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=jobfile
172 Convert `jobfile` 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 With :option:`kb_base`\=1024 (the default), the unit prefixes are opposite
548 from those specified in the SI and IEC 80000-13 standards to provide
549 compatibility with old scripts. For example, 4k means 4096.
551 For quantities of data, an optional unit of 'B' may be included
552 (e.g., 'kB' is the same as 'k').
554 The *integer suffix* is not case sensitive (e.g., m/mi mean mebi/mega,
555 not milli). 'b' and 'B' both mean byte, not bit.
557 Examples with :option:`kb_base`\=1000:
559 * *4 KiB*: 4096, 4096b, 4096B, 4ki, 4kib, 4kiB, 4Ki, 4KiB
560 * *1 MiB*: 1048576, 1mi, 1024ki
561 * *1 MB*: 1000000, 1m, 1000k
562 * *1 TiB*: 1099511627776, 1ti, 1024gi, 1048576mi
563 * *1 TB*: 1000000000, 1t, 1000m, 1000000k
565 Examples with :option:`kb_base`\=1024 (default):
567 * *4 KiB*: 4096, 4096b, 4096B, 4k, 4kb, 4kB, 4K, 4KB
568 * *1 MiB*: 1048576, 1m, 1024k
569 * *1 MB*: 1000000, 1mi, 1000ki
570 * *1 TiB*: 1099511627776, 1t, 1024g, 1048576m
571 * *1 TB*: 1000000000, 1ti, 1000mi, 1000000ki
573 To specify times (units are not case sensitive):
577 * *M* -- means minutes
578 * *s* -- or sec means seconds (default)
579 * *ms* -- or *msec* means milliseconds
580 * *us* -- or *usec* means microseconds
582 If the option accepts an upper and lower range, use a colon ':' or
583 minus '-' to separate such values. See :ref:`irange <irange>`.
584 If the lower value specified happens to be larger than the upper value
585 the two values are swapped.
590 Boolean. Usually parsed as an integer, however only defined for
591 true and false (1 and 0).
596 Integer range with suffix. Allows value range to be given, such as
597 1024-4096. A colon may also be used as the separator, e.g. 1k:4k. If the
598 option allows two sets of ranges, they can be specified with a ',' or '/'
599 delimiter: 1k-4k/8k-32k. Also see :ref:`int <int>`.
602 A list of floating point numbers, separated by a ':' character.
604 With the above in mind, here follows the complete list of fio job parameters.
610 .. option:: kb_base=int
612 Select the interpretation of unit prefixes in input parameters.
615 Inputs comply with IEC 80000-13 and the International
616 System of Units (SI). Use:
618 - power-of-2 values with IEC prefixes (e.g., KiB)
619 - power-of-10 values with SI prefixes (e.g., kB)
622 Compatibility mode (default). To avoid breaking old scripts:
624 - power-of-2 values with SI prefixes
625 - power-of-10 values with IEC prefixes
627 See :option:`bs` for more details on input parameters.
629 Outputs always use correct prefixes. Most outputs include both
632 bw=2383.3kB/s (2327.4KiB/s)
634 If only one value is reported, then kb_base selects the one to use:
636 **1000** -- SI prefixes
638 **1024** -- IEC prefixes
640 .. option:: unit_base=int
642 Base unit for reporting. Allowed values are:
645 Use auto-detection (default).
657 ASCII name of the job. This may be used to override the name printed by fio
658 for this job. Otherwise the job name is used. On the command line this
659 parameter has the special purpose of also signaling the start of a new job.
661 .. option:: description=str
663 Text description of the job. Doesn't do anything except dump this text
664 description when this job is run. It's not parsed.
666 .. option:: loops=int
668 Run the specified number of iterations of this job. Used to repeat the same
669 workload a given number of times. Defaults to 1.
671 .. option:: numjobs=int
673 Create the specified number of clones of this job. Each clone of job
674 is spawned as an independent thread or process. May be used to setup a
675 larger number of threads/processes doing the same thing. Each thread is
676 reported separately; to see statistics for all clones as a whole, use
677 :option:`group_reporting` in conjunction with :option:`new_group`.
