. option:: zone_reset_threshold=float
1104 A number between zero and one that indicates the ratio of written bytes
1105 in the zones with write pointers in the IO range to the size of the IO
1106 range. When current ratio is above this ratio, zones are reset
1107 periodically as :option:`zone_reset_frequency` specifies. If there are
1108 multiple jobs when using this option, the IO range for all write jobs
1111 .. option:: zone_reset_frequency=float
1113 A number between zero and one that indicates how often a zone reset
1114 should be issued if the zone reset threshold has been exceeded. A zone
1115 reset is submitted after each (1 / zone_reset_frequency) write
1116 requests. This and the previous parameter can be used to simulate
1117 garbage collection activity.
1123 .. option:: direct=bool
1125 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
1126 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
1127 ioengines don't support direct I/O. Default: false.
1129 .. option:: buffered=bool
1131 If value is true, use buffered I/O. This is the opposite of the
1132 :option:`direct` option. Defaults to true.
1134 .. option:: readwrite=str, rw=str
1136 Type of I/O pattern. Accepted values are:
1143 Sequential trims (Linux block devices and SCSI
1144 character devices only).
1150 Random trims (Linux block devices and SCSI
1151 character devices only).
1153 Sequential mixed reads and writes.
1155 Random mixed reads and writes.
1157 Sequential trim+write sequences. Blocks will be trimmed first,
1158 then the same blocks will be written to. So if ``io_size=64K``
1159 is specified, Fio will trim a total of 64K bytes and also
1160 write 64K bytes on the same trimmed blocks. This behaviour
1161 will be consistent with ``number_ios`` or other Fio options
1162 limiting the total bytes or number of I/O's.
1164 Like trimwrite, but uses random offsets rather
1165 than sequential writes.
1167 Fio defaults to read if the option is not specified. For the mixed I/O
1168 types, the default is to split them 50/50. For certain types of I/O the
1169 result may still be skewed a bit, since the speed may be different.
1171 It is possible to specify the number of I/Os to do before getting a new
1172 offset by appending ``:<nr>`` to the end of the string given. For a
1173 random read, it would look like ``rw=randread:8`` for passing in an offset
1174 modifier with a value of 8. If the suffix is used with a sequential I/O
1175 pattern, then the *<nr>* value specified will be **added** to the generated
1176 offset for each I/O turning sequential I/O into sequential I/O with holes.
1177 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1178 the :option:`rw_sequencer` option.
1180 .. option:: rw_sequencer=str
1182 If an offset modifier is given by appending a number to the ``rw=<str>``
1183 line, then this option controls how that number modifies the I/O offset
1184 being generated. Accepted values are:
1187 Generate sequential offset.
1189 Generate the same offset.
1191 ``sequential`` is only useful for random I/O, where fio would normally
1192 generate a new random offset for every I/O. If you append e.g. 8 to
1193 randread, i.e. ``rw=randread:8`` you would get a new random offset for
1194 every 8 I/Os. The result would be a sequence of 8 sequential offsets
1195 with a random starting point. However this behavior may change if a
1196 sequential I/O reaches end of the file. As sequential I/O is already
1197 sequential, setting ``sequential`` for that would not result in any
1198 difference. ``identical`` behaves in a similar fashion, except it sends
1199 the same offset 8 number of times before generating a new offset.
1204 rw_sequencer=sequential
1207 The generated sequence of offsets will look like this:
1208 4k, 8k, 12k, 16k, 20k, 24k, 28k, 32k, 92k, 96k, 100k, 104k, 108k,
1209 112k, 116k, 120k, 48k, 52k ...
1214 rw_sequencer=identical
1217 The generated sequence of offsets will look like this:
1218 4k, 4k, 4k, 4k, 4k, 4k, 4k, 4k, 92k, 92k, 92k, 92k, 92k, 92k, 92k, 92k,
1221 .. option:: unified_rw_reporting=str
1223 Fio normally reports statistics on a per data direction basis, meaning that
1224 reads, writes, and trims are accounted and reported separately. This option
1225 determines whether fio reports the results normally, summed together, or as
1227 Accepted values are:
1230 Normal statistics reporting.
1233 Statistics are summed per data direction and reported together.
1236 Statistics are reported normally, followed by the mixed statistics.
1239 Backward-compatible alias for **none**.
1242 Backward-compatible alias for **mixed**.
1247 .. option:: randrepeat=bool
1249 Seed all random number generators in a predictable way so the pattern
1250 is repeatable across runs. Default: true.
1252 .. option:: allrandrepeat=bool
1254 Alias for :option:`randrepeat`. Default: true.
1256 .. option:: randseed=int
1258 Seed the random number generators based on this seed value, to be able to
1259 control what sequence of output is being generated. If not set, the random
1260 sequence depends on the :option:`randrepeat` setting.
1262 .. option:: fallocate=str
1264 Whether pre-allocation is performed when laying down files.
1265 Accepted values are:
1268 Do not pre-allocate space.
1271 Use a platform's native pre-allocation call but fall back to
1272 **none** behavior if it fails/is not implemented.
1275 Pre-allocate via :manpage:`posix_fallocate(3)`.
1278 Pre-allocate via :manpage:`fallocate(2)` with
1279 FALLOC_FL_KEEP_SIZE set.
1282 Extend file to final size via :manpage:`ftruncate(2)`
1283 instead of allocating.
1286 Backward-compatible alias for **none**.
1289 Backward-compatible alias for **posix**.
1291 May not be available on all supported platforms. **keep** is only available
1292 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1293 because ZFS doesn't support pre-allocation. Default: **native** if any
1294 pre-allocation methods except **truncate** are available, **none** if not.
1296 Note that using **truncate** on Windows will interact surprisingly
1297 with non-sequential write patterns. When writing to a file that has
1298 been extended by setting the end-of-file information, Windows will
1299 backfill the unwritten portion of the file up to that offset with
1300 zeroes before issuing the new write. This means that a single small
1301 write to the end of an extended file will stall until the entire
1302 file has been filled with zeroes.
1304 .. option:: fadvise_hint=str
1306 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1307 advise the kernel on what I/O patterns are likely to be issued.
1308 Accepted values are:
1311 Backwards-compatible hint for "no hint".
1314 Backwards compatible hint for "advise with fio workload type". This
1315 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1316 for a sequential workload.
1319 Advise using **FADV_SEQUENTIAL**.
1322 Advise using **FADV_RANDOM**.
1325 Advise using **FADV_NOREUSE**. This may be a no-op on older Linux
1326 kernels. Since Linux 6.3, it provides a hint to the LRU algorithm.
1327 See the :manpage:`posix_fadvise(2)` man page.
1329 .. option:: write_hint=str
1331 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1332 from a write. Only supported on Linux, as of version 4.13. Accepted
1336 No particular life time associated with this file.
1339 Data written to this file has a short life time.
1342 Data written to this file has a medium life time.
1345 Data written to this file has a long life time.
1348 Data written to this file has a very long life time.
1350 The values are all relative to each other, and no absolute meaning
1351 should be associated with them.
1353 .. option:: offset=int
1355 Start I/O at the provided offset in the file, given as either a fixed size in
1356 bytes, zones or a percentage. If a percentage is given, the generated offset will be
1357 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1358 provided. Data before the given offset will not be touched. This
1359 effectively caps the file size at `real_size - offset`. Can be combined with
1360 :option:`size` to constrain the start and end range of the I/O workload.
1361 A percentage can be specified by a number between 1 and 100 followed by '%',
1362 for example, ``offset=20%`` to specify 20%. In ZBD mode, value can be set as
1363 number of zones using 'z'.
1365 .. option:: offset_align=int
1367 If set to non-zero value, the byte offset generated by a percentage ``offset``
1368 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1369 offset is aligned to the minimum block size.
1371 .. option:: offset_increment=int
1373 If this is provided, then the real offset becomes `offset + offset_increment
1374 * thread_number`, where the thread number is a counter that starts at 0 and
1375 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1376 specified). This option is useful if there are several jobs which are
1377 intended to operate on a file in parallel disjoint segments, with even
1378 spacing between the starting points. Percentages can be used for this option.
1379 If a percentage is given, the generated offset will be aligned to the minimum
1380 ``blocksize`` or to the value of ``offset_align`` if provided. In ZBD mode, value can
1381 also be set as number of zones using 'z'.
1383 .. option:: number_ios=int
1385 Fio will normally perform I/Os until it has exhausted the size of the region
1386 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1387 condition). With this setting, the range/size can be set independently of
1388 the number of I/Os to perform. When fio reaches this number, it will exit
1389 normally and report status. Note that this does not extend the amount of I/O
1390 that will be done, it will only stop fio if this condition is met before
1391 other end-of-job criteria.
1393 .. option:: fsync=int
1395 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1396 the dirty data for every number of blocks given. For example, if you give 32
1397 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1398 using non-buffered I/O, we may not sync the file. The exception is the sg
1399 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1400 means fio does not periodically issue and wait for a sync to complete. Also
1401 see :option:`end_fsync` and :option:`fsync_on_close`.
1403 .. option:: fdatasync=int
1405 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1406 not metadata blocks. In Windows, DragonFlyBSD or OSX there is no
1407 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1408 Defaults to 0, which means fio does not periodically issue and wait for a
1409 data-only sync to complete.
1411 .. option:: write_barrier=int
1413 Make every `N-th` write a barrier write.
1415 .. option:: sync_file_range=str:int
1417 Use :manpage:`sync_file_range(2)` for every `int` number of write
1418 operations. Fio will track range of writes that have happened since the last
1419 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1422 SYNC_FILE_RANGE_WAIT_BEFORE
1424 SYNC_FILE_RANGE_WRITE
1426 SYNC_FILE_RANGE_WAIT_AFTER
1428 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1429 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1430 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1433 .. option:: overwrite=bool
1435 If true, writes to a file will always overwrite existing data. If the file
1436 doesn't already exist, it will be created before the write phase begins. If
1437 the file exists and is large enough for the specified write phase, nothing
1438 will be done. Default: false.
1440 .. option:: end_fsync=bool
1442 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1445 .. option:: fsync_on_close=bool
1447 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1448 from :option:`end_fsync` in that it will happen on every file close, not
1449 just at the end of the job. Default: false.
1451 .. option:: rwmixread=int
1453 Percentage of a mixed workload that should be reads. Default: 50.
1455 .. option:: rwmixwrite=int
1457 Percentage of a mixed workload that should be writes. If both
1458 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1459 add up to 100%, the latter of the two will be used to override the
1460 first. This may interfere with a given rate setting, if fio is asked to
1461 limit reads or writes to a certain rate. If that is the case, then the
1462 distribution may be skewed. Default: 50.
1464 .. option:: random_distribution=str:float[:float][,str:float][,str:float]
1466 By default, fio will use a completely uniform random distribution when asked
1467 to perform random I/O. Sometimes it is useful to skew the distribution in
1468 specific ways, ensuring that some parts of the data is more hot than others.
1469 fio includes the following distribution models:
1472 Uniform random distribution
1481 Normal (Gaussian) distribution
1484 Zoned random distribution
1487 Zone absolute random distribution
1489 When using a **zipf** or **pareto** distribution, an input value is also
1490 needed to define the access pattern. For **zipf**, this is the `Zipf
1491 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1492 program, :command:`fio-genzipf`, that can be used visualize what the given input
1493 values will yield in terms of hit rates. If you wanted to use **zipf** with
1494 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1495 option. If a non-uniform model is used, fio will disable use of the random
1496 map. For the **normal** distribution, a normal (Gaussian) deviation is
1497 supplied as a value between 0 and 100.
1499 The second, optional float is allowed for **pareto**, **zipf** and **normal** distributions.
1500 It allows one to set base of distribution in non-default place, giving more control
1501 over most probable outcome. This value is in range [0-1] which maps linearly to
1502 range of possible random values.
1503 Defaults are: random for **pareto** and **zipf**, and 0.5 for **normal**.
1504 If you wanted to use **zipf** with a `theta` of 1.2 centered on 1/4 of allowed value range,
1505 you would use ``random_distribution=zipf:1.2:0.25``.
1507 For a **zoned** distribution, fio supports specifying percentages of I/O
1508 access that should fall within what range of the file or device. For
1509 example, given a criteria of:
1511 * 60% of accesses should be to the first 10%
1512 * 30% of accesses should be to the next 20%
1513 * 8% of accesses should be to the next 30%
1514 * 2% of accesses should be to the next 40%
1516 we can define that through zoning of the random accesses. For the above
1517 example, the user would do::
1519 random_distribution=zoned:60/10:30/20:8/30:2/40
1521 A **zoned_abs** distribution works exactly like the **zoned**, except
1522 that it takes absolute sizes. For example, let's say you wanted to
1523 define access according to the following criteria:
1525 * 60% of accesses should be to the first 20G
1526 * 30% of accesses should be to the next 100G
1527 * 10% of accesses should be to the next 500G
1529 we can define an absolute zoning distribution with:
1531 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1533 For both **zoned** and **zoned_abs**, fio supports defining up to
1536 Similarly to how :option:`bssplit` works for setting ranges and
1537 percentages of block sizes. Like :option:`bssplit`, it's possible to
1538 specify separate zones for reads, writes, and trims. If just one set
1539 is given, it'll apply to all of them. This goes for both **zoned**
1540 **zoned_abs** distributions.
1542 .. option:: percentage_random=int[,int][,int]
1544 For a random workload, set how big a percentage should be random. This
1545 defaults to 100%, in which case the workload is fully random. It can be set
1546 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1547 sequential. Any setting in between will result in a random mix of sequential
1548 and random I/O, at the given percentages. Comma-separated values may be
1549 specified for reads, writes, and trims as described in :option:`blocksize`.
1551 .. option:: norandommap
1553 Normally fio will cover every block of the file when doing random I/O. If
1554 this option is given, fio will just get a new random offset without looking
1555 at past I/O history. This means that some blocks may not be read or written,
1556 and that some blocks may be read/written more than once. If this option is
1557 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1558 only intact blocks are verified, i.e., partially-overwritten blocks are
1559 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1560 the same block to be overwritten, which can cause verification errors. Either
1561 do not use norandommap in this case, or also use the lfsr random generator.
1563 .. option:: softrandommap=bool
1565 See :option:`norandommap`. If fio runs with the random block map enabled and
1566 it fails to allocate the map, if this option is set it will continue without
1567 a random block map. As coverage will not be as complete as with random maps,
1568 this option is disabled by default.
1570 .. option:: random_generator=str
1572 Fio supports the following engines for generating I/O offsets for random I/O:
1575 Strong 2^88 cycle random number generator.
1577 Linear feedback shift register generator.
1579 Strong 64-bit 2^258 cycle random number generator.
1581 **tausworthe** is a strong random number generator, but it requires tracking
1582 on the side if we want to ensure that blocks are only read or written
1583 once. **lfsr** guarantees that we never generate the same offset twice, and
1584 it's also less computationally expensive. It's not a true random generator,
1585 however, though for I/O purposes it's typically good enough. **lfsr** only
1586 works with single block sizes, not with workloads that use multiple block
1587 sizes. If used with such a workload, fio may read or write some blocks
1588 multiple times. The default value is **tausworthe**, unless the required
1589 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1590 selected automatically.
1596 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1598 The block size in bytes used for I/O units. Default: 4096. A single value
1599 applies to reads, writes, and trims. Comma-separated values may be
1600 specified for reads, writes, and trims. A value not terminated in a comma
1601 applies to subsequent types.
1606 means 256k for reads, writes and trims.
1609 means 8k for reads, 32k for writes and trims.
1612 means 8k for reads, 32k for writes, and default for trims.
1615 means default for reads, 8k for writes and trims.
1618 means default for reads, 8k for writes, and default for trims.
1620 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1622 A range of block sizes in bytes for I/O units. The issued I/O unit will
1623 always be a multiple of the minimum size, unless
1624 :option:`blocksize_unaligned` is set.
1626 Comma-separated ranges may be specified for reads, writes, and trims as
1627 described in :option:`blocksize`.