678 See :option:`--max-jobs`. Default: 1.
681 Time related parameters
682 ~~~~~~~~~~~~~~~~~~~~~~~
684 .. option:: runtime=time
686 Tell fio to terminate processing after the specified period of time. It
687 can be quite hard to determine for how long a specified job will run, so
688 this parameter is handy to cap the total runtime to a given time. When
689 the unit is omitted, the value is interpreted in seconds.
691 .. option:: time_based
693 If set, fio will run for the duration of the :option:`runtime` specified
694 even if the file(s) are completely read or written. It will simply loop over
695 the same workload as many times as the :option:`runtime` allows.
697 .. option:: startdelay=irange(time)
699 Delay the start of job for the specified amount of time. Can be a single
700 value or a range. When given as a range, each thread will choose a value
701 randomly from within the range. Value is in seconds if a unit is omitted.
703 .. option:: ramp_time=time
705 If set, fio will run the specified workload for this amount of time before
706 logging any performance numbers. Useful for letting performance settle
707 before logging results, thus minimizing the runtime required for stable
708 results. Note that the ``ramp_time`` is considered lead in time for a job,
709 thus it will increase the total runtime if a special timeout or
710 :option:`runtime` is specified. When the unit is omitted, the value is
713 .. option:: clocksource=str
715 Use the given clocksource as the base of timing. The supported options are:
718 :manpage:`gettimeofday(2)`
721 :manpage:`clock_gettime(2)`
724 Internal CPU clock source
726 cpu is the preferred clocksource if it is reliable, as it is very fast (and
727 fio is heavy on time calls). Fio will automatically use this clocksource if
728 it's supported and considered reliable on the system it is running on,
729 unless another clocksource is specifically set. For x86/x86-64 CPUs, this
730 means supporting TSC Invariant.
732 .. option:: gtod_reduce=bool
734 Enable all of the :manpage:`gettimeofday(2)` reducing options
735 (:option:`disable_clat`, :option:`disable_slat`, :option:`disable_bw_measurement`) plus
736 reduce precision of the timeout somewhat to really shrink the
737 :manpage:`gettimeofday(2)` call count. With this option enabled, we only do
738 about 0.4% of the :manpage:`gettimeofday(2)` calls we would have done if all
739 time keeping was enabled.
741 .. option:: gtod_cpu=int
743 Sometimes it's cheaper to dedicate a single thread of execution to just
744 getting the current time. Fio (and databases, for instance) are very
745 intensive on :manpage:`gettimeofday(2)` calls. With this option, you can set
746 one CPU aside for doing nothing but logging current time to a shared memory
747 location. Then the other threads/processes that run I/O workloads need only
748 copy that segment, instead of entering the kernel with a
749 :manpage:`gettimeofday(2)` call. The CPU set aside for doing these time
750 calls will be excluded from other uses. Fio will manually clear it from the
751 CPU mask of other jobs.
757 .. option:: directory=str
759 Prefix filenames with this directory. Used to place files in a different
760 location than :file:`./`. You can specify a number of directories by
761 separating the names with a ':' character. These directories will be
762 assigned equally distributed to job clones created by :option:`numjobs` as
763 long as they are using generated filenames. If specific `filename(s)` are
764 set fio will use the first listed directory, and thereby matching the
765 `filename` semantic (which generates a file for each clone if not
766 specified, but lets all clones use the same file if set).
768 See the :option:`filename` option for information on how to escape "``:``"
769 characters within the directory path itself.
771 Note: To control the directory fio will use for internal state files
772 use :option:`--aux-path`.
774 .. option:: filename=str
776 Fio normally makes up a `filename` based on the job name, thread number, and
777 file number (see :option:`filename_format`). If you want to share files
778 between threads in a job or several
779 jobs with fixed file paths, specify a `filename` for each of them to override
780 the default. If the ioengine is file based, you can specify a number of files
781 by separating the names with a ':' colon. So if you wanted a job to open
782 :file:`/dev/sda` and :file:`/dev/sdb` as the two working files, you would use
783 ``filename=/dev/sda:/dev/sdb``. This also means that whenever this option is
784 specified, :option:`nrfiles` is ignored. The size of regular files specified
785 by this option will be :option:`size` divided by number of files unless an
786 explicit size is specified by :option:`filesize`.