1629 Example: ``bsrange=1k-4k,2k-8k``.
1631 .. option:: bssplit=str[,str][,str]
1633 Sometimes you want even finer grained control of the block sizes
1634 issued, not just an even split between them. This option allows you to
1635 weight various block sizes, so that you are able to define a specific
1636 amount of block sizes issued. The format for this option is::
1638 bssplit=blocksize/percentage:blocksize/percentage
1640 for as many block sizes as needed. So if you want to define a workload
1641 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1644 bssplit=4k/10:64k/50:32k/40
1646 Ordering does not matter. If the percentage is left blank, fio will
1647 fill in the remaining values evenly. So a bssplit option like this one::
1649 bssplit=4k/50:1k/:32k/
1651 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1652 add up to 100, if bssplit is given a range that adds up to more, it
1655 Comma-separated values may be specified for reads, writes, and trims as
1656 described in :option:`blocksize`.
1658 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1659 having 90% 4k writes and 10% 8k writes, you would specify::
1661 bssplit=2k/50:4k/50,4k/90:8k/10
1663 Fio supports defining up to 64 different weights for each data
1666 .. option:: blocksize_unaligned, bs_unaligned
1668 If set, fio will issue I/O units with any size within
1669 :option:`blocksize_range`, not just multiples of the minimum size. This
1670 typically won't work with direct I/O, as that normally requires sector
1673 .. option:: bs_is_seq_rand=bool
1675 If this option is set, fio will use the normal read,write blocksize settings
1676 as sequential,random blocksize settings instead. Any random read or write
1677 will use the WRITE blocksize settings, and any sequential read or write will
1678 use the READ blocksize settings.
1680 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1682 Boundary to which fio will align random I/O units. Default:
1683 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1684 I/O, though it usually depends on the hardware block size. This option is
1685 mutually exclusive with using a random map for files, so it will turn off
1686 that option. Comma-separated values may be specified for reads, writes, and
1687 trims as described in :option:`blocksize`.
1693 .. option:: zero_buffers
1695 Initialize buffers with all zeros. Default: fill buffers with random data.
1697 .. option:: refill_buffers
1699 If this option is given, fio will refill the I/O buffers on every
1700 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1701 naturally. Defaults to being unset i.e., the buffer is only filled at
1702 init time and the data in it is reused when possible but if any of
1703 :option:`verify`, :option:`buffer_compress_percentage` or
1704 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1705 automatically enabled.
1707 .. option:: scramble_buffers=bool
1709 If :option:`refill_buffers` is too costly and the target is using data
1710 deduplication, then setting this option will slightly modify the I/O buffer
1711 contents to defeat normal de-dupe attempts. This is not enough to defeat
1712 more clever block compression attempts, but it will stop naive dedupe of
1713 blocks. Default: true.
1715 .. option:: buffer_compress_percentage=int
1717 If this is set, then fio will attempt to provide I/O buffer content
1718 (on WRITEs) that compresses to the specified level. Fio does this by
1719 providing a mix of random data followed by fixed pattern data. The
1720 fixed pattern is either zeros, or the pattern specified by
1721 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1722 might skew the compression ratio slightly. Setting
1723 `buffer_compress_percentage` to a value other than 100 will also
1724 enable :option:`refill_buffers` in order to reduce the likelihood that
1725 adjacent blocks are so similar that they over compress when seen
1726 together. See :option:`buffer_compress_chunk` for how to set a finer or
1727 coarser granularity for the random/fixed data region. Defaults to unset
1728 i.e., buffer data will not adhere to any compression level.
1730 .. option:: buffer_compress_chunk=int
1732 This setting allows fio to manage how big the random/fixed data region
1733 is when using :option:`buffer_compress_percentage`. When
1734 `buffer_compress_chunk` is set to some non-zero value smaller than the
1735 block size, fio can repeat the random/fixed region throughout the I/O
1736 buffer at the specified interval (which particularly useful when
1737 bigger block sizes are used for a job). When set to 0, fio will use a
1738 chunk size that matches the block size resulting in a single
1739 random/fixed region within the I/O buffer. Defaults to 512. When the
1740 unit is omitted, the value is interpreted in bytes.
1742 .. option:: buffer_pattern=str
1744 If set, fio will fill the I/O buffers with this pattern or with the contents
1745 of a file. If not set, the contents of I/O buffers are defined by the other
1746 options related to buffer contents. The setting can be any pattern of bytes,
1747 and can be prefixed with 0x for hex values. It may also be a string, where
1748 the string must then be wrapped with ``""``. Or it may also be a filename,
1749 where the filename must be wrapped with ``''`` in which case the file is
1750 opened and read. Note that not all the file contents will be read if that
1751 would cause the buffers to overflow. So, for example::
1753 buffer_pattern='filename'
1757 buffer_pattern="abcd"
1765 buffer_pattern=0xdeadface
1767 Also you can combine everything together in any order::
1769 buffer_pattern=0xdeadface"abcd"-12'filename'
1771 .. option:: dedupe_percentage=int
1773 If set, fio will generate this percentage of identical buffers when
1774 writing. These buffers will be naturally dedupable. The contents of the
1775 buffers depend on what other buffer compression settings have been set. It's
1776 possible to have the individual buffers either fully compressible, or not at
1777 all -- this option only controls the distribution of unique buffers. Setting
1778 this option will also enable :option:`refill_buffers` to prevent every buffer
1781 .. option:: dedupe_mode=str
1783 If ``dedupe_percentage=<int>`` is given, then this option controls how fio
1784 generates the dedupe buffers.
1787 Generate dedupe buffers by repeating previous writes
1789 Generate dedupe buffers from working set
1791 ``repeat`` is the default option for fio. Dedupe buffers are generated
1792 by repeating previous unique write.
1794 ``working_set`` is a more realistic workload.
1795 With ``working_set``, ``dedupe_working_set_percentage=<int>`` should be provided.
1796 Given that, fio will use the initial unique write buffers as its working set.
1797 Upon deciding to dedupe, fio will randomly choose a buffer from the working set.
1798 Note that by using ``working_set`` the dedupe percentage will converge
1799 to the desired over time while ``repeat`` maintains the desired percentage
1802 .. option:: dedupe_working_set_percentage=int
1804 If ``dedupe_mode=<str>`` is set to ``working_set``, then this controls
1805 the percentage of size of the file or device used as the buffers
1806 fio will choose to generate the dedupe buffers from
1808 Note that size needs to be explicitly provided and only 1 file per
1811 .. option:: dedupe_global=bool
1813 This controls whether the deduplication buffers will be shared amongst
1814 all jobs that have this option set. The buffers are spread evenly between
1817 .. option:: invalidate=bool
1819 Invalidate the buffer/page cache parts of the files to be used prior to
1820 starting I/O if the platform and file type support it. Defaults to true.
1821 This will be ignored if :option:`pre_read` is also specified for the
1824 .. option:: sync=str
1826 Whether, and what type, of synchronous I/O to use for writes. The allowed
1830 Do not use synchronous IO, the default.
1836 Use synchronous file IO. For the majority of I/O engines,
1837 this means using O_SYNC.
1843 Use synchronous data IO. For the majority of I/O engines,
1844 this means using O_DSYNC.
1847 .. option:: iomem=str, mem=str
1849 Fio can use various types of memory as the I/O unit buffer. The allowed
1853 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1857 Use shared memory as the buffers. Allocated through
1858 :manpage:`shmget(2)`.
1861 Same as shm, but use huge pages as backing.
1864 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1865 be file backed if a filename is given after the option. The format
1866 is `mem=mmap:/path/to/file`.
1869 Use a memory mapped huge file as the buffer backing. Append filename
1870 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1873 Same as mmap, but use a MMAP_SHARED mapping.
1876 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1877 The :option:`ioengine` must be `rdma`.
1879 The area allocated is a function of the maximum allowed bs size for the job,
1880 multiplied by the I/O depth given. Note that for **shmhuge** and
1881 **mmaphuge** to work, the system must have free huge pages allocated. This
1882 can normally be checked and set by reading/writing
1883 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1884 is 2 or 4MiB in size depending on the platform. So to calculate the
1885 number of huge pages you need for a given job file, add up the I/O
1886 depth of all jobs (normally one unless :option:`iodepth` is used) and
1887 multiply by the maximum bs set. Then divide that number by the huge
1888 page size. You can see the size of the huge pages in
1889 :file:`/proc/meminfo`. If no huge pages are allocated by having a
1890 non-zero number in `nr_hugepages`, using **mmaphuge** or **shmhuge**
1891 will fail. Also see :option:`hugepage-size`.
1893 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1894 should point there. So if it's mounted in :file:`/huge`, you would use
1895 `mem=mmaphuge:/huge/somefile`.
1897 .. option:: iomem_align=int, mem_align=int
1899 This indicates the memory alignment of the I/O memory buffers. Note that
1900 the given alignment is applied to the first I/O unit buffer, if using
1901 :option:`iodepth` the alignment of the following buffers are given by the
1902 :option:`bs` used. In other words, if using a :option:`bs` that is a
1903 multiple of the page sized in the system, all buffers will be aligned to
1904 this value. If using a :option:`bs` that is not page aligned, the alignment
1905 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1908 .. option:: hugepage-size=int
1910 Defines the size of a huge page. Must at least be equal to the system
1911 setting, see :file:`/proc/meminfo` and
1912 :file:`/sys/kernel/mm/hugepages/`. Defaults to 2 or 4MiB depending on
1913 the platform. Should probably always be a multiple of megabytes, so
1914 using ``hugepage-size=Xm`` is the preferred way to set this to avoid
1915 setting a non-pow-2 bad value.
1917 .. option:: lockmem=int
1919 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1920 simulate a smaller amount of memory. The amount specified is per worker.
1926 .. option:: size=int
1928 The total size of file I/O for each thread of this job. Fio will run until
1929 this many bytes has been transferred, unless runtime is altered by other means
1930 such as (1) :option:`runtime`, (2) :option:`io_size` (3) :option:`number_ios`,
1931 (4) gaps/holes while doing I/O's such as ``rw=read:16K``, or (5) sequential
1932 I/O reaching end of the file which is possible when :option:`percentage_random`
1934 Fio will divide this size between the available files determined by options
1935 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1936 specified by the job. If the result of division happens to be 0, the size is
1937 set to the physical size of the given files or devices if they exist.
1938 If this option is not specified, fio will use the full size of the given
1939 files or devices. If the files do not exist, size must be given. It is also
1940 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1941 given, fio will use 20% of the full size of the given files or devices.
1942 In ZBD mode, value can also be set as number of zones using 'z'.
1943 Can be combined with :option:`offset` to constrain the start and end range
1944 that I/O will be done within.
1946 .. option:: io_size=int, io_limit=int
1948 Normally fio operates within the region set by :option:`size`, which means
1949 that the :option:`size` option sets both the region and size of I/O to be
1950 performed. Sometimes that is not what you want. With this option, it is
1951 possible to define just the amount of I/O that fio should do. For instance,
1952 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1953 will perform I/O within the first 20GiB but exit when 5GiB have been
1954 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1955 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1956 the 0..20GiB region.
1958 .. option:: filesize=irange(int)
1960 Individual file sizes. May be a range, in which case fio will select sizes for
1961 files at random within the given range. If not given, each created file is the
1962 same size. This option overrides :option:`size` in terms of file size, i.e. if
1963 :option:`filesize` is specified then :option:`size` becomes merely the default
1964 for :option:`io_size` and has no effect at all if :option:`io_size` is set
1967 .. option:: file_append=bool
1969 Perform I/O after the end of the file. Normally fio will operate within the
1970 size of a file. If this option is set, then fio will append to the file
1971 instead. This has identical behavior to setting :option:`offset` to the size
1972 of a file. This option is ignored on non-regular files.
1974 .. option:: fill_device=bool, fill_fs=bool
1976 Sets size to something really large and waits for ENOSPC (no space left on
1977 device) or EDQUOT (disk quota exceeded)
1978 as the terminating condition. Only makes sense with sequential
1979 write. For a read workload, the mount point will be filled first then I/O
1980 started on the result. This option doesn't make sense if operating on a raw
1981 device node, since the size of that is already known by the file system.
1982 Additionally, writing beyond end-of-device will not return ENOSPC there.
1988 .. option:: ioengine=str
1990 Defines how the job issues I/O to the file. The following types are defined:
1993 Basic :manpage:`read(2)` or :manpage:`write(2)`
1994 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1995 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1998 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1999 all supported operating systems except for Windows.
2002 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
2003 queuing by coalescing adjacent I/Os into a single submission.
2006 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
2009 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
2012 Fast Linux native asynchronous I/O. Supports async IO
2013 for both direct and buffered IO.
2014 This engine defines engine specific options.
2017 Fast Linux native asynchronous I/O for pass through commands.
2018 This engine defines engine specific options.
2021 Linux native asynchronous I/O. Note that Linux may only support
2022 queued behavior with non-buffered I/O (set ``direct=1`` or
2024 This engine defines engine specific options.
2027 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
2028 :manpage:`aio_write(3)`.
2031 Solaris native asynchronous I/O.
2034 Windows native asynchronous I/O. Default on Windows.
2037 File is memory mapped with :manpage:`mmap(2)` and data copied
2038 to/from using :manpage:`memcpy(3)`.
2041 :manpage:`splice(2)` is used to transfer the data and
2042 :manpage:`vmsplice(2)` to transfer data from user space to the
2046 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
2047 ioctl, or if the target is an sg character device we use
2048 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
2049 I/O. Requires :option:`filename` option to specify either block or
2050 character devices. This engine supports trim operations.
2051 The sg engine includes engine specific options.
2054 Read, write, trim and ZBC/ZAC operations to a zoned
2055 block device using libzbc library. The target can be
2056 either an SG character device or a block device file.
2059 Doesn't transfer any data, just pretends to. This is mainly used to
2060 exercise fio itself and for debugging/testing purposes.
2063 Transfer over the network to given ``host:port``. Depending on the
2064 :option:`protocol` used, the :option:`hostname`, :option:`port`,
2065 :option:`listen` and :option:`filename` options are used to specify
2066 what sort of connection to make, while the :option:`protocol` option
2067 determines which protocol will be used. This engine defines engine
2071 Like **net**, but uses :manpage:`splice(2)` and
2072 :manpage:`vmsplice(2)` to map data and send/receive.
2073 This engine defines engine specific options.
2076 Doesn't transfer any data, but burns CPU cycles according to the
2077 :option:`cpuload`, :option:`cpuchunks` and :option:`cpumode` options.
2078 Setting :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
2079 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
2080 to get desired CPU usage, as the cpuload only loads a
2081 single CPU at the desired rate. A job never finishes unless there is
2082 at least one non-cpuio job.
2083 Setting :option:`cpumode`\=qsort replace the default noop instructions loop
2084 by a qsort algorithm to consume more energy.
2087 The RDMA I/O engine supports both RDMA memory semantics
2088 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
2089 InfiniBand, RoCE and iWARP protocols. This engine defines engine
2093 I/O engine that does regular fallocate to simulate data transfer as
2097 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
2100 does fallocate(,mode = 0).
2103 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
2106 I/O engine that sends :manpage:`ftruncate(2)` operations in response
2107 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
2108 size to the current block offset. :option:`blocksize` is ignored.
2111 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
2112 defragment activity in request to DDIR_WRITE event.
2115 I/O engine supporting direct access to Ceph Reliable Autonomic
2116 Distributed Object Store (RADOS) via librados. This ioengine
2117 defines engine specific options.
2120 I/O engine supporting direct access to Ceph Rados Block Devices
2121 (RBD) via librbd without the need to use the kernel rbd driver. This
2122 ioengine defines engine specific options.
2125 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
2126 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
2128 This engine only supports direct IO of iodepth=1; you need to scale this
2129 via numjobs. blocksize defines the size of the objects to be created.
2131 TRIM is translated to object deletion.
2134 Using GlusterFS libgfapi sync interface to direct access to
2135 GlusterFS volumes without having to go through FUSE. This ioengine
2136 defines engine specific options.
2139 Using GlusterFS libgfapi async interface to direct access to
2140 GlusterFS volumes without having to go through FUSE. This ioengine
2141 defines engine specific options.