788 Each colon in the wanted path must be escaped with a ``\``
789 character. For instance, if the path is :file:`/dev/dsk/foo@3,0:c` then you
790 would use ``filename=/dev/dsk/foo@3,0\:c`` and if the path is
791 :file:`F:\\filename` then you would use ``filename=F\:\filename``.
793 On Windows, disk devices are accessed as :file:`\\\\.\\PhysicalDrive0` for
794 the first device, :file:`\\\\.\\PhysicalDrive1` for the second etc.
795 Note: Windows and FreeBSD prevent write access to areas
796 of the disk containing in-use data (e.g. filesystems).
798 The filename "`-`" is a reserved name, meaning *stdin* or *stdout*. Which
799 of the two depends on the read/write direction set.
801 .. option:: filename_format=str
803 If sharing multiple files between jobs, it is usually necessary to have fio
804 generate the exact names that you want. By default, fio will name a file
805 based on the default file format specification of
806 :file:`jobname.jobnumber.filenumber`. With this option, that can be
807 customized. Fio will recognize and replace the following keywords in this
811 The name of the worker thread or process.
813 The incremental number of the worker thread or process.
815 The incremental number of the file for that worker thread or
818 To have dependent jobs share a set of files, this option can be set to have
819 fio generate filenames that are shared between the two. For instance, if
820 :file:`testfiles.$filenum` is specified, file number 4 for any job will be
821 named :file:`testfiles.4`. The default of :file:`$jobname.$jobnum.$filenum`
822 will be used if no other format specifier is given.
824 If you specify a path then the directories will be created up to the
825 main directory for the file. So for example if you specify
826 ``filename_format=a/b/c/$jobnum`` then the directories a/b/c will be
827 created before the file setup part of the job. If you specify
828 :option:`directory` then the path will be relative that directory,
829 otherwise it is treated as the absolute path.
831 .. option:: unique_filename=bool
833 To avoid collisions between networked clients, fio defaults to prefixing any
834 generated filenames (with a directory specified) with the source of the
835 client connecting. To disable this behavior, set this option to 0.
837 .. option:: opendir=str
839 Recursively open any files below directory `str`.
841 .. option:: lockfile=str
843 Fio defaults to not locking any files before it does I/O to them. If a file
844 or file descriptor is shared, fio can serialize I/O to that file to make the
845 end result consistent. This is usual for emulating real workloads that share
846 files. The lock modes are:
849 No locking. The default.
851 Only one thread or process may do I/O at a time, excluding all
854 Read-write locking on the file. Many readers may
855 access the file at the same time, but writes get exclusive access.
857 .. option:: nrfiles=int
859 Number of files to use for this job. Defaults to 1. The size of files
860 will be :option:`size` divided by this unless explicit size is specified by
861 :option:`filesize`. Files are created for each thread separately, and each
862 file will have a file number within its name by default, as explained in
863 :option:`filename` section.
866 .. option:: openfiles=int
868 Number of files to keep open at the same time. Defaults to the same as
869 :option:`nrfiles`, can be set smaller to limit the number simultaneous
872 .. option:: file_service_type=str
874 Defines how fio decides which file from a job to service next. The following
878 Choose a file at random.
881 Round robin over opened files. This is the default.
884 Finish one file before moving on to the next. Multiple files can
885 still be open depending on :option:`openfiles`.
888 Use a *Zipf* distribution to decide what file to access.
891 Use a *Pareto* distribution to decide what file to access.
894 Use a *Gaussian* (normal) distribution to decide what file to
900 For *random*, *roundrobin*, and *sequential*, a postfix can be appended to
901 tell fio how many I/Os to issue before switching to a new file. For example,
902 specifying ``file_service_type=random:8`` would cause fio to issue
903 8 I/Os before selecting a new file at random. For the non-uniform
904 distributions, a floating point postfix can be given to influence how the
905 distribution is skewed. See :option:`random_distribution` for a description
906 of how that would work.
908 .. option:: ioscheduler=str
910 Attempt to switch the device hosting the file to the specified I/O scheduler
913 .. option:: create_serialize=bool
915 If true, serialize the file creation for the jobs. This may be handy to
916 avoid interleaving of data files, which may greatly depend on the filesystem
917 used and even the number of processors in the system. Default: true.
919 .. option:: create_fsync=bool
921 :manpage:`fsync(2)` the data file after creation. This is the default.
923 .. option:: create_on_open=bool
925 If true, don't pre-create files but allow the job's open() to create a file
926 when it's time to do I/O. Default: false -- pre-create all necessary files
929 .. option:: create_only=bool
931 If true, fio will only run the setup phase of the job. If files need to be
932 laid out or updated on disk, only that will be done -- the actual job contents
933 are not executed. Default: false.
935 .. option:: allow_file_create=bool
937 If true, fio is permitted to create files as part of its workload. If this
938 option is false, then fio will error out if
939 the files it needs to use don't already exist. Default: true.
941 .. option:: allow_mounted_write=bool
943 If this isn't set, fio will abort jobs that are destructive (e.g. that write)
944 to what appears to be a mounted device or partition. This should help catch
945 creating inadvertently destructive tests, not realizing that the test will
946 destroy data on the mounted file system. Note that some platforms don't allow
947 writing against a mounted device regardless of this option. Default: false.
949 .. option:: pre_read=bool
951 If this is given, files will be pre-read into memory before starting the
952 given I/O operation. This will also clear the :option:`invalidate` flag,
953 since it is pointless to pre-read and then drop the cache. This will only
954 work for I/O engines that are seek-able, since they allow you to read the
955 same data multiple times. Thus it will not work on non-seekable I/O engines
956 (e.g. network, splice). Default: false.
958 .. option:: unlink=bool
960 Unlink the job files when done. Not the default, as repeated runs of that
961 job would then waste time recreating the file set again and again. Default:
964 .. option:: unlink_each_loop=bool
966 Unlink job files after each iteration or loop. Default: false.
968 .. option:: zonemode=str
973 The :option:`zonerange`, :option:`zonesize`,
974 :option `zonecapacity` and option:`zoneskip`
975 parameters are ignored.
977 I/O happens in a single zone until
978 :option:`zonesize` bytes have been transferred.
979 After that number of bytes has been
980 transferred processing of the next zone
981 starts. :option `zonecapacity` is ignored.
983 Zoned block device mode. I/O happens
984 sequentially in each zone, even if random I/O
985 has been selected. Random I/O happens across
986 all zones instead of being restricted to a
987 single zone. The :option:`zoneskip` parameter
988 is ignored. :option:`zonerange` and
989 :option:`zonesize` must be identical.
991 .. option:: zonerange=int
993 Size of a single zone. See also :option:`zonesize` and
996 .. option:: zonesize=int
998 For :option:`zonemode` =strided, this is the number of bytes to
999 transfer before skipping :option:`zoneskip` bytes. If this parameter
1000 is smaller than :option:`zonerange` then only a fraction of each zone
1001 with :option:`zonerange` bytes will be accessed. If this parameter is
1002 larger than :option:`zonerange` then each zone will be accessed
1003 multiple times before skipping to the next zone.
1005 For :option:`zonemode` =zbd, this is the size of a single zone. The
1006 :option:`zonerange` parameter is ignored in this mode.
1009 .. option:: zonecapacity=int
1011 For :option:`zonemode` =zbd, this defines the capacity of a single zone,
1012 which is the accessible area starting from the zone start address.
1013 This parameter only applies when using :option:`zonemode` =zbd in
1014 combination with regular block devices. If not specified it defaults to
1015 the zone size. If the target device is a zoned block device, the zone
1016 capacity is obtained from the device information and this option is
1019 .. option:: zoneskip=int
1021 For :option:`zonemode` =strided, the number of bytes to skip after
1022 :option:`zonesize` bytes of data have been transferred. This parameter
1023 must be zero for :option:`zonemode` =zbd.