2144 Read and write through Hadoop (HDFS). The :option:`filename` option
2145 is used to specify host,port of the hdfs name-node to connect. This
2146 engine interprets offsets a little differently. In HDFS, files once
2147 created cannot be modified so random writes are not possible. To
2148 imitate this the libhdfs engine expects a bunch of small files to be
2149 created over HDFS and will randomly pick a file from them
2150 based on the offset generated by fio backend (see the example
2151 job file to create such files, use ``rw=write`` option). Please
2152 note, it may be necessary to set environment variables to work
2153 with HDFS/libhdfs properly. Each job uses its own connection to
2157 Read, write and erase an MTD character device (e.g.,
2158 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
2159 underlying device type, the I/O may have to go in a certain pattern,
2160 e.g., on NAND, writing sequentially to erase blocks and discarding
2161 before overwriting. The `trimwrite` mode works well for this
2165 Read and write using device DAX to a persistent memory device (e.g.,
2166 /dev/dax0.0) through the PMDK libpmem library.
2169 Prefix to specify loading an external I/O engine object file. Append
2170 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
2171 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
2172 absolute or relative. See :file:`engines/skeleton_external.c` for
2173 details of writing an external I/O engine.
2176 Simply create the files and do no I/O to them. You still need to
2177 set `filesize` so that all the accounting still occurs, but no
2178 actual I/O will be done other than creating the file.
2181 Simply do stat() and do no I/O to the file. You need to set 'filesize'
2182 and 'nrfiles', so that files will be created.
2183 This engine is to measure file lookup and meta data access.
2186 Simply delete the files by unlink() and do no I/O to them. You need to set 'filesize'
2187 and 'nrfiles', so that the files will be created.
2188 This engine is to measure file delete.
2191 Read and write using mmap I/O to a file on a filesystem
2192 mounted with DAX on a persistent memory device through the PMDK
2196 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2197 This engine is very basic and issues calls to IME whenever an IO is
2201 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2202 This engine uses iovecs and will try to stack as much IOs as possible
2203 (if the IOs are "contiguous" and the IO depth is not exceeded)
2204 before issuing a call to IME.
2207 Asynchronous read and write using DDN's Infinite Memory Engine (IME).
2208 This engine will try to stack as much IOs as possible by creating
2209 requests for IME. FIO will then decide when to commit these requests.
2212 Read and write iscsi lun with libiscsi.
2215 Read and write a Network Block Device (NBD).
2218 I/O engine supporting libcufile synchronous access to nvidia-fs and a
2219 GPUDirect Storage-supported filesystem. This engine performs
2220 I/O without transferring buffers between user-space and the kernel,
2221 unless :option:`verify` is set or :option:`cuda_io` is `posix`.
2222 :option:`iomem` must not be `cudamalloc`. This ioengine defines
2223 engine specific options.
2226 I/O engine supporting asynchronous read and write operations to the
2227 DAOS File System (DFS) via libdfs.
2230 I/O engine supporting asynchronous read and write operations to
2231 NFS filesystems from userspace via libnfs. This is useful for
2232 achieving higher concurrency and thus throughput than is possible
2236 Execute 3rd party tools. Could be used to perform monitoring during jobs runtime.
2239 I/O engine using the xNVMe C API, for NVMe devices. The xnvme engine provides
2240 flexibility to access GNU/Linux Kernel NVMe driver via libaio, IOCTLs, io_uring,
2241 the SPDK NVMe driver, or your own custom NVMe driver. The xnvme engine includes
2242 engine specific options. (See https://xnvme.io).
2245 Use the libblkio library
2246 (https://gitlab.com/libblkio/libblkio). The specific
2247 *driver* to use must be set using
2248 :option:`libblkio_driver`. If
2249 :option:`mem`/:option:`iomem` is not specified, memory
2250 allocation is delegated to libblkio (and so is
2251 guaranteed to work with the selected *driver*). One
2252 libblkio instance is used per process, so all jobs
2253 setting option :option:`thread` will share a single
2254 instance (with one queue per thread) and must specify
2255 compatible options. Note that some drivers don't allow
2256 several instances to access the same device or file
2257 simultaneously, but allow it for threads.
2259 I/O engine specific parameters
2260 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2262 In addition, there are some parameters which are only valid when a specific
2263 :option:`ioengine` is in use. These are used identically to normal parameters,
2264 with the caveat that when used on the command line, they must come after the
2265 :option:`ioengine` that defines them is selected.
2267 .. option:: cmdprio_percentage=int[,int] : [io_uring] [libaio]
2269 Set the percentage of I/O that will be issued with the highest priority.
2270 Default: 0. A single value applies to reads and writes. Comma-separated
2271 values may be specified for reads and writes. For this option to be
2272 effective, NCQ priority must be supported and enabled, and the :option:`direct`
2273 option must be set. fio must also be run as the root user. Unlike
2274 slat/clat/lat stats, which can be tracked and reported independently, per
2275 priority stats only track and report a single type of latency. By default,
2276 completion latency (clat) will be reported, if :option:`lat_percentiles` is
2277 set, total latency (lat) will be reported.
2279 .. option:: cmdprio_class=int[,int] : [io_uring] [libaio]
2281 Set the I/O priority class to use for I/Os that must be issued with
2282 a priority when :option:`cmdprio_percentage` or
2283 :option:`cmdprio_bssplit` is set. If not specified when
2284 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2285 this defaults to the highest priority class. A single value applies
2286 to reads and writes. Comma-separated values may be specified for
2287 reads and writes. See :manpage:`ionice(1)`. See also the
2288 :option:`prioclass` option.
2290 .. option:: cmdprio=int[,int] : [io_uring] [libaio]
2292 Set the I/O priority value to use for I/Os that must be issued with
2293 a priority when :option:`cmdprio_percentage` or
2294 :option:`cmdprio_bssplit` is set. If not specified when
2295 :option:`cmdprio_percentage` or :option:`cmdprio_bssplit` is set,
2297 Linux limits us to a positive value between 0 and 7, with 0 being the
2298 highest. A single value applies to reads and writes. Comma-separated
2299 values may be specified for reads and writes. See :manpage:`ionice(1)`.
2300 Refer to an appropriate manpage for other operating systems since
2301 meaning of priority may differ. See also the :option:`prio` option.
2303 .. option:: cmdprio_bssplit=str[,str] : [io_uring] [libaio]
2305 To get a finer control over I/O priority, this option allows
2306 specifying the percentage of IOs that must have a priority set
2307 depending on the block size of the IO. This option is useful only
2308 when used together with the :option:`bssplit` option, that is,
2309 multiple different block sizes are used for reads and writes.
2311 The first accepted format for this option is the same as the format of
2312 the :option:`bssplit` option:
2314 cmdprio_bssplit=blocksize/percentage:blocksize/percentage
2316 In this case, each entry will use the priority class and priority
2317 level defined by the options :option:`cmdprio_class` and
2318 :option:`cmdprio` respectively.
2320 The second accepted format for this option is:
2322 cmdprio_bssplit=blocksize/percentage/class/level:blocksize/percentage/class/level
2324 In this case, the priority class and priority level is defined inside
2325 each entry. In comparison with the first accepted format, the second
2326 accepted format does not restrict all entries to have the same priority
2327 class and priority level.
2329 For both formats, only the read and write data directions are supported,
2330 values for trim IOs are ignored. This option is mutually exclusive with
2331 the :option:`cmdprio_percentage` option.
2333 .. option:: fixedbufs : [io_uring] [io_uring_cmd]
2335 If fio is asked to do direct IO, then Linux will map pages for each
2336 IO call, and release them when IO is done. If this option is set, the
2337 pages are pre-mapped before IO is started. This eliminates the need to
2338 map and release for each IO. This is more efficient, and reduces the
2341 .. option:: nonvectored=int : [io_uring] [io_uring_cmd]
2343 With this option, fio will use non-vectored read/write commands, where
2344 address must contain the address directly. Default is -1.
2346 .. option:: force_async=int : [io_uring] [io_uring_cmd]
2348 Normal operation for io_uring is to try and issue an sqe as
2349 non-blocking first, and if that fails, execute it in an async manner.
2350 With this option set to N, then every N request fio will ask sqe to
2351 be issued in an async manner. Default is 0.
2353 .. option:: registerfiles : [io_uring] [io_uring_cmd]
2355 With this option, fio registers the set of files being used with the
2356 kernel. This avoids the overhead of managing file counts in the kernel,
2357 making the submission and completion part more lightweight. Required
2358 for the below :option:`sqthread_poll` option.
2360 .. option:: sqthread_poll : [io_uring] [io_uring_cmd] [xnvme]
2362 Normally fio will submit IO by issuing a system call to notify the
2363 kernel of available items in the SQ ring. If this option is set, the
2364 act of submitting IO will be done by a polling thread in the kernel.
2365 This frees up cycles for fio, at the cost of using more CPU in the
2366 system. As submission is just the time it takes to fill in the sqe
2367 entries and any syscall required to wake up the idle kernel thread,
2368 fio will not report submission latencies.
2370 .. option:: sqthread_poll_cpu=int : [io_uring] [io_uring_cmd]
2372 When :option:`sqthread_poll` is set, this option provides a way to
2373 define which CPU should be used for the polling thread.
2375 .. option:: cmd_type=str : [io_uring_cmd]
2377 Specifies the type of uring passthrough command to be used. Supported
2378 value is nvme. Default is nvme.
2382 [io_uring] [io_uring_cmd] [xnvme]
2384 If this option is set, fio will attempt to use polled IO completions.
2385 Normal IO completions generate interrupts to signal the completion of
2386 IO, polled completions do not. Hence they are require active reaping
2387 by the application. The benefits are more efficient IO for high IOPS
2388 scenarios, and lower latencies for low queue depth IO.
2392 Use poll queues. This is incompatible with
2393 :option:`libblkio_wait_mode=eventfd <libblkio_wait_mode>` and
2394 :option:`libblkio_force_enable_completion_eventfd`.
2398 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2403 If this option is set, fio will attempt to use polled IO completions.
2404 This will have a similar effect as (io_uring)hipri. Only SCSI READ and
2405 WRITE commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor
2406 VERIFY). Older versions of the Linux sg driver that do not support
2407 hipri will simply ignore this flag and do normal IO. The Linux SCSI
2408 Low Level Driver (LLD) that "owns" the device also needs to support
2409 hipri (also known as iopoll and mq_poll). The MegaRAID driver is an
2410 example of a SCSI LLD. Default: clear (0) which does normal
2411 (interrupted based) IO.
2413 .. option:: userspace_reap : [libaio]
2415 Normally, with the libaio engine in use, fio will use the
2416 :manpage:`io_getevents(2)` system call to reap newly returned events. With
2417 this flag turned on, the AIO ring will be read directly from user-space to
2418 reap events. The reaping mode is only enabled when polling for a minimum of
2419 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
2421 .. option:: hipri_percentage : [pvsync2]
2423 When hipri is set this determines the probability of a pvsync2 I/O being high
2424 priority. The default is 100%.
2426 .. option:: nowait=bool : [pvsync2] [libaio] [io_uring] [io_uring_cmd]
2428 By default if a request cannot be executed immediately (e.g. resource starvation,
2429 waiting on locks) it is queued and the initiating process will be blocked until
2430 the required resource becomes free.
2432 This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and
2433 the call will return instantly with EAGAIN or a partial result rather than waiting.
2435 It is useful to also use ignore_error=EAGAIN when using this option.
2437 Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2.
2438 They return EOPNOTSUP instead of EAGAIN.
2440 For cached I/O, using this option usually means a request operates only with
2441 cached data. Currently the RWF_NOWAIT flag does not supported for cached write.
2443 For direct I/O, requests will only succeed if cache invalidation isn't required,
2444 file blocks are fully allocated and the disk request could be issued immediately.
2446 .. option:: fdp=bool : [io_uring_cmd] [xnvme]
2448 Enable Flexible Data Placement mode for write commands.
2450 .. option:: fdp_pli_select=str : [io_uring_cmd] [xnvme]
2452 Defines how fio decides which placement ID to use next. The following
2456 Choose a placement ID at random (uniform).
2459 Round robin over available placement IDs. This is the
2462 The available placement ID index/indices is defined by the option
2465 .. option:: fdp_pli=str : [io_uring_cmd] [xnvme]
2467 Select which Placement ID Index/Indicies this job is allowed to use for
2468 writes. By default, the job will cycle through all available Placement
2469 IDs, so use this to isolate these identifiers to specific jobs. If you
2470 want fio to use placement identifier only at indices 0, 2 and 5 specify
2473 .. option:: cpuload=int : [cpuio]
2475 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2476 option when using cpuio I/O engine.
2478 .. option:: cpuchunks=int : [cpuio]
2480 Split the load into cycles of the given time. In microseconds.
2482 .. option:: cpumode=str : [cpuio]
2484 Specify how to stress the CPU. It can take these two values:
2487 This is the default where the CPU executes noop instructions.
2489 Replace the default noop instructions loop with a qsort algorithm to
2490 consume more energy.
2492 .. option:: exit_on_io_done=bool : [cpuio]
2494 Detect when I/O threads are done, then exit.
2496 .. option:: namenode=str : [libhdfs]
2498 The hostname or IP address of a HDFS cluster namenode to contact.
2500 .. option:: port=int
2504 The listening port of the HFDS cluster namenode.
2508 The TCP or UDP port to bind to or connect to. If this is used with
2509 :option:`numjobs` to spawn multiple instances of the same job type, then
2510 this will be the starting port number since fio will use a range of
2515 The port to use for RDMA-CM communication. This should be the same value
2516 on the client and the server side.
2518 .. option:: hostname=str : [netsplice] [net] [rdma]
2520 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
2521 is a TCP listener or UDP reader, the hostname is not used and must be omitted
2522 unless it is a valid UDP multicast address.
2524 .. option:: serverip=str : [librpma_*]
2526 The IP address to be used for RDMA-CM based I/O.
2528 .. option:: direct_write_to_pmem=bool : [librpma_*]
2530 Set to 1 only when Direct Write to PMem from the remote host is possible.
2531 Otherwise, set to 0.
2533 .. option:: busy_wait_polling=bool : [librpma_*_server]
2535 Set to 0 to wait for completion instead of busy-wait polling completion.
2538 .. option:: interface=str : [netsplice] [net]
2540 The IP address of the network interface used to send or receive UDP
2543 .. option:: ttl=int : [netsplice] [net]
2545 Time-to-live value for outgoing UDP multicast packets. Default: 1.
2547 .. option:: nodelay=bool : [netsplice] [net]
2549 Set TCP_NODELAY on TCP connections.
2551 .. option:: protocol=str, proto=str : [netsplice] [net]
2553 The network protocol to use. Accepted values are:
2556 Transmission control protocol.
2558 Transmission control protocol V6.
2560 User datagram protocol.
2562 User datagram protocol V6.
2566 When the protocol is TCP or UDP, the port must also be given, as well as the
2567 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
2568 normal :option:`filename` option should be used and the port is invalid.
2570 .. option:: listen : [netsplice] [net]
2572 For TCP network connections, tell fio to listen for incoming connections
2573 rather than initiating an outgoing connection. The :option:`hostname` must
2574 be omitted if this option is used.
2576 .. option:: pingpong : [netsplice] [net]
2578 Normally a network writer will just continue writing data, and a network
2579 reader will just consume packages. If ``pingpong=1`` is set, a writer will
2580 send its normal payload to the reader, then wait for the reader to send the
2581 same payload back. This allows fio to measure network latencies. The
2582 submission and completion latencies then measure local time spent sending or
2583 receiving, and the completion latency measures how long it took for the
2584 other end to receive and send back. For UDP multicast traffic
2585 ``pingpong=1`` should only be set for a single reader when multiple readers
2586 are listening to the same address.