1025 .. option:: read_beyond_wp=bool
1027 This parameter applies to :option:`zonemode` =zbd only.
1029 Zoned block devices are block devices that consist of multiple zones.
1030 Each zone has a type, e.g. conventional or sequential. A conventional
1031 zone can be written at any offset that is a multiple of the block
1032 size. Sequential zones must be written sequentially. The position at
1033 which a write must occur is called the write pointer. A zoned block
1034 device can be either drive managed, host managed or host aware. For
1035 host managed devices the host must ensure that writes happen
1036 sequentially. Fio recognizes host managed devices and serializes
1037 writes to sequential zones for these devices.
1039 If a read occurs in a sequential zone beyond the write pointer then
1040 the zoned block device will complete the read without reading any data
1041 from the storage medium. Since such reads lead to unrealistically high
1042 bandwidth and IOPS numbers fio only reads beyond the write pointer if
1043 explicitly told to do so. Default: false.
1045 .. option:: max_open_zones=int
1047 When running a random write test across an entire drive many more
1048 zones will be open than in a typical application workload. Hence this
1049 command line option that allows to limit the number of open zones. The
1050 number of open zones is defined as the number of zones to which write
1051 commands are issued.
1053 .. option:: zone_reset_threshold=float
1055 A number between zero and one that indicates the ratio of logical
1056 blocks with data to the total number of logical blocks in the test
1057 above which zones should be reset periodically.
1059 .. option:: zone_reset_frequency=float
1061 A number between zero and one that indicates how often a zone reset
1062 should be issued if the zone reset threshold has been exceeded. A zone
1063 reset is submitted after each (1 / zone_reset_frequency) write
1064 requests. This and the previous parameter can be used to simulate
1065 garbage collection activity.
1071 .. option:: direct=bool
1073 If value is true, use non-buffered I/O. This is usually O_DIRECT. Note that
1074 OpenBSD and ZFS on Solaris don't support direct I/O. On Windows the synchronous
1075 ioengines don't support direct I/O. Default: false.
1077 .. option:: atomic=bool
1079 If value is true, attempt to use atomic direct I/O. Atomic writes are
1080 guaranteed to be stable once acknowledged by the operating system. Only
1081 Linux supports O_ATOMIC right now.
1083 .. option:: buffered=bool
1085 If value is true, use buffered I/O. This is the opposite of the
1086 :option:`direct` option. Defaults to true.
1088 .. option:: readwrite=str, rw=str
1090 Type of I/O pattern. Accepted values are:
1097 Sequential trims (Linux block devices and SCSI
1098 character devices only).
1104 Random trims (Linux block devices and SCSI
1105 character devices only).
1107 Sequential mixed reads and writes.
1109 Random mixed reads and writes.
1111 Sequential trim+write sequences. Blocks will be trimmed first,
1112 then the same blocks will be written to.
1114 Fio defaults to read if the option is not specified. For the mixed I/O
1115 types, the default is to split them 50/50. For certain types of I/O the
1116 result may still be skewed a bit, since the speed may be different.
1118 It is possible to specify the number of I/Os to do before getting a new
1119 offset by appending ``:<nr>`` to the end of the string given. For a
1120 random read, it would look like ``rw=randread:8`` for passing in an offset
1121 modifier with a value of 8. If the suffix is used with a sequential I/O
1122 pattern, then the *<nr>* value specified will be **added** to the generated
1123 offset for each I/O turning sequential I/O into sequential I/O with holes.
1124 For instance, using ``rw=write:4k`` will skip 4k for every write. Also see
1125 the :option:`rw_sequencer` option.
1127 .. option:: rw_sequencer=str
1129 If an offset modifier is given by appending a number to the ``rw=<str>``
1130 line, then this option controls how that number modifies the I/O offset
1131 being generated. Accepted values are:
1134 Generate sequential offset.
1136 Generate the same offset.
1138 ``sequential`` is only useful for random I/O, where fio would normally
1139 generate a new random offset for every I/O. If you append e.g. 8 to randread,
1140 you would get a new random offset for every 8 I/Os. The result would be a
1141 seek for only every 8 I/Os, instead of for every I/O. Use ``rw=randread:8``
1142 to specify that. As sequential I/O is already sequential, setting
1143 ``sequential`` for that would not result in any differences. ``identical``
1144 behaves in a similar fashion, except it sends the same offset 8 number of
1145 times before generating a new offset.
1147 .. option:: unified_rw_reporting=bool
1149 Fio normally reports statistics on a per data direction basis, meaning that
1150 reads, writes, and trims are accounted and reported separately. If this
1151 option is set fio sums the results and report them as "mixed" instead.
1153 .. option:: randrepeat=bool
1155 Seed the random number generator used for random I/O patterns in a
1156 predictable way so the pattern is repeatable across runs. Default: true.
1158 .. option:: allrandrepeat=bool
1160 Seed all random number generators in a predictable way so results are
1161 repeatable across runs. Default: false.
1163 .. option:: randseed=int
1165 Seed the random number generators based on this seed value, to be able to
1166 control what sequence of output is being generated. If not set, the random
1167 sequence depends on the :option:`randrepeat` setting.
1169 .. option:: fallocate=str
1171 Whether pre-allocation is performed when laying down files.
1172 Accepted values are:
1175 Do not pre-allocate space.
1178 Use a platform's native pre-allocation call but fall back to
1179 **none** behavior if it fails/is not implemented.
1182 Pre-allocate via :manpage:`posix_fallocate(3)`.
1185 Pre-allocate via :manpage:`fallocate(2)` with
1186 FALLOC_FL_KEEP_SIZE set.
1189 Extend file to final size via :manpage:`ftruncate(2)`
1190 instead of allocating.
1193 Backward-compatible alias for **none**.
1196 Backward-compatible alias for **posix**.
1198 May not be available on all supported platforms. **keep** is only available
1199 on Linux. If using ZFS on Solaris this cannot be set to **posix**
1200 because ZFS doesn't support pre-allocation. Default: **native** if any
1201 pre-allocation methods except **truncate** are available, **none** if not.
1203 Note that using **truncate** on Windows will interact surprisingly
1204 with non-sequential write patterns. When writing to a file that has
1205 been extended by setting the end-of-file information, Windows will
1206 backfill the unwritten portion of the file up to that offset with
1207 zeroes before issuing the new write. This means that a single small
1208 write to the end of an extended file will stall until the entire
1209 file has been filled with zeroes.
1211 .. option:: fadvise_hint=str
1213 Use :manpage:`posix_fadvise(2)` or :manpage:`posix_fadvise(2)` to
1214 advise the kernel on what I/O patterns are likely to be issued.
1215 Accepted values are:
1218 Backwards-compatible hint for "no hint".
1221 Backwards compatible hint for "advise with fio workload type". This
1222 uses **FADV_RANDOM** for a random workload, and **FADV_SEQUENTIAL**
1223 for a sequential workload.
1226 Advise using **FADV_SEQUENTIAL**.
1229 Advise using **FADV_RANDOM**.
1231 .. option:: write_hint=str
1233 Use :manpage:`fcntl(2)` to advise the kernel what life time to expect
1234 from a write. Only supported on Linux, as of version 4.13. Accepted
1238 No particular life time associated with this file.
1241 Data written to this file has a short life time.
1244 Data written to this file has a medium life time.
1247 Data written to this file has a long life time.
1250 Data written to this file has a very long life time.
1252 The values are all relative to each other, and no absolute meaning
1253 should be associated with them.
1255 .. option:: offset=int
1257 Start I/O at the provided offset in the file, given as either a fixed size in
1258 bytes or a percentage. If a percentage is given, the generated offset will be
1259 aligned to the minimum ``blocksize`` or to the value of ``offset_align`` if
1260 provided. Data before the given offset will not be touched. This
1261 effectively caps the file size at `real_size - offset`. Can be combined with
1262 :option:`size` to constrain the start and end range of the I/O workload.