2588 .. option:: window_size : [netsplice] [net]
2590 Set the desired socket buffer size for the connection.
2592 .. option:: mss : [netsplice] [net]
2594 Set the TCP maximum segment size (TCP_MAXSEG).
2596 .. option:: donorname=str : [e4defrag]
2598 File will be used as a block donor (swap extents between files).
2600 .. option:: inplace=int : [e4defrag]
2602 Configure donor file blocks allocation strategy:
2605 Default. Preallocate donor's file on init.
2607 Allocate space immediately inside defragment event, and free right
2610 .. option:: clustername=str : [rbd,rados]
2612 Specifies the name of the Ceph cluster.
2614 .. option:: rbdname=str : [rbd]
2616 Specifies the name of the RBD.
2618 .. option:: clientname=str : [rbd,rados]
2620 Specifies the username (without the 'client.' prefix) used to access the
2621 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2622 the full *type.id* string. If no type. prefix is given, fio will add
2623 'client.' by default.
2625 .. option:: conf=str : [rados]
2627 Specifies the configuration path of ceph cluster, so conf file does not
2628 have to be /etc/ceph/ceph.conf.
2630 .. option:: busy_poll=bool : [rbd,rados]
2632 Poll store instead of waiting for completion. Usually this provides better
2633 throughput at cost of higher(up to 100%) CPU utilization.
2635 .. option:: touch_objects=bool : [rados]
2637 During initialization, touch (create if do not exist) all objects (files).
2638 Touching all objects affects ceph caches and likely impacts test results.
2641 .. option:: pool=str :
2645 Specifies the name of the Ceph pool containing RBD or RADOS data.
2649 Specify the label or UUID of the DAOS pool to connect to.
2651 .. option:: cont=str : [dfs]
2653 Specify the label or UUID of the DAOS container to open.
2655 .. option:: chunk_size=int
2659 Specify a different chunk size (in bytes) for the dfs file.
2660 Use DAOS container's chunk size by default.
2664 The size of the chunk to use for each file.
2666 .. option:: object_class=str : [dfs]
2668 Specify a different object class for the dfs file.
2669 Use DAOS container's object class by default.
2671 .. option:: skip_bad=bool : [mtd]
2673 Skip operations against known bad blocks.
2675 .. option:: hdfsdirectory : [libhdfs]
2677 libhdfs will create chunk in this HDFS directory.
2679 .. option:: verb=str : [rdma]
2681 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2682 values are write, read, send and recv. These correspond to the equivalent
2683 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2684 specified on the client side of the connection. See the examples folder.
2686 .. option:: bindname=str : [rdma]
2688 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2689 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2690 will be passed into the rdma_bind_addr() function and on the client site it
2691 will be used in the rdma_resolve_add() function. This can be useful when
2692 multiple paths exist between the client and the server or in certain loopback
2695 .. option:: stat_type=str : [filestat]
2697 Specify stat system call type to measure lookup/getattr performance.
2698 Default is **stat** for :manpage:`stat(2)`.
2700 .. option:: readfua=bool : [sg]
2702 With readfua option set to 1, read operations include
2703 the force unit access (fua) flag. Default is 0.
2705 .. option:: writefua=bool : [sg]
2707 With writefua option set to 1, write operations include
2708 the force unit access (fua) flag. Default is 0.
2710 .. option:: sg_write_mode=str : [sg]
2712 Specify the type of write commands to issue. This option can take three values:
2715 This is the default where write opcodes are issued as usual.
2716 **write_and_verify**
2717 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2718 directs the device to carry out a medium verification with no data
2719 comparison. The writefua option is ignored with this selection.
2721 This option is deprecated. Use write_and_verify instead.
2723 Issue WRITE SAME commands. This transfers a single block to the device
2724 and writes this same block of data to a contiguous sequence of LBAs
2725 beginning at the specified offset. fio's block size parameter specifies
2726 the amount of data written with each command. However, the amount of data
2727 actually transferred to the device is equal to the device's block
2728 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2729 write 16 sectors with each command. fio will still generate 8k of data
2730 for each command but only the first 512 bytes will be used and
2731 transferred to the device. The writefua option is ignored with this
2734 This option is deprecated. Use write_same instead.
2736 Issue WRITE SAME(16) commands as above but with the No Data Output
2737 Buffer (NDOB) bit set. No data will be transferred to the device with
2738 this bit set. Data written will be a pre-determined pattern such as
2741 Issue WRITE STREAM(16) commands. Use the **stream_id** option to specify
2742 the stream identifier.
2743 **verify_bytchk_00**
2744 Issue VERIFY commands with BYTCHK set to 00. This directs the
2745 device to carry out a medium verification with no data comparison.
2746 **verify_bytchk_01**
2747 Issue VERIFY commands with BYTCHK set to 01. This directs the device to
2748 compare the data on the device with the data transferred to the device.
2749 **verify_bytchk_11**
2750 Issue VERIFY commands with BYTCHK set to 11. This transfers a
2751 single block to the device and compares the contents of this block with the
2752 data on the device beginning at the specified offset. fio's block size
2753 parameter specifies the total amount of data compared with this command.
2754 However, only one block (sector) worth of data is transferred to the device.
2755 This is similar to the WRITE SAME command except that data is compared instead
2758 .. option:: stream_id=int : [sg]
2760 Set the stream identifier for WRITE STREAM commands. If this is set to 0 (which is not
2761 a valid stream identifier) fio will open a stream and then close it when done. Default
2764 .. option:: http_host=str : [http]
2766 Hostname to connect to. For S3, this could be the bucket hostname.
2767 Default is **localhost**
2769 .. option:: http_user=str : [http]
2771 Username for HTTP authentication.
2773 .. option:: http_pass=str : [http]
2775 Password for HTTP authentication.
2777 .. option:: https=str : [http]
2779 Enable HTTPS instead of http. *on* enables HTTPS; *insecure*
2780 will enable HTTPS, but disable SSL peer verification (use with
2781 caution!). Default is **off**
2783 .. option:: http_mode=str : [http]
2785 Which HTTP access mode to use: *webdav*, *swift*, or *s3*.
2786 Default is **webdav**
2788 .. option:: http_s3_region=str : [http]
2790 The S3 region/zone string.
2791 Default is **us-east-1**
2793 .. option:: http_s3_key=str : [http]
2797 .. option:: http_s3_keyid=str : [http]
2799 The S3 key/access id.
2801 .. option:: http_s3_sse_customer_key=str : [http]
2803 The encryption customer key in SSE server side.
2805 .. option:: http_s3_sse_customer_algorithm=str : [http]
2807 The encryption customer algorithm in SSE server side.
2808 Default is **AES256**
2810 .. option:: http_s3_storage_class=str : [http]
2812 Which storage class to access. User-customizable settings.
2813 Default is **STANDARD**
2815 .. option:: http_swift_auth_token=str : [http]
2817 The Swift auth token. See the example configuration file on how
2820 .. option:: http_verbose=int : [http]
2822 Enable verbose requests from libcurl. Useful for debugging. 1
2823 turns on verbose logging from libcurl, 2 additionally enables
2824 HTTP IO tracing. Default is **0**
2826 .. option:: uri=str : [nbd]
2828 Specify the NBD URI of the server to test. The string
2829 is a standard NBD URI
2830 (see https://github.com/NetworkBlockDevice/nbd/tree/master/doc).
2831 Example URIs: nbd://localhost:10809
2832 nbd+unix:///?socket=/tmp/socket
2833 nbds://tlshost/exportname
2835 .. option:: gpu_dev_ids=str : [libcufile]
2837 Specify the GPU IDs to use with CUDA. This is a colon-separated list of
2838 int. GPUs are assigned to workers roundrobin. Default is 0.
2840 .. option:: cuda_io=str : [libcufile]
2842 Specify the type of I/O to use with CUDA. Default is **cufile**.
2845 Use libcufile and nvidia-fs. This option performs I/O directly
2846 between a GPUDirect Storage filesystem and GPU buffers,
2847 avoiding use of a bounce buffer. If :option:`verify` is set,
2848 cudaMemcpy is used to copy verificaton data between RAM and GPU.
2849 Verification data is copied from RAM to GPU before a write
2850 and from GPU to RAM after a read. :option:`direct` must be 1.
2852 Use POSIX to perform I/O with a RAM buffer, and use cudaMemcpy
2853 to transfer data between RAM and the GPUs. Data is copied from
2854 GPU to RAM before a write and copied from RAM to GPU after a
2855 read. :option:`verify` does not affect use of cudaMemcpy.
2857 .. option:: nfs_url=str : [nfs]
2859 URL in libnfs format, eg nfs://<server|ipv4|ipv6>/path[?arg=val[&arg=val]*]
2860 Refer to the libnfs README for more details.
2862 .. option:: program=str : [exec]
2864 Specify the program to execute.
2866 .. option:: arguments=str : [exec]
2868 Specify arguments to pass to program.
2869 Some special variables can be expanded to pass fio's job details to the program.
2872 Replaced by the duration of the job in seconds.
2874 Replaced by the name of the job.
2876 .. option:: grace_time=int : [exec]
2878 Specify the time between the SIGTERM and SIGKILL signals. Default is 1 second.
2880 .. option:: std_redirect=bool : [exec]
2882 If set, stdout and stderr streams are redirected to files named from the job name. Default is true.
2884 .. option:: xnvme_async=str : [xnvme]
2886 Select the xnvme async command interface. This can take these values.
2889 This is default and use to emulate asynchronous I/O by using a
2890 single thread to create a queue pair on top of a synchronous
2891 I/O interface using the NVMe driver IOCTL.
2893 Emulate an asynchronous I/O interface with a pool of userspace
2894 threads on top of a synchronous I/O interface using the NVMe
2895 driver IOCTL. By default four threads are used.
2897 Linux native asynchronous I/O interface which supports both
2898 direct and buffered I/O.
2900 Fast Linux native asynchronous I/O interface for NVMe pass
2901 through commands. This only works with NVMe character device
2904 Use Linux aio for Asynchronous I/O.
2906 Use the posix asynchronous I/O interface to perform one or
2907 more I/O operations asynchronously.
2909 Use the user-space VFIO-based backend, implemented using
2910 libvfn instead of SPDK.
2912 Do not transfer any data; just pretend to. This is mainly used
2913 for introspective performance evaluation.
2915 .. option:: xnvme_sync=str : [xnvme]
2917 Select the xnvme synchronous command interface. This can take these values.
2920 This is default and uses Linux NVMe Driver ioctl() for
2923 This supports regular as well as vectored pread() and pwrite()
2926 This is the same as psync except that it also supports zone
2927 management commands using Linux block layer IOCTLs.
2929 .. option:: xnvme_admin=str : [xnvme]
2931 Select the xnvme admin command interface. This can take these values.
2934 This is default and uses linux NVMe Driver ioctl() for admin
2937 Use Linux Block Layer ioctl() and sysfs for admin commands.
2939 .. option:: xnvme_dev_nsid=int : [xnvme]
2941 xnvme namespace identifier for userspace NVMe driver, SPDK or vfio.
2943 .. option:: xnvme_dev_subnqn=str : [xnvme]
2945 Sets the subsystem NQN for fabrics. This is for xNVMe to utilize a
2946 fabrics target with multiple systems.
2948 .. option:: xnvme_mem=str : [xnvme]
2950 Select the xnvme memory backend. This can take these values.
2953 This is the default posix memory backend for linux NVMe driver.
2955 Use hugepages, instead of existing posix memory backend. The
2956 memory backend uses hugetlbfs. This require users to allocate
2957 hugepages, mount hugetlbfs and set an enviornment variable for
2960 Uses SPDK's memory allocator.
2962 Uses libvfn's memory allocator. This also specifies the use
2963 of libvfn backend instead of SPDK.
2965 .. option:: xnvme_iovec=int : [xnvme]
2967 If this option is set. xnvme will use vectored read/write commands.
2969 .. option:: libblkio_driver=str : [libblkio]
2971 The libblkio *driver* to use. Different drivers access devices through
2972 different underlying interfaces. Available drivers depend on the
2973 libblkio version in use and are listed at
2974 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
2976 .. option:: libblkio_path=str : [libblkio]
2978 Sets the value of the driver-specific "path" property before connecting
2979 the libblkio instance, which identifies the target device or file on
2980 which to perform I/O. Its exact semantics are driver-dependent and not
2981 all drivers may support it; see
2982 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
2984 .. option:: libblkio_pre_connect_props=str : [libblkio]
2986 A colon-separated list of additional libblkio properties to be set after
2987 creating but before connecting the libblkio instance. Each property must
2988 have the format ``<name>=<value>``. Colons can be escaped as ``\:``.
2989 These are set after the engine sets any other properties, so those can
2990 be overriden. Available properties depend on the libblkio version in use
2992 https://libblkio.gitlab.io/libblkio/blkio.html#properties
2994 .. option:: libblkio_num_entries=int : [libblkio]
2996 Sets the value of the driver-specific "num-entries" property before
2997 starting the libblkio instance. Its exact semantics are driver-dependent
2998 and not all drivers may support it; see
2999 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3001 .. option:: libblkio_queue_size=int : [libblkio]
3003 Sets the value of the driver-specific "queue-size" property before
3004 starting the libblkio instance. Its exact semantics are driver-dependent
3005 and not all drivers may support it; see
3006 https://libblkio.gitlab.io/libblkio/blkio.html#drivers
3008 .. option:: libblkio_pre_start_props=str : [libblkio]
3010 A colon-separated list of additional libblkio properties to be set after
3011 connecting but before starting the libblkio instance. Each property must
3012 have the format ``<name>=<value>``. Colons can be escaped as ``\:``.
3013 These are set after the engine sets any other properties, so those can
3014 be overriden. Available properties depend on the libblkio version in use
3016 https://libblkio.gitlab.io/libblkio/blkio.html#properties
3018 .. option:: libblkio_vectored : [libblkio]
3020 Submit vectored read and write requests.
3022 .. option:: libblkio_write_zeroes_on_trim : [libblkio]
3024 Submit trims as "write zeroes" requests instead of discard requests.
3026 .. option:: libblkio_wait_mode=str : [libblkio]
3028 How to wait for completions:
3031 Use a blocking call to ``blkioq_do_io()``.
3033 Use a blocking call to ``read()`` on the completion eventfd.
3035 Use a busy loop with a non-blocking call to ``blkioq_do_io()``.
3037 .. option:: libblkio_force_enable_completion_eventfd : [libblkio]
3039 Enable the queue's completion eventfd even when unused. This may impact
3040 performance. The default is to enable it only if
3041 :option:`libblkio_wait_mode=eventfd <libblkio_wait_mode>`.
3043 .. option:: no_completion_thread : [windowsaio]
3045 Avoid using a separate thread for completion polling.
3050 .. option:: iodepth=int
3052 Number of I/O units to keep in flight against the file. Note that
3053 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
3054 for small degrees when :option:`verify_async` is in use). Even async
3055 engines may impose OS restrictions causing the desired depth not to be
3056 achieved. This may happen on Linux when using libaio and not setting
3057 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
3058 eye on the I/O depth distribution in the fio output to verify that the
3059 achieved depth is as expected. Default: 1.
3061 .. option:: iodepth_batch_submit=int, iodepth_batch=int
3063 This defines how many pieces of I/O to submit at once. It defaults to 1
3064 which means that we submit each I/O as soon as it is available, but can be
3065 raised to submit bigger batches of I/O at the time. If it is set to 0 the
3066 :option:`iodepth` value will be used.
3068 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
3070 This defines how many pieces of I/O to retrieve at once. It defaults to 1
3071 which means that we'll ask for a minimum of 1 I/O in the retrieval process
3072 from the kernel. The I/O retrieval will go on until we hit the limit set by
3073 :option:`iodepth_low`. If this variable is set to 0, then fio will always
3074 check for completed events before queuing more I/O. This helps reduce I/O
3075 latency, at the cost of more retrieval system calls.
3077 .. option:: iodepth_batch_complete_max=int
3079 This defines maximum pieces of I/O to retrieve at once. This variable should
3080 be used along with :option:`iodepth_batch_complete_min`\=int variable,
3081 specifying the range of min and max amount of I/O which should be
3082 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
3087 iodepth_batch_complete_min=1
3088 iodepth_batch_complete_max=<iodepth>
3090 which means that we will retrieve at least 1 I/O and up to the whole
3091 submitted queue depth. If none of I/O has been completed yet, we will wait.