1263 A percentage can be specified by a number between 1 and 100 followed by '%',
1264 for example, ``offset=20%`` to specify 20%.
1266 .. option:: offset_align=int
1268 If set to non-zero value, the byte offset generated by a percentage ``offset``
1269 is aligned upwards to this value. Defaults to 0 meaning that a percentage
1270 offset is aligned to the minimum block size.
1272 .. option:: offset_increment=int
1274 If this is provided, then the real offset becomes `offset + offset_increment
1275 * thread_number`, where the thread number is a counter that starts at 0 and
1276 is incremented for each sub-job (i.e. when :option:`numjobs` option is
1277 specified). This option is useful if there are several jobs which are
1278 intended to operate on a file in parallel disjoint segments, with even
1279 spacing between the starting points. Percentages can be used for this option.
1280 If a percentage is given, the generated offset will be aligned to the minimum
1281 ``blocksize`` or to the value of ``offset_align`` if provided.
1283 .. option:: number_ios=int
1285 Fio will normally perform I/Os until it has exhausted the size of the region
1286 set by :option:`size`, or if it exhaust the allocated time (or hits an error
1287 condition). With this setting, the range/size can be set independently of
1288 the number of I/Os to perform. When fio reaches this number, it will exit
1289 normally and report status. Note that this does not extend the amount of I/O
1290 that will be done, it will only stop fio if this condition is met before
1291 other end-of-job criteria.
1293 .. option:: fsync=int
1295 If writing to a file, issue an :manpage:`fsync(2)` (or its equivalent) of
1296 the dirty data for every number of blocks given. For example, if you give 32
1297 as a parameter, fio will sync the file after every 32 writes issued. If fio is
1298 using non-buffered I/O, we may not sync the file. The exception is the sg
1299 I/O engine, which synchronizes the disk cache anyway. Defaults to 0, which
1300 means fio does not periodically issue and wait for a sync to complete. Also
1301 see :option:`end_fsync` and :option:`fsync_on_close`.
1303 .. option:: fdatasync=int
1305 Like :option:`fsync` but uses :manpage:`fdatasync(2)` to only sync data and
1306 not metadata blocks. In Windows, FreeBSD, DragonFlyBSD or OSX there is no
1307 :manpage:`fdatasync(2)` so this falls back to using :manpage:`fsync(2)`.
1308 Defaults to 0, which means fio does not periodically issue and wait for a
1309 data-only sync to complete.
1311 .. option:: write_barrier=int
1313 Make every `N-th` write a barrier write.
1315 .. option:: sync_file_range=str:int
1317 Use :manpage:`sync_file_range(2)` for every `int` number of write
1318 operations. Fio will track range of writes that have happened since the last
1319 :manpage:`sync_file_range(2)` call. `str` can currently be one or more of:
1322 SYNC_FILE_RANGE_WAIT_BEFORE
1324 SYNC_FILE_RANGE_WRITE
1326 SYNC_FILE_RANGE_WAIT_AFTER
1328 So if you do ``sync_file_range=wait_before,write:8``, fio would use
1329 ``SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE`` for every 8
1330 writes. Also see the :manpage:`sync_file_range(2)` man page. This option is
1333 .. option:: overwrite=bool
1335 If true, writes to a file will always overwrite existing data. If the file
1336 doesn't already exist, it will be created before the write phase begins. If
1337 the file exists and is large enough for the specified write phase, nothing
1338 will be done. Default: false.
1340 .. option:: end_fsync=bool
1342 If true, :manpage:`fsync(2)` file contents when a write stage has completed.
1345 .. option:: fsync_on_close=bool
1347 If true, fio will :manpage:`fsync(2)` a dirty file on close. This differs
1348 from :option:`end_fsync` in that it will happen on every file close, not
1349 just at the end of the job. Default: false.
1351 .. option:: rwmixread=int
1353 Percentage of a mixed workload that should be reads. Default: 50.
1355 .. option:: rwmixwrite=int
1357 Percentage of a mixed workload that should be writes. If both
1358 :option:`rwmixread` and :option:`rwmixwrite` is given and the values do not
1359 add up to 100%, the latter of the two will be used to override the
1360 first. This may interfere with a given rate setting, if fio is asked to
1361 limit reads or writes to a certain rate. If that is the case, then the
1362 distribution may be skewed. Default: 50.