3095 iodepth_batch_complete_min=0
3096 iodepth_batch_complete_max=<iodepth>
3098 which means that we can retrieve up to the whole submitted queue depth, but
3099 if none of I/O has been completed yet, we will NOT wait and immediately exit
3100 the system call. In this example we simply do polling.
3102 .. option:: iodepth_low=int
3104 The low water mark indicating when to start filling the queue
3105 again. Defaults to the same as :option:`iodepth`, meaning that fio will
3106 attempt to keep the queue full at all times. If :option:`iodepth` is set to
3107 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
3108 16 requests, it will let the depth drain down to 4 before starting to fill
3111 .. option:: serialize_overlap=bool
3113 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
3114 When two or more I/Os are submitted simultaneously, there is no guarantee that
3115 the I/Os will be processed or completed in the submitted order. Further, if
3116 two or more of those I/Os are writes, any overlapping region between them can
3117 become indeterminate/undefined on certain storage. These issues can cause
3118 verification to fail erratically when at least one of the racing I/Os is
3119 changing data and the overlapping region has a non-zero size. Setting
3120 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
3121 serializing in-flight I/Os that have a non-zero overlap. Note that setting
3122 this option can reduce both performance and the :option:`iodepth` achieved.
3124 This option only applies to I/Os issued for a single job except when it is
3125 enabled along with :option:`io_submit_mode`\=offload. In offload mode, fio
3126 will check for overlap among all I/Os submitted by offload jobs with :option:`serialize_overlap`
3131 .. option:: io_submit_mode=str
3133 This option controls how fio submits the I/O to the I/O engine. The default
3134 is `inline`, which means that the fio job threads submit and reap I/O
3135 directly. If set to `offload`, the job threads will offload I/O submission
3136 to a dedicated pool of I/O threads. This requires some coordination and thus
3137 has a bit of extra overhead, especially for lower queue depth I/O where it
3138 can increase latencies. The benefit is that fio can manage submission rates
3139 independently of the device completion rates. This avoids skewed latency
3140 reporting if I/O gets backed up on the device side (the coordinated omission
3141 problem). Note that this option cannot reliably be used with async IO
3148 .. option:: thinktime=time
3150 Stall the job for the specified period of time after an I/O has completed before issuing the
3151 next. May be used to simulate processing being done by an application.
3152 When the unit is omitted, the value is interpreted in microseconds. See
3153 :option:`thinktime_blocks`, :option:`thinktime_iotime` and :option:`thinktime_spin`.
3155 .. option:: thinktime_spin=time
3157 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
3158 something with the data received, before falling back to sleeping for the
3159 rest of the period specified by :option:`thinktime`. When the unit is
3160 omitted, the value is interpreted in microseconds.
3162 .. option:: thinktime_blocks=int
3164 Only valid if :option:`thinktime` is set - control how many blocks to issue,
3165 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
3166 fio wait :option:`thinktime` usecs after every block. This effectively makes any
3167 queue depth setting redundant, since no more than 1 I/O will be queued
3168 before we have to complete it and do our :option:`thinktime`. In other words, this
3169 setting effectively caps the queue depth if the latter is larger.
3171 .. option:: thinktime_blocks_type=str
3173 Only valid if :option:`thinktime` is set - control how :option:`thinktime_blocks`
3174 triggers. The default is `complete`, which triggers thinktime when fio completes
3175 :option:`thinktime_blocks` blocks. If this is set to `issue`, then the trigger happens
3178 .. option:: thinktime_iotime=time
3180 Only valid if :option:`thinktime` is set - control :option:`thinktime`
3181 interval by time. The :option:`thinktime` stall is repeated after IOs
3182 are executed for :option:`thinktime_iotime`. For example,
3183 ``--thinktime_iotime=9s --thinktime=1s`` repeat 10-second cycle with IOs
3184 for 9 seconds and stall for 1 second. When the unit is omitted,
3185 :option:`thinktime_iotime` is interpreted as a number of seconds. If
3186 this option is used together with :option:`thinktime_blocks`, the
3187 :option:`thinktime` stall is repeated after :option:`thinktime_iotime`
3188 or after :option:`thinktime_blocks` IOs, whichever happens first.
3190 .. option:: rate=int[,int][,int]
3192 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
3193 suffix rules apply. Comma-separated values may be specified for reads,
3194 writes, and trims as described in :option:`blocksize`.
3196 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
3197 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
3198 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
3199 latter will only limit reads.
3201 .. option:: rate_min=int[,int][,int]
3203 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
3204 to meet this requirement will cause the job to exit. Comma-separated values
3205 may be specified for reads, writes, and trims as described in
3206 :option:`blocksize`.
3208 .. option:: rate_iops=int[,int][,int]
3210 Cap the bandwidth to this number of IOPS. Basically the same as
3211 :option:`rate`, just specified independently of bandwidth. If the job is
3212 given a block size range instead of a fixed value, the smallest block size
3213 is used as the metric. Comma-separated values may be specified for reads,
3214 writes, and trims as described in :option:`blocksize`.
3216 .. option:: rate_iops_min=int[,int][,int]
3218 If fio doesn't meet this rate of I/O, it will cause the job to exit.
3219 Comma-separated values may be specified for reads, writes, and trims as
3220 described in :option:`blocksize`.
3222 .. option:: rate_process=str
3224 This option controls how fio manages rated I/O submissions. The default is
3225 `linear`, which submits I/O in a linear fashion with fixed delays between
3226 I/Os that gets adjusted based on I/O completion rates. If this is set to
3227 `poisson`, fio will submit I/O based on a more real world random request
3228 flow, known as the Poisson process
3229 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
3230 10^6 / IOPS for the given workload.
3232 .. option:: rate_ignore_thinktime=bool
3234 By default, fio will attempt to catch up to the specified rate setting,
3235 if any kind of thinktime setting was used. If this option is set, then
3236 fio will ignore the thinktime and continue doing IO at the specified
3237 rate, instead of entering a catch-up mode after thinktime is done.
3239 .. option:: rate_cycle=int
3241 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
3242 of milliseconds. Defaults to 1000.
3248 .. option:: latency_target=time
3250 If set, fio will attempt to find the max performance point that the given
3251 workload will run at while maintaining a latency below this target. When
3252 the unit is omitted, the value is interpreted in microseconds. See
3253 :option:`latency_window` and :option:`latency_percentile`.
3255 .. option:: latency_window=time
3257 Used with :option:`latency_target` to specify the sample window that the job
3258 is run at varying queue depths to test the performance. When the unit is
3259 omitted, the value is interpreted in microseconds.
3261 .. option:: latency_percentile=float
3263 The percentage of I/Os that must fall within the criteria specified by
3264 :option:`latency_target` and :option:`latency_window`. If not set, this
3265 defaults to 100.0, meaning that all I/Os must be equal or below to the value
3266 set by :option:`latency_target`.
3268 .. option:: latency_run=bool
3270 Used with :option:`latency_target`. If false (default), fio will find
3271 the highest queue depth that meets :option:`latency_target` and exit. If
3272 true, fio will continue running and try to meet :option:`latency_target`
3273 by adjusting queue depth.
3275 .. option:: max_latency=time[,time][,time]
3277 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
3278 maximum latency. When the unit is omitted, the value is interpreted in
3279 microseconds. Comma-separated values may be specified for reads, writes,
3280 and trims as described in :option:`blocksize`.
3286 .. option:: write_iolog=str
3288 Write the issued I/O patterns to the specified file. See
3289 :option:`read_iolog`. Specify a separate file for each job, otherwise the
3290 iologs will be interspersed and the file may be corrupt. This file will
3291 be opened in append mode.
3293 .. option:: read_iolog=str
3295 Open an iolog with the specified filename and replay the I/O patterns it
3296 contains. This can be used to store a workload and replay it sometime
3297 later. The iolog given may also be a blktrace binary file, which allows fio
3298 to replay a workload captured by :command:`blktrace`. See
3299 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
3300 replay, the file needs to be turned into a blkparse binary data file first
3301 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
3302 You can specify a number of files by separating the names with a ':'
3303 character. See the :option:`filename` option for information on how to
3304 escape ':' characters within the file names. These files will
3305 be sequentially assigned to job clones created by :option:`numjobs`.
3306 '-' is a reserved name, meaning read from stdin, notably if
3307 :option:`filename` is set to '-' which means stdin as well, then
3308 this flag can't be set to '-'.
3310 .. option:: read_iolog_chunked=bool
3312 Determines how iolog is read. If false(default) entire :option:`read_iolog`
3313 will be read at once. If selected true, input from iolog will be read
3314 gradually. Useful when iolog is very large, or it is generated.
3316 .. option:: merge_blktrace_file=str
3318 When specified, rather than replaying the logs passed to :option:`read_iolog`,
3319 the logs go through a merge phase which aggregates them into a single
3320 blktrace. The resulting file is then passed on as the :option:`read_iolog`
3321 parameter. The intention here is to make the order of events consistent.
3322 This limits the influence of the scheduler compared to replaying multiple
3323 blktraces via concurrent jobs.
3325 .. option:: merge_blktrace_scalars=float_list
3327 This is a percentage based option that is index paired with the list of
3328 files passed to :option:`read_iolog`. When merging is performed, scale
3329 the time of each event by the corresponding amount. For example,
3330 ``--merge_blktrace_scalars="50:100"`` runs the first trace in halftime
3331 and the second trace in realtime. This knob is separately tunable from
3332 :option:`replay_time_scale` which scales the trace during runtime and
3333 does not change the output of the merge unlike this option.
3335 .. option:: merge_blktrace_iters=float_list
3337 This is a whole number option that is index paired with the list of files
3338 passed to :option:`read_iolog`. When merging is performed, run each trace
3339 for the specified number of iterations. For example,
3340 ``--merge_blktrace_iters="2:1"`` runs the first trace for two iterations
3341 and the second trace for one iteration.
3343 .. option:: replay_no_stall=bool
3345 When replaying I/O with :option:`read_iolog` the default behavior is to
3346 attempt to respect the timestamps within the log and replay them with the
3347 appropriate delay between IOPS. By setting this variable fio will not
3348 respect the timestamps and attempt to replay them as fast as possible while
3349 still respecting ordering. The result is the same I/O pattern to a given
3350 device, but different timings.
3352 .. option:: replay_time_scale=int
3354 When replaying I/O with :option:`read_iolog`, fio will honor the
3355 original timing in the trace. With this option, it's possible to scale
3356 the time. It's a percentage option, if set to 50 it means run at 50%
3357 the original IO rate in the trace. If set to 200, run at twice the
3358 original IO rate. Defaults to 100.
3360 .. option:: replay_redirect=str
3362 While replaying I/O patterns using :option:`read_iolog` the default behavior
3363 is to replay the IOPS onto the major/minor device that each IOP was recorded
3364 from. This is sometimes undesirable because on a different machine those
3365 major/minor numbers can map to a different device. Changing hardware on the
3366 same system can also result in a different major/minor mapping.
3367 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
3368 device regardless of the device it was recorded
3369 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
3370 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
3371 multiple devices will be replayed onto a single device, if the trace
3372 contains multiple devices. If you want multiple devices to be replayed
3373 concurrently to multiple redirected devices you must blkparse your trace
3374 into separate traces and replay them with independent fio invocations.
3375 Unfortunately this also breaks the strict time ordering between multiple
3378 .. option:: replay_align=int
3380 Force alignment of the byte offsets in a trace to this value. The value
3381 must be a power of 2.
3383 .. option:: replay_scale=int
3385 Scale byte offsets down by this factor when replaying traces. Should most
3386 likely use :option:`replay_align` as well.
3388 .. option:: replay_skip=str
3390 Sometimes it's useful to skip certain IO types in a replay trace.
3391 This could be, for instance, eliminating the writes in the trace.
3392 Or not replaying the trims/discards, if you are redirecting to
3393 a device that doesn't support them. This option takes a comma
3394 separated list of read, write, trim, sync.
3397 Threads, processes and job synchronization
3398 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3402 Fio defaults to creating jobs by using fork, however if this option is
3403 given, fio will create jobs by using POSIX Threads' function
3404 :manpage:`pthread_create(3)` to create threads instead.
3406 .. option:: wait_for=str
3408 If set, the current job won't be started until all workers of the specified
3409 waitee job are done.
3411 ``wait_for`` operates on the job name basis, so there are a few
3412 limitations. First, the waitee must be defined prior to the waiter job
3413 (meaning no forward references). Second, if a job is being referenced as a
3414 waitee, it must have a unique name (no duplicate waitees).
3416 .. option:: nice=int
3418 Run the job with the given nice value. See man :manpage:`nice(2)`.
3420 On Windows, values less than -15 set the process class to "High"; -1 through
3421 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
3424 .. option:: prio=int
3426 Set the I/O priority value of this job. Linux limits us to a positive value
3427 between 0 and 7, with 0 being the highest. See man
3428 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
3429 systems since meaning of priority may differ. For per-command priority
3430 setting, see I/O engine specific :option:`cmdprio_percentage` and
3431 :option:`cmdprio` options.
3433 .. option:: prioclass=int
3435 Set the I/O priority class. See man :manpage:`ionice(1)`. For per-command
3436 priority setting, see I/O engine specific :option:`cmdprio_percentage`
3437 and :option:`cmdprio_class` options.
3439 .. option:: cpus_allowed=str
3441 Controls the same options as :option:`cpumask`, but accepts a textual
3442 specification of the permitted CPUs instead and CPUs are indexed from 0. So
3443 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
3444 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
3445 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
3447 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
3448 processor group will be used and affinity settings are inherited from the
3449 system. An fio build configured to target Windows 7 makes options that set
3450 CPUs processor group aware and values will set both the processor group
3451 and a CPU from within that group. For example, on a system where processor
3452 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
3453 values between 0 and 39 will bind CPUs from processor group 0 and
3454 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
3455 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
3456 single ``cpus_allowed`` option must be from the same processor group. For
3457 Windows fio builds not built for Windows 7, CPUs will only be selected from
3458 (and be relative to) whatever processor group fio happens to be running in
3459 and CPUs from other processor groups cannot be used.
3461 .. option:: cpus_allowed_policy=str
3463 Set the policy of how fio distributes the CPUs specified by
3464 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
3467 All jobs will share the CPU set specified.
3469 Each job will get a unique CPU from the CPU set.
3471 **shared** is the default behavior, if the option isn't specified. If
3472 **split** is specified, then fio will assign one cpu per job. If not
3473 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
3476 .. option:: cpumask=int
3478 Set the CPU affinity of this job. The parameter given is a bit mask of
3479 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
3480 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
3481 :manpage:`sched_setaffinity(2)`. This may not work on all supported
3482 operating systems or kernel versions. This option doesn't work well for a
3483 higher CPU count than what you can store in an integer mask, so it can only
3484 control cpus 1-32. For boxes with larger CPU counts, use
3485 :option:`cpus_allowed`.
3487 .. option:: numa_cpu_nodes=str
3489 Set this job running on specified NUMA nodes' CPUs. The arguments allow
3490 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
3491 NUMA options support, fio must be built on a system with libnuma-dev(el)
3494 .. option:: numa_mem_policy=str
3496 Set this job's memory policy and corresponding NUMA nodes. Format of the
3501 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
3502 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
3503 policies, no node needs to be specified. For ``prefer``, only one node is
3504 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
3505 follows: a comma delimited list of numbers, A-B ranges, or `all`.
3507 .. option:: cgroup=str
3509 Add job to this control group. If it doesn't exist, it will be created. The
3510 system must have a mounted cgroup blkio mount point for this to work. If
3511 your system doesn't have it mounted, you can do so with::
3513 # mount -t cgroup -o blkio none /cgroup
3515 .. option:: cgroup_weight=int
3517 Set the weight of the cgroup to this value. See the documentation that comes
3518 with the kernel, allowed values are in the range of 100..1000.
3520 .. option:: cgroup_nodelete=bool
3522 Normally fio will delete the cgroups it has created after the job
3523 completion. To override this behavior and to leave cgroups around after the
3524 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
3525 to inspect various cgroup files after job completion. Default: false.
3527 .. option:: flow_id=int
3529 The ID of the flow. If not specified, it defaults to being a global
3530 flow. See :option:`flow`.
3532 .. option:: flow=int
3534 Weight in token-based flow control. If this value is used, then fio
3535 regulates the activity between two or more jobs sharing the same
3536 flow_id. Fio attempts to keep each job activity proportional to other
3537 jobs' activities in the same flow_id group, with respect to requested
3538 weight per job. That is, if one job has `flow=3', another job has
3539 `flow=2' and another with `flow=1`, then there will be a roughly 3:2:1
3540 ratio in how much one runs vs the others.
3542 .. option:: flow_sleep=int
3544 The period of time, in microseconds, to wait after the flow counter
3545 has exceeded its proportion before retrying operations.
3547 .. option:: stonewall, wait_for_previous
3549 Wait for preceding jobs in the job file to exit, before starting this
3550 one. Can be used to insert serialization points in the job file. A stone
3551 wall also implies starting a new reporting group, see
3552 :option:`group_reporting`.
3556 By default, fio will continue running all other jobs when one job finishes.
3557 Sometimes this is not the desired action. Setting ``exitall`` will instead
3558 make fio terminate all jobs in the same group, as soon as one job of that
3561 .. option:: exit_what=str
3563 By default, fio will continue running all other jobs when one job finishes.
3564 Sometimes this is not the desired action. Setting ``exitall`` will
3565 instead make fio terminate all jobs in the same group. The option
3566 ``exit_what`` allows to control which jobs get terminated when ``exitall`` is
3567 enabled. The default is ``group`` and does not change the behaviour of
3568 ``exitall``. The setting ``all`` terminates all jobs. The setting ``stonewall``
3569 terminates all currently running jobs across all groups and continues execution
3570 with the next stonewalled group.
3572 .. option:: exec_prerun=str
3574 Before running this job, issue the command specified through
3575 :manpage:`system(3)`. Output is redirected in a file called
3576 :file:`jobname.prerun.txt`.
3578 .. option:: exec_postrun=str
3580 After the job completes, issue the command specified though
3581 :manpage:`system(3)`. Output is redirected in a file called
3582 :file:`jobname.postrun.txt`.
3586 Instead of running as the invoking user, set the user ID to this value
3587 before the thread/process does any work.
3591 Set group ID, see :option:`uid`.
3597 .. option:: verify_only
3599 Do not perform specified workload, only verify data still matches previous
3600 invocation of this workload. This option allows one to check data multiple
3601 times at a later date without overwriting it. This option makes sense only
3602 for workloads that write data, and does not support workloads with the
3603 :option:`time_based` option set.
3605 .. option:: do_verify=bool
3607 Run the verify phase after a write phase. Only valid if :option:`verify` is
3610 .. option:: verify=str
3612 If writing to a file, fio can verify the file contents after each iteration
3613 of the job. Each verification method also implies verification of special
3614 header, which is written to the beginning of each block. This header also
3615 includes meta information, like offset of the block, block number, timestamp
3616 when block was written, etc. :option:`verify` can be combined with
3617 :option:`verify_pattern` option. The allowed values are:
3620 Use an md5 sum of the data area and store it in the header of
3624 Use an experimental crc64 sum of the data area and store it in the
3625 header of each block.
3628 Use a crc32c sum of the data area and store it in the header of
3629 each block. This will automatically use hardware acceleration
3630 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
3631 fall back to software crc32c if none is found. Generally the
3632 fastest checksum fio supports when hardware accelerated.
3638 Use a crc32 sum of the data area and store it in the header of each
3642 Use a crc16 sum of the data area and store it in the header of each
3646 Use a crc7 sum of the data area and store it in the header of each
3650 Use xxhash as the checksum function. Generally the fastest software
3651 checksum that fio supports.
3654 Use sha512 as the checksum function.
3657 Use sha256 as the checksum function.
3660 Use optimized sha1 as the checksum function.
3663 Use optimized sha3-224 as the checksum function.
3666 Use optimized sha3-256 as the checksum function.
3669 Use optimized sha3-384 as the checksum function.
3672 Use optimized sha3-512 as the checksum function.
3675 This option is deprecated, since now meta information is included in
3676 generic verification header and meta verification happens by
3677 default. For detailed information see the description of the
3678 :option:`verify` setting. This option is kept because of
3679 compatibility's sake with old configurations. Do not use it.
3682 Verify a strict pattern. Normally fio includes a header with some
3683 basic information and checksumming, but if this option is set, only
3684 the specific pattern set with :option:`verify_pattern` is verified.
3687 Only pretend to verify. Useful for testing internals with
3688 :option:`ioengine`\=null, not for much else.
3690 This option can be used for repeated burn-in tests of a system to make sure
3691 that the written data is also correctly read back. If the data direction
3692 given is a read or random read, fio will assume that it should verify a
3693 previously written file. If the data direction includes any form of write,
3694 the verify will be of the newly written data.
3696 To avoid false verification errors, do not use the norandommap option when
3697 verifying data with async I/O engines and I/O depths > 1. Or use the
3698 norandommap and the lfsr random generator together to avoid writing to the
3699 same offset with multiple outstanding I/Os.
3701 .. option:: verify_offset=int
3703 Swap the verification header with data somewhere else in the block before
3704 writing. It is swapped back before verifying.
3706 .. option:: verify_interval=int
3708 Write the verification header at a finer granularity than the
3709 :option:`blocksize`. It will be written for chunks the size of
3710 ``verify_interval``. :option:`blocksize` should divide this evenly.
3712 .. option:: verify_pattern=str
3714 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
3715 filling with totally random bytes, but sometimes it's interesting to fill
3716 with a known pattern for I/O verification purposes. Depending on the width
3717 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
3718 be either a decimal or a hex number). The ``verify_pattern`` if larger than
3719 a 32-bit quantity has to be a hex number that starts with either "0x" or
3720 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
3721 format, which means that for each block offset will be written and then
3722 verified back, e.g.::
3726 Or use combination of everything::
3728 verify_pattern=0xff%o"abcd"-12
3730 .. option:: verify_fatal=bool
3732 Normally fio will keep checking the entire contents before quitting on a
3733 block verification failure. If this option is set, fio will exit the job on
3734 the first observed failure. Default: false.
3736 .. option:: verify_dump=bool
3738 If set, dump the contents of both the original data block and the data block
3739 we read off disk to files. This allows later analysis to inspect just what
3740 kind of data corruption occurred. Off by default.
3742 .. option:: verify_async=int
3744 Fio will normally verify I/O inline from the submitting thread. This option
3745 takes an integer describing how many async offload threads to create for I/O
3746 verification instead, causing fio to offload the duty of verifying I/O
3747 contents to one or more separate threads. If using this offload option, even
3748 sync I/O engines can benefit from using an :option:`iodepth` setting higher
3749 than 1, as it allows them to have I/O in flight while verifies are running.
3750 Defaults to 0 async threads, i.e. verification is not asynchronous.
3752 .. option:: verify_async_cpus=str
3754 Tell fio to set the given CPU affinity on the async I/O verification
3755 threads. See :option:`cpus_allowed` for the format used.
3757 .. option:: verify_backlog=int
3759 Fio will normally verify the written contents of a job that utilizes verify
3760 once that job has completed. In other words, everything is written then
3761 everything is read back and verified. You may want to verify continually
3762 instead for a variety of reasons. Fio stores the meta data associated with
3763 an I/O block in memory, so for large verify workloads, quite a bit of memory
3764 would be used up holding this meta data. If this option is enabled, fio will
3765 write only N blocks before verifying these blocks.
3767 .. option:: verify_backlog_batch=int
3769 Control how many blocks fio will verify if :option:`verify_backlog` is
3770 set. If not set, will default to the value of :option:`verify_backlog`
3771 (meaning the entire queue is read back and verified). If
3772 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
3773 blocks will be verified, if ``verify_backlog_batch`` is larger than
3774 :option:`verify_backlog`, some blocks will be verified more than once.
3776 .. option:: verify_state_save=bool
3778 When a job exits during the write phase of a verify workload, save its
3779 current state. This allows fio to replay up until that point, if the verify
3780 state is loaded for the verify read phase. The format of the filename is,
3783 <type>-<jobname>-<jobindex>-verify.state.
3785 <type> is "local" for a local run, "sock" for a client/server socket
3786 connection, and "ip" (192.168.0.1, for instance) for a networked
3787 client/server connection. Defaults to true.
3789 .. option:: verify_state_load=bool
3791 If a verify termination trigger was used, fio stores the current write state
3792 of each thread. This can be used at verification time so that fio knows how
3793 far it should verify. Without this information, fio will run a full
3794 verification pass, according to the settings in the job file used. Default
3797 .. option:: experimental_verify=bool
3799 Enable experimental verification. Standard verify records I/O metadata
3800 for later use during the verification phase. Experimental verify
3801 instead resets the file after the write phase and then replays I/Os for
3802 the verification phase.
3804 .. option:: trim_percentage=int
3806 Number of verify blocks to discard/trim.
3808 .. option:: trim_verify_zero=bool
3810 Verify that trim/discarded blocks are returned as zeros.
3812 .. option:: trim_backlog=int
3814 Trim after this number of blocks are written.
3816 .. option:: trim_backlog_batch=int
3818 Trim this number of I/O blocks.
3823 .. option:: steadystate=str:float, ss=str:float
3825 Define the criterion and limit for assessing steady state performance. The
3826 first parameter designates the criterion whereas the second parameter sets
3827 the threshold. When the criterion falls below the threshold for the
3828 specified duration, the job will stop. For example, `iops_slope:0.1%` will
3829 direct fio to terminate the job when the least squares regression slope
3830 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
3831 this will apply to all jobs in the group. Below is the list of available
3832 steady state assessment criteria. All assessments are carried out using only
3833 data from the rolling collection window. Threshold limits can be expressed
3834 as a fixed value or as a percentage of the mean in the collection window.
3836 When using this feature, most jobs should include the :option:`time_based`
3837 and :option:`runtime` options or the :option:`loops` option so that fio does not
3838 stop running after it has covered the full size of the specified file(s) or device(s).
3841 Collect IOPS data. Stop the job if all individual IOPS measurements
3842 are within the specified limit of the mean IOPS (e.g., ``iops:2``
3843 means that all individual IOPS values must be within 2 of the mean,
3844 whereas ``iops:0.2%`` means that all individual IOPS values must be
3845 within 0.2% of the mean IOPS to terminate the job).
3848 Collect IOPS data and calculate the least squares regression
3849 slope. Stop the job if the slope falls below the specified limit.
3852 Collect bandwidth data. Stop the job if all individual bandwidth
3853 measurements are within the specified limit of the mean bandwidth.
3856 Collect bandwidth data and calculate the least squares regression
3857 slope. Stop the job if the slope falls below the specified limit.
3859 .. option:: steadystate_duration=time, ss_dur=time
3861 A rolling window of this duration will be used to judge whether steady
3862 state has been reached. Data will be collected every
3863 :option:`ss_interval`. The default is 0 which disables steady state
3864 detection. When the unit is omitted, the value is interpreted in
3867 .. option:: steadystate_ramp_time=time, ss_ramp=time
3869 Allow the job to run for the specified duration before beginning data
3870 collection for checking the steady state job termination criterion. The
3871 default is 0. When the unit is omitted, the value is interpreted in seconds.
3873 .. option:: steadystate_check_interval=time, ss_interval=time
3875 The values during the rolling window will be collected with a period of
3876 this value. If :option:`ss_interval` is 30s and :option:`ss_dur` is
3877 300s, 10 measurements will be taken. Default is 1s but that might not
3878 converge, especially for slower devices, so set this accordingly. When
3879 the unit is omitted, the value is interpreted in seconds.
3882 Measurements and reporting
3883 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3885 .. option:: per_job_logs=bool
3887 If set, this generates bw/clat/iops log with per file private filenames. If
3888 not set, jobs with identical names will share the log filename. Default:
3891 .. option:: group_reporting
3893 It may sometimes be interesting to display statistics for groups of jobs as
3894 a whole instead of for each individual job. This is especially true if
3895 :option:`numjobs` is used; looking at individual thread/process output
3896 quickly becomes unwieldy. To see the final report per-group instead of
3897 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
3898 same reporting group, unless if separated by a :option:`stonewall`, or by
3899 using :option:`new_group`.
3901 .. option:: new_group
3903 Start a new reporting group. See: :option:`group_reporting`. If not given,
3904 all jobs in a file will be part of the same reporting group, unless
3905 separated by a :option:`stonewall`.
3907 .. option:: stats=bool
3909 By default, fio collects and shows final output results for all jobs
3910 that run. If this option is set to 0, then fio will ignore it in
3911 the final stat output.
3913 .. option:: write_bw_log=str
3915 If given, write a bandwidth log for this job. Can be used to store data of
3916 the bandwidth of the jobs in their lifetime.
3918 If no str argument is given, the default filename of
3919 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
3920 will still append the type of log. So if one specifies::
3924 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
3925 of the job (`1..N`, where `N` is the number of jobs). If
3926 :option:`per_job_logs` is false, then the filename will not include the
3929 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
3930 text files into nice graphs. See `Log File Formats`_ for how data is
3931 structured within the file.
3933 .. option:: write_lat_log=str
3935 Same as :option:`write_bw_log`, except this option creates I/O
3936 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
3937 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
3938 latency files instead. See :option:`write_bw_log` for details about
3939 the filename format and `Log File Formats`_ for how data is structured
3942 .. option:: write_hist_log=str
3944 Same as :option:`write_bw_log` but writes an I/O completion latency
3945 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
3946 file will be empty unless :option:`log_hist_msec` has also been set.
3947 See :option:`write_bw_log` for details about the filename format and
3948 `Log File Formats`_ for how data is structured within the file.
3950 .. option:: write_iops_log=str
3952 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
3953 :file:`name_iops.x.log`) instead. Because fio defaults to individual
3954 I/O logging, the value entry in the IOPS log will be 1 unless windowed
3955 logging (see :option:`log_avg_msec`) has been enabled. See
3956 :option:`write_bw_log` for details about the filename format and `Log
3957 File Formats`_ for how data is structured within the file.
3959 .. option:: log_entries=int
3961 By default, fio will log an entry in the iops, latency, or bw log for
3962 every I/O that completes. The initial number of I/O log entries is 1024.
3963 When the log entries are all used, new log entries are dynamically
3964 allocated. This dynamic log entry allocation may negatively impact
3965 time-related statistics such as I/O tail latencies (e.g. 99.9th percentile
3966 completion latency). This option allows specifying a larger initial
3967 number of log entries to avoid run-time allocations of new log entries,
3968 resulting in more precise time-related I/O statistics.
3969 Also see :option:`log_avg_msec`. Defaults to 1024.
3971 .. option:: log_avg_msec=int
3973 By default, fio will log an entry in the iops, latency, or bw log for every
3974 I/O that completes. When writing to the disk log, that can quickly grow to a
3975 very large size. Setting this option makes fio average the each log entry
3976 over the specified period of time, reducing the resolution of the log. See
3977 :option:`log_max_value` as well. Defaults to 0, logging all entries.
3978 Also see `Log File Formats`_.
3980 .. option:: log_hist_msec=int
3982 Same as :option:`log_avg_msec`, but logs entries for completion latency
3983 histograms. Computing latency percentiles from averages of intervals using
3984 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
3985 histogram entries over the specified period of time, reducing log sizes for
3986 high IOPS devices while retaining percentile accuracy. See
3987 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
3988 Defaults to 0, meaning histogram logging is disabled.
3990 .. option:: log_hist_coarseness=int
3992 Integer ranging from 0 to 6, defining the coarseness of the resolution of
3993 the histogram logs enabled with :option:`log_hist_msec`. For each increment
3994 in coarseness, fio outputs half as many bins. Defaults to 0, for which
3995 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
3996 and `Log File Formats`_.
3998 .. option:: log_max_value=bool
4000 If :option:`log_avg_msec` is set, fio logs the average over that window. If
4001 you instead want to log the maximum value, set this option to 1. Defaults to
4002 0, meaning that averaged values are logged.
4004 .. option:: log_offset=bool
4006 If this is set, the iolog options will include the byte offset for the I/O
4007 entry as well as the other data values. Defaults to 0 meaning that
4008 offsets are not present in logs. Also see `Log File Formats`_.
4010 .. option:: log_compression=int
4012 If this is set, fio will compress the I/O logs as it goes, to keep the
4013 memory footprint lower. When a log reaches the specified size, that chunk is
4014 removed and compressed in the background. Given that I/O logs are fairly
4015 highly compressible, this yields a nice memory savings for longer runs. The
4016 downside is that the compression will consume some background CPU cycles, so
4017 it may impact the run. This, however, is also true if the logging ends up
4018 consuming most of the system memory. So pick your poison. The I/O logs are
4019 saved normally at the end of a run, by decompressing the chunks and storing
4020 them in the specified log file. This feature depends on the availability of
4023 .. option:: log_compression_cpus=str
4025 Define the set of CPUs that are allowed to handle online log compression for
4026 the I/O jobs. This can provide better isolation between performance
4027 sensitive jobs, and background compression work. See
4028 :option:`cpus_allowed` for the format used.
4030 .. option:: log_store_compressed=bool
4032 If set, fio will store the log files in a compressed format. They can be
4033 decompressed with fio, using the :option:`--inflate-log` command line
4034 parameter. The files will be stored with a :file:`.fz` suffix.
4036 .. option:: log_unix_epoch=bool
4038 If set, fio will log Unix timestamps to the log files produced by enabling
4039 write_type_log for each log type, instead of the default zero-based
4042 .. option:: log_alternate_epoch=bool
4044 If set, fio will log timestamps based on the epoch used by the clock specified
4045 in the log_alternate_epoch_clock_id option, to the log files produced by
4046 enabling write_type_log for each log type, instead of the default zero-based
4049 .. option:: log_alternate_epoch_clock_id=int
4051 Specifies the clock_id to be used by clock_gettime to obtain the alternate epoch
4052 if either log_unix_epoch or log_alternate_epoch are true. Otherwise has no
4053 effect. Default value is 0, or CLOCK_REALTIME.
4055 .. option:: block_error_percentiles=bool
4057 If set, record errors in trim block-sized units from writes and trims and
4058 output a histogram of how many trims it took to get to errors, and what kind
4059 of error was encountered.
4061 .. option:: bwavgtime=int
4063 Average the calculated bandwidth over the given time. Value is specified in
4064 milliseconds. If the job also does bandwidth logging through
4065 :option:`write_bw_log`, then the minimum of this option and
4066 :option:`log_avg_msec` will be used. Default: 500ms.
4068 .. option:: iopsavgtime=int
4070 Average the calculated IOPS over the given time. Value is specified in
4071 milliseconds. If the job also does IOPS logging through
4072 :option:`write_iops_log`, then the minimum of this option and
4073 :option:`log_avg_msec` will be used. Default: 500ms.
4075 .. option:: disk_util=bool
4077 Generate disk utilization statistics, if the platform supports it.
4080 .. option:: disable_lat=bool
4082 Disable measurements of total latency numbers. Useful only for cutting back
4083 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
4084 performance at really high IOPS rates. Note that to really get rid of a
4085 large amount of these calls, this option must be used with
4086 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
4088 .. option:: disable_clat=bool
4090 Disable measurements of completion latency numbers. See
4091 :option:`disable_lat`.
4093 .. option:: disable_slat=bool
4095 Disable measurements of submission latency numbers. See
4096 :option:`disable_lat`.
4098 .. option:: disable_bw_measurement=bool, disable_bw=bool
4100 Disable measurements of throughput/bandwidth numbers. See
4101 :option:`disable_lat`.
4103 .. option:: slat_percentiles=bool
4105 Report submission latency percentiles. Submission latency is not recorded
4106 for synchronous ioengines.
4108 .. option:: clat_percentiles=bool
4110 Report completion latency percentiles.
4112 .. option:: lat_percentiles=bool
4114 Report total latency percentiles. Total latency is the sum of submission
4115 latency and completion latency.
4117 .. option:: percentile_list=float_list
4119 Overwrite the default list of percentiles for latencies and the block error
4120 histogram. Each number is a floating point number in the range (0,100], and
4121 the maximum length of the list is 20. Use ``:`` to separate the numbers. For
4122 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
4123 latency durations below which 99.5% and 99.9% of the observed latencies fell,
4126 .. option:: significant_figures=int
4128 If using :option:`--output-format` of `normal`, set the significant
4129 figures to this value. Higher values will yield more precise IOPS and
4130 throughput units, while lower values will round. Requires a minimum
4131 value of 1 and a maximum value of 10. Defaults to 4.
4137 .. option:: exitall_on_error
4139 When one job finishes in error, terminate the rest. The default is to wait
4140 for each job to finish.
4142 .. option:: continue_on_error=str
4144 Normally fio will exit the job on the first observed failure. If this option
4145 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
4146 EILSEQ) until the runtime is exceeded or the I/O size specified is
4147 completed. If this option is used, there are two more stats that are
4148 appended, the total error count and the first error. The error field given
4149 in the stats is the first error that was hit during the run.
4151 Note: a write error from the device may go unnoticed by fio when using
4152 buffered IO, as the write() (or similar) system call merely dirties the
4153 kernel pages, unless :option:`sync` or :option:`direct` is used. Device IO
4154 errors occur when the dirty data is actually written out to disk. If fully
4155 sync writes aren't desirable, :option:`fsync` or :option:`fdatasync` can be
4156 used as well. This is specific to writes, as reads are always synchronous.
4158 The allowed values are:
4161 Exit on any I/O or verify errors.
4164 Continue on read errors, exit on all others.
4167 Continue on write errors, exit on all others.
4170 Continue on any I/O error, exit on all others.
4173 Continue on verify errors, exit on all others.
4176 Continue on all errors.
4179 Backward-compatible alias for 'none'.
4182 Backward-compatible alias for 'all'.
4184 .. option:: ignore_error=str
4186 Sometimes you want to ignore some errors during test in that case you can
4187 specify error list for each error type, instead of only being able to
4188 ignore the default 'non-fatal error' using :option:`continue_on_error`.
4189 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
4190 given error type is separated with ':'. Error may be symbol ('ENOSPC',
4191 'ENOMEM') or integer. Example::
4193 ignore_error=EAGAIN,ENOSPC:122
4195 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
4196 WRITE. This option works by overriding :option:`continue_on_error` with
4197 the list of errors for each error type if any.
4199 .. option:: error_dump=bool
4201 If set dump every error even if it is non fatal, true by default. If
4202 disabled only fatal error will be dumped.
4204 Running predefined workloads
4205 ----------------------------
4207 Fio includes predefined profiles that mimic the I/O workloads generated by
4210 .. option:: profile=str
4212 The predefined workload to run. Current profiles are:
4215 Threaded I/O bench (tiotest/tiobench) like workload.
4218 Aerospike Certification Tool (ACT) like workload.
4220 To view a profile's additional options use :option:`--cmdhelp` after specifying
4221 the profile. For example::
4223 $ fio --profile=act --cmdhelp
4228 .. option:: device-names=str
4233 .. option:: load=int
4236 ACT load multiplier. Default: 1.
4238 .. option:: test-duration=time
4241 How long the entire test takes to run. When the unit is omitted, the value
4242 is given in seconds. Default: 24h.
4244 .. option:: threads-per-queue=int
4247 Number of read I/O threads per device. Default: 8.
4249 .. option:: read-req-num-512-blocks=int
4252 Number of 512B blocks to read at the time. Default: 3.
4254 .. option:: large-block-op-kbytes=int
4257 Size of large block ops in KiB (writes). Default: 131072.
4262 Set to run ACT prep phase.
4264 Tiobench profile options
4265 ~~~~~~~~~~~~~~~~~~~~~~~~
4267 .. option:: size=str
4272 .. option:: block=int
4275 Block size in bytes. Default: 4096.
4277 .. option:: numruns=int
4287 .. option:: threads=int
4292 Interpreting the output
4293 -----------------------
4296 Example output was based on the following:
4297 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
4298 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
4299 --runtime=2m --rw=rw
4301 Fio spits out a lot of output. While running, fio will display the status of the
4302 jobs created. An example of that would be::
4304 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]
4306 The characters inside the first set of square brackets denote the current status of
4307 each thread. The first character is the first job defined in the job file, and so
4308 forth. The possible values (in typical life cycle order) are:
4310 +------+-----+-----------------------------------------------------------+
4312 +======+=====+===========================================================+
4313 | P | | Thread setup, but not started. |
4314 +------+-----+-----------------------------------------------------------+
4315 | C | | Thread created. |
4316 +------+-----+-----------------------------------------------------------+
4317 | I | | Thread initialized, waiting or generating necessary data. |
4318 +------+-----+-----------------------------------------------------------+
4319 | | p | Thread running pre-reading file(s). |
4320 +------+-----+-----------------------------------------------------------+
4321 | | / | Thread is in ramp period. |
4322 +------+-----+-----------------------------------------------------------+
4323 | | R | Running, doing sequential reads. |
4324 +------+-----+-----------------------------------------------------------+
4325 | | r | Running, doing random reads. |
4326 +------+-----+-----------------------------------------------------------+
4327 | | W | Running, doing sequential writes. |
4328 +------+-----+-----------------------------------------------------------+
4329 | | w | Running, doing random writes. |
4330 +------+-----+-----------------------------------------------------------+
4331 | | M | Running, doing mixed sequential reads/writes. |
4332 +------+-----+-----------------------------------------------------------+
4333 | | m | Running, doing mixed random reads/writes. |
4334 +------+-----+-----------------------------------------------------------+
4335 | | D | Running, doing sequential trims. |
4336 +------+-----+-----------------------------------------------------------+
4337 | | d | Running, doing random trims. |
4338 +------+-----+-----------------------------------------------------------+
4339 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
4340 +------+-----+-----------------------------------------------------------+
4341 | | V | Running, doing verification of written data. |
4342 +------+-----+-----------------------------------------------------------+
4343 | f | | Thread finishing. |
4344 +------+-----+-----------------------------------------------------------+
4345 | E | | Thread exited, not reaped by main thread yet. |
4346 +------+-----+-----------------------------------------------------------+
4347 | _ | | Thread reaped. |
4348 +------+-----+-----------------------------------------------------------+
4349 | X | | Thread reaped, exited with an error. |
4350 +------+-----+-----------------------------------------------------------+
4351 | K | | Thread reaped, exited due to signal. |
4352 +------+-----+-----------------------------------------------------------+
4355 Example output was based on the following:
4356 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
4357 --time_based --rate=2512k --bs=256K --numjobs=10 \
4358 --name=readers --rw=read --name=writers --rw=write
4360 Fio will condense the thread string as not to take up more space on the command
4361 line than needed. For instance, if you have 10 readers and 10 writers running,
4362 the output would look like this::
4364 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]
4366 Note that the status string is displayed in order, so it's possible to tell which of
4367 the jobs are currently doing what. In the example above this means that jobs 1--10
4368 are readers and 11--20 are writers.
4370 The other values are fairly self explanatory -- number of threads currently
4371 running and doing I/O, the number of currently open files (f=), the estimated
4372 completion percentage, the rate of I/O since last check (read speed listed first,
4373 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
4374 and time to completion for the current running group. It's impossible to estimate
4375 runtime of the following groups (if any).
4378 Example output was based on the following:
4379 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
4380 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
4381 --bs=7K --name=Client1 --rw=write
4383 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
4384 each thread, group of threads, and disks in that order. For each overall thread (or
4385 group) the output looks like::
4387 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
4388 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
4389 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
4390 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
4391 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
4392 clat percentiles (usec):
4393 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
4394 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
4395 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
4396 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
4398 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
4399 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
4400 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
4401 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
4402 lat (msec) : 100=0.65%
4403 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
4404 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
4405 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4406 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
4407 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
4408 latency : target=0, window=0, percentile=100.00%, depth=8
4410 The job name (or first job's name when using :option:`group_reporting`) is printed,
4411 along with the group id, count of jobs being aggregated, last error id seen (which
4412 is 0 when there are no errors), pid/tid of that thread and the time the job/group
4413 completed. Below are the I/O statistics for each data direction performed (showing
4414 writes in the example above). In the order listed, they denote:
4417 The string before the colon shows the I/O direction the statistics
4418 are for. **IOPS** is the average I/Os performed per second. **BW**
4419 is the average bandwidth rate shown as: value in power of 2 format
4420 (value in power of 10 format). The last two values show: (**total
4421 I/O performed** in power of 2 format / **runtime** of that thread).
4424 Submission latency (**min** being the minimum, **max** being the
4425 maximum, **avg** being the average, **stdev** being the standard
4426 deviation). This is the time from when fio initialized the I/O
4427 to submission. For synchronous ioengines this includes the time
4428 up until just before the ioengine's queue function is called.
4429 For asynchronous ioengines this includes the time up through the
4430 completion of the ioengine's queue function (and commit function
4431 if it is defined). For sync I/O this row is not displayed as the
4432 slat is negligible. This value can be in nanoseconds,
4433 microseconds or milliseconds --- fio will choose the most
4434 appropriate base and print that (in the example above
4435 nanoseconds was the best scale). Note: in :option:`--minimal`
4436 mode latencies are always expressed in microseconds.
4439 Completion latency. Same names as slat, this denotes the time from
4440 submission to completion of the I/O pieces. For sync I/O, this
4441 represents the time from when the I/O was submitted to the
4442 operating system to when it was completed. For asynchronous
4443 ioengines this is the time from when the ioengine's queue (and
4444 commit if available) functions were completed to when the I/O's
4445 completion was reaped by fio.
4448 Total latency. Same names as slat and clat, this denotes the time from
4449 when fio created the I/O unit to completion of the I/O operation.
4450 It is the sum of submission and completion latency.
4453 Bandwidth statistics based on measurements from discrete
4454 intervals. Fio continuously monitors bytes transferred and I/O
4455 operations completed. By default fio calculates bandwidth in
4456 each half-second interval (see :option:`bwavgtime`) and reports
4457 descriptive statistics for the measurements here. Same names as
4458 the xlat stats, but also includes the number of samples taken
4459 (**samples**) and an approximate percentage of total aggregate
4460 bandwidth this thread received in its group (**per**). This
4461 last value is only really useful if the threads in this group
4462 are on the same disk, since they are then competing for disk
4466 IOPS statistics based on measurements from discrete intervals.
4467 For details see the description for bw above. See
4468 :option:`iopsavgtime` to control the duration of the intervals.
4469 Same values reported here as for bw except for percentage.
4471 **lat (nsec/usec/msec)**
4472 The distribution of I/O completion latencies. This is the time from when
4473 I/O leaves fio and when it gets completed. Unlike the separate
4474 read/write/trim sections above, the data here and in the remaining
4475 sections apply to all I/Os for the reporting group. 250=0.04% means that
4476 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
4477 of the I/Os required 250 to 499us for completion.
4480 CPU usage. User and system time, along with the number of context
4481 switches this thread went through, usage of system and user time, and
4482 finally the number of major and minor page faults. The CPU utilization
4483 numbers are averages for the jobs in that reporting group, while the
4484 context and fault counters are summed.
4487 The distribution of I/O depths over the job lifetime. The numbers are
4488 divided into powers of 2 and each entry covers depths from that value
4489 up to those that are lower than the next entry -- e.g., 16= covers
4490 depths from 16 to 31. Note that the range covered by a depth
4491 distribution entry can be different to the range covered by the
4492 equivalent submit/complete distribution entry.
4495 How many pieces of I/O were submitting in a single submit call. Each
4496 entry denotes that amount and below, until the previous entry -- e.g.,
4497 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
4498 call. Note that the range covered by a submit distribution entry can
4499 be different to the range covered by the equivalent depth distribution
4503 Like the above submit number, but for completions instead.
4506 The number of read/write/trim requests issued, and how many of them were
4510 These values are for :option:`latency_target` and related options. When
4511 these options are engaged, this section describes the I/O depth required
4512 to meet the specified latency target.
4515 Example output was based on the following:
4516 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
4517 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
4518 --rate=11M --name=write --rw=write --bs=2k --rate=700k
4520 After each client has been listed, the group statistics are printed. They
4521 will look like this::
4523 Run status group 0 (all jobs):
4524 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
4525 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
4527 For each data direction it prints:
4530 Aggregate bandwidth of threads in this group followed by the
4531 minimum and maximum bandwidth of all the threads in this group.
4532 Values outside of brackets are power-of-2 format and those
4533 within are the equivalent value in a power-of-10 format.
4535 Aggregate I/O performed of all threads in this group. The
4536 format is the same as bw.
4538 The smallest and longest runtimes of the threads in this group.
4540 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
4542 Disk stats (read/write):
4543 sda: ios=16398/16511, sectors=32321/65472, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
4545 Each value is printed for both reads and writes, with reads first. The
4549 Number of I/Os performed by all groups.
4551 Amount of data transferred in units of 512 bytes for all groups.
4553 Number of merges performed by the I/O scheduler.
4555 Number of ticks we kept the disk busy.
4557 Total time spent in the disk queue.
4559 The disk utilization. A value of 100% means we kept the disk
4560 busy constantly, 50% would be a disk idling half of the time.
4562 It is also possible to get fio to dump the current output while it is running,
4563 without terminating the job. To do that, send fio the **USR1** signal. You can
4564 also get regularly timed dumps by using the :option:`--status-interval`
4565 parameter, or by creating a file in :file:`/tmp` named
4566 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
4567 current output status.
4573 For scripted usage where you typically want to generate tables or graphs of the
4574 results, fio can output the results in a semicolon separated format. The format
4575 is one long line of values, such as::
4577 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%
4578 A description of this job goes here.
4580 The job description (if provided) follows on a second line for terse v2.
4581 It appears on the same line for other terse versions.
4583 To enable terse output, use the :option:`--minimal` or
4584 :option:`--output-format`\=terse command line options. The
4585 first value is the version of the terse output format. If the output has to be
4586 changed for some reason, this number will be incremented by 1 to signify that
4589 Split up, the format is as follows (comments in brackets denote when a
4590 field was introduced or whether it's specific to some terse version):
4594 terse version, fio version [v3], jobname, groupid, error
4598 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4599 Submission latency: min, max, mean, stdev (usec)
4600 Completion latency: min, max, mean, stdev (usec)
4601 Completion latency percentiles: 20 fields (see below)
4602 Total latency: min, max, mean, stdev (usec)
4603 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4604 IOPS [v5]: min, max, mean, stdev, number of samples
4610 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
4611 Submission latency: min, max, mean, stdev (usec)
4612 Completion latency: min, max, mean, stdev (usec)
4613 Completion latency percentiles: 20 fields (see below)
4614 Total latency: min, max, mean, stdev (usec)
4615 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
4616 IOPS [v5]: min, max, mean, stdev, number of samples
4618 TRIM status [all but version 3]:
4620 Fields are similar to READ/WRITE status.
4624 user, system, context switches, major faults, minor faults
4628 <=1, 2, 4, 8, 16, 32, >=64
4630 I/O latencies microseconds::
4632 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
4634 I/O latencies milliseconds::
4636 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
4638 Disk utilization [v3]::
4640 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
4641 time spent in queue, disk utilization percentage
4643 Additional Info (dependent on continue_on_error, default off)::
4645 total # errors, first error code
4647 Additional Info (dependent on description being set)::
4651 Completion latency percentiles can be a grouping of up to 20 sets, so for the
4652 terse output fio writes all of them. Each field will look like this::
4656 which is the Xth percentile, and the `usec` latency associated with it.
4658 For `Disk utilization`, all disks used by fio are shown. So for each disk there
4659 will be a disk utilization section.
4661 Below is a single line containing short names for each of the fields in the
4662 minimal output v3, separated by semicolons::
4664 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
4666 In client/server mode terse output differs from what appears when jobs are run
4667 locally. Disk utilization data is omitted from the standard terse output and
4668 for v3 and later appears on its own separate line at the end of each terse
4675 The `json` output format is intended to be both human readable and convenient
4676 for automated parsing. For the most part its sections mirror those of the
4677 `normal` output. The `runtime` value is reported in msec and the `bw` value is
4678 reported in 1024 bytes per second units.
4684 The `json+` output format is identical to the `json` output format except that it
4685 adds a full dump of the completion latency bins. Each `bins` object contains a
4686 set of (key, value) pairs where keys are latency durations and values count how
4687 many I/Os had completion latencies of the corresponding duration. For example,
4690 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
4692 This data indicates that one I/O required 87,552ns to complete, two I/Os required
4693 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
4695 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
4696 json+ output and generates CSV-formatted latency data suitable for plotting.
4698 The latency durations actually represent the midpoints of latency intervals.
4699 For details refer to :file:`stat.h`.
4705 There are two trace file format that you can encounter. The older (v1) format is
4706 unsupported since version 1.20-rc3 (March 2008). It will still be described
4707 below in case that you get an old trace and want to understand it.
4709 In any case the trace is a simple text file with a single action per line.
4712 Trace file format v1
4713 ~~~~~~~~~~~~~~~~~~~~
4715 Each line represents a single I/O action in the following format::
4719 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
4721 This format is not supported in fio versions >= 1.20-rc3.
4724 Trace file format v2
4725 ~~~~~~~~~~~~~~~~~~~~
4727 The second version of the trace file format was added in fio version 1.17. It
4728 allows one to access more than one file per trace and has a bigger set of possible
4731 The first line of the trace file has to be::
4735 Following this can be lines in two different formats, which are described below.
4737 The file management format::
4741 The `filename` is given as an absolute path. The `action` can be one of these:
4744 Add the given `filename` to the trace.
4746 Open the file with the given `filename`. The `filename` has to have
4747 been added with the **add** action before.
4749 Close the file with the given `filename`. The file has to have been
4753 The file I/O action format::
4755 filename action offset length
4757 The `filename` is given as an absolute path, and has to have been added and
4758 opened before it can be used with this format. The `offset` and `length` are
4759 given in bytes. The `action` can be one of these:
4762 Wait for `offset` microseconds. Everything below 100 is discarded.
4763 The time is relative to the previous `wait` statement. Note that
4764 action `wait` is not allowed as of version 3, as the same behavior
4765 can be achieved using timestamps.
4767 Read `length` bytes beginning from `offset`.
4769 Write `length` bytes beginning from `offset`.
4771 :manpage:`fsync(2)` the file.
4773 :manpage:`fdatasync(2)` the file.
4775 Trim the given file from the given `offset` for `length` bytes.
4778 Trace file format v3
4779 ~~~~~~~~~~~~~~~~~~~~
4781 The third version of the trace file format was added in fio version 3.31. It
4782 forces each action to have a timestamp associated with it.
4784 The first line of the trace file has to be::
4788 Following this can be lines in two different formats, which are described below.
4790 The file management format::
4792 timestamp filename action
4794 The file I/O action format::
4796 timestamp filename action offset length
4798 The `timestamp` is relative to the beginning of the run (ie starts at 0). The
4799 `filename`, `action`, `offset` and `length` are identical to version 2, except
4800 that version 3 does not allow the `wait` action.
4803 I/O Replay - Merging Traces
4804 ---------------------------
4806 Colocation is a common practice used to get the most out of a machine.
4807 Knowing which workloads play nicely with each other and which ones don't is
4808 a much harder task. While fio can replay workloads concurrently via multiple
4809 jobs, it leaves some variability up to the scheduler making results harder to
4810 reproduce. Merging is a way to make the order of events consistent.
4812 Merging is integrated into I/O replay and done when a
4813 :option:`merge_blktrace_file` is specified. The list of files passed to
4814 :option:`read_iolog` go through the merge process and output a single file
4815 stored to the specified file. The output file is passed on as if it were the
4816 only file passed to :option:`read_iolog`. An example would look like::
4818 $ fio --read_iolog="<file1>:<file2>" --merge_blktrace_file="<output_file>"
4820 Creating only the merged file can be done by passing the command line argument
4821 :option:`--merge-blktrace-only`.
4823 Scaling traces can be done to see the relative impact of any particular trace
4824 being slowed down or sped up. :option:`merge_blktrace_scalars` takes in a colon
4825 separated list of percentage scalars. It is index paired with the files passed
4826 to :option:`read_iolog`.
4828 With scaling, it may be desirable to match the running time of all traces.
4829 This can be done with :option:`merge_blktrace_iters`. It is index paired with
4830 :option:`read_iolog` just like :option:`merge_blktrace_scalars`.
4832 In an example, given two traces, A and B, each 60s long. If we want to see
4833 the impact of trace A issuing IOs twice as fast and repeat trace A over the
4834 runtime of trace B, the following can be done::
4836 $ fio --read_iolog="<trace_a>:"<trace_b>" --merge_blktrace_file"<output_file>" --merge_blktrace_scalars="50:100" --merge_blktrace_iters="2:1"
4838 This runs trace A at 2x the speed twice for approximately the same runtime as
4839 a single run of trace B.
4842 CPU idleness profiling
4843 ----------------------
4845 In some cases, we want to understand CPU overhead in a test. For example, we
4846 test patches for the specific goodness of whether they reduce CPU usage.
4847 Fio implements a balloon approach to create a thread per CPU that runs at idle
4848 priority, meaning that it only runs when nobody else needs the cpu.
4849 By measuring the amount of work completed by the thread, idleness of each CPU
4850 can be derived accordingly.
4852 An unit work is defined as touching a full page of unsigned characters. Mean and
4853 standard deviation of time to complete an unit work is reported in "unit work"
4854 section. Options can be chosen to report detailed percpu idleness or overall
4855 system idleness by aggregating percpu stats.
4858 Verification and triggers
4859 -------------------------
4861 Fio is usually run in one of two ways, when data verification is done. The first
4862 is a normal write job of some sort with verify enabled. When the write phase has
4863 completed, fio switches to reads and verifies everything it wrote. The second
4864 model is running just the write phase, and then later on running the same job
4865 (but with reads instead of writes) to repeat the same I/O patterns and verify
4866 the contents. Both of these methods depend on the write phase being completed,
4867 as fio otherwise has no idea how much data was written.
4869 With verification triggers, fio supports dumping the current write state to
4870 local files. Then a subsequent read verify workload can load this state and know
4871 exactly where to stop. This is useful for testing cases where power is cut to a
4872 server in a managed fashion, for instance.
4874 A verification trigger consists of two things:
4876 1) Storing the write state of each job.
4877 2) Executing a trigger command.
4879 The write state is relatively small, on the order of hundreds of bytes to single
4880 kilobytes. It contains information on the number of completions done, the last X
4883 A trigger is invoked either through creation ('touch') of a specified file in
4884 the system, or through a timeout setting. If fio is run with
4885 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
4886 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
4887 will fire off the trigger (thus saving state, and executing the trigger
4890 For client/server runs, there's both a local and remote trigger. If fio is
4891 running as a server backend, it will send the job states back to the client for
4892 safe storage, then execute the remote trigger, if specified. If a local trigger
4893 is specified, the server will still send back the write state, but the client
4894 will then execute the trigger.
4896 Verification trigger example
4897 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4899 Let's say we want to run a powercut test on the remote Linux machine 'server'.
4900 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
4901 some point during the run, and we'll run this test from the safety or our local
4902 machine, 'localbox'. On the server, we'll start the fio backend normally::
4904 server# fio --server
4906 and on the client, we'll fire off the workload::
4908 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
4910 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
4912 echo b > /proc/sysrq-trigger
4914 on the server once it has received the trigger and sent us the write state. This
4915 will work, but it's not **really** cutting power to the server, it's merely
4916 abruptly rebooting it. If we have a remote way of cutting power to the server
4917 through IPMI or similar, we could do that through a local trigger command
4918 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
4919 ipmi-reboot. On localbox, we could then have run fio with a local trigger
4922 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
4924 For this case, fio would wait for the server to send us the write state, then
4925 execute ``ipmi-reboot server`` when that happened.
4927 Loading verify state
4928 ~~~~~~~~~~~~~~~~~~~~
4930 To load stored write state, a read verification job file must contain the
4931 :option:`verify_state_load` option. If that is set, fio will load the previously
4932 stored state. For a local fio run this is done by loading the files directly,
4933 and on a client/server run, the server backend will ask the client to send the
4934 files over and load them from there.
4940 Fio supports a variety of log file formats, for logging latencies, bandwidth,
4941 and IOPS. The logs share a common format, which looks like this:
4943 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
4944 *offset* (`bytes`), *command priority*
4946 *Time* for the log entry is always in milliseconds. The *value* logged depends
4947 on the type of log, it will be one of the following:
4950 Value is latency in nsecs
4956 *Data direction* is one of the following:
4965 The entry's *block size* is always in bytes. The *offset* is the position in bytes
4966 from the start of the file for that particular I/O. The logging of the offset can be
4967 toggled with :option:`log_offset`.
4969 *Command priority* is 0 for normal priority and 1 for high priority. This is controlled
4970 by the ioengine specific :option:`cmdprio_percentage`.
4972 Fio defaults to logging every individual I/O but when windowed logging is set
4973 through :option:`log_avg_msec`, either the average (by default) or the maximum
4974 (:option:`log_max_value` is set) *value* seen over the specified period of time
4975 is recorded. Each *data direction* seen within the window period will aggregate
4976 its values in a separate row. Further, when using windowed logging the *block
4977 size* and *offset* entries will always contain 0.
4983 Normally fio is invoked as a stand-alone application on the machine where the
4984 I/O workload should be generated. However, the backend and frontend of fio can
4985 be run separately i.e., the fio server can generate an I/O workload on the "Device
4986 Under Test" while being controlled by a client on another machine.
4988 Start the server on the machine which has access to the storage DUT::
4992 where `args` defines what fio listens to. The arguments are of the form
4993 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
4994 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
4995 *hostname* is either a hostname or IP address, and *port* is the port to listen
4996 to (only valid for TCP/IP, not a local socket). Some examples:
5000 Start a fio server, listening on all interfaces on the default port (8765).
5002 2) ``fio --server=ip:hostname,4444``
5004 Start a fio server, listening on IP belonging to hostname and on port 4444.
5006 3) ``fio --server=ip6:::1,4444``
5008 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
5010 4) ``fio --server=,4444``
5012 Start a fio server, listening on all interfaces on port 4444.
5014 5) ``fio --server=1.2.3.4``
5016 Start a fio server, listening on IP 1.2.3.4 on the default port.
5018 6) ``fio --server=sock:/tmp/fio.sock``
5020 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
5022 Once a server is running, a "client" can connect to the fio server with::
5024 fio <local-args> --client=<server> <remote-args> <job file(s)>
5026 where `local-args` are arguments for the client where it is running, `server`
5027 is the connect string, and `remote-args` and `job file(s)` are sent to the
5028 server. The `server` string follows the same format as it does on the server
5029 side, to allow IP/hostname/socket and port strings.
5031 Fio can connect to multiple servers this way::
5033 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
5035 If the job file is located on the fio server, then you can tell the server to
5036 load a local file as well. This is done by using :option:`--remote-config` ::
5038 fio --client=server --remote-config /path/to/file.fio
5040 Then fio will open this local (to the server) job file instead of being passed
5041 one from the client.
5043 If you have many servers (example: 100 VMs/containers), you can input a pathname
5044 of a file containing host IPs/names as the parameter value for the
5045 :option:`--client` option. For example, here is an example :file:`host.list`
5046 file containing 2 hostnames::
5048 host1.your.dns.domain
5049 host2.your.dns.domain
5051 The fio command would then be::
5053 fio --client=host.list <job file(s)>
5055 In this mode, you cannot input server-specific parameters or job files -- all
5056 servers receive the same job file.
5058 In order to let ``fio --client`` runs use a shared filesystem from multiple
5059 hosts, ``fio --client`` now prepends the IP address of the server to the
5060 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
5061 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
5062 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
5063 192.168.10.121, then fio will create two files::
5065 /mnt/nfs/fio/192.168.10.120.fileio.tmp
5066 /mnt/nfs/fio/192.168.10.121.fileio.tmp
5068 Terse output in client/server mode will differ slightly from what is produced
5069 when fio is run in stand-alone mode. See the terse output section for details.