1364 .. option:: random_distribution=str:float[:float][,str:float][,str:float]
1366 By default, fio will use a completely uniform random distribution when asked
1367 to perform random I/O. Sometimes it is useful to skew the distribution in
1368 specific ways, ensuring that some parts of the data is more hot than others.
1369 fio includes the following distribution models:
1372 Uniform random distribution
1381 Normal (Gaussian) distribution
1384 Zoned random distribution
1387 Zone absolute random distribution
1389 When using a **zipf** or **pareto** distribution, an input value is also
1390 needed to define the access pattern. For **zipf**, this is the `Zipf
1391 theta`. For **pareto**, it's the `Pareto power`. Fio includes a test
1392 program, :command:`fio-genzipf`, that can be used visualize what the given input
1393 values will yield in terms of hit rates. If you wanted to use **zipf** with
1394 a `theta` of 1.2, you would use ``random_distribution=zipf:1.2`` as the
1395 option. If a non-uniform model is used, fio will disable use of the random
1396 map. For the **normal** distribution, a normal (Gaussian) deviation is
1397 supplied as a value between 0 and 100.
1399 The second, optional float is allowed for **pareto**, **zipf** and **normal** distributions.
1400 It allows to set base of distribution in non-default place, giving more control
1401 over most probable outcome. This value is in range [0-1] which maps linearly to
1402 range of possible random values.
1403 Defaults are: random for **pareto** and **zipf**, and 0.5 for **normal**.
1404 If you wanted to use **zipf** with a `theta` of 1.2 centered on 1/4 of allowed value range,
1405 you would use ``random_distibution=zipf:1.2:0.25``.
1407 For a **zoned** distribution, fio supports specifying percentages of I/O
1408 access that should fall within what range of the file or device. For
1409 example, given a criteria of:
1411 * 60% of accesses should be to the first 10%
1412 * 30% of accesses should be to the next 20%
1413 * 8% of accesses should be to the next 30%
1414 * 2% of accesses should be to the next 40%
1416 we can define that through zoning of the random accesses. For the above
1417 example, the user would do::
1419 random_distribution=zoned:60/10:30/20:8/30:2/40
1421 A **zoned_abs** distribution works exactly like the **zoned**, except
1422 that it takes absolute sizes. For example, let's say you wanted to
1423 define access according to the following criteria:
1425 * 60% of accesses should be to the first 20G
1426 * 30% of accesses should be to the next 100G
1427 * 10% of accesses should be to the next 500G
1429 we can define an absolute zoning distribution with:
1431 random_distribution=zoned_abs=60/20G:30/100G:10/500g
1433 For both **zoned** and **zoned_abs**, fio supports defining up to
1436 Similarly to how :option:`bssplit` works for setting ranges and
1437 percentages of block sizes. Like :option:`bssplit`, it's possible to
1438 specify separate zones for reads, writes, and trims. If just one set
1439 is given, it'll apply to all of them. This goes for both **zoned**
1440 **zoned_abs** distributions.
1442 .. option:: percentage_random=int[,int][,int]
1444 For a random workload, set how big a percentage should be random. This
1445 defaults to 100%, in which case the workload is fully random. It can be set
1446 from anywhere from 0 to 100. Setting it to 0 would make the workload fully
1447 sequential. Any setting in between will result in a random mix of sequential
1448 and random I/O, at the given percentages. Comma-separated values may be
1449 specified for reads, writes, and trims as described in :option:`blocksize`.
1451 .. option:: norandommap
1453 Normally fio will cover every block of the file when doing random I/O. If
1454 this option is given, fio will just get a new random offset without looking
1455 at past I/O history. This means that some blocks may not be read or written,
1456 and that some blocks may be read/written more than once. If this option is
1457 used with :option:`verify` and multiple blocksizes (via :option:`bsrange`),
1458 only intact blocks are verified, i.e., partially-overwritten blocks are
1459 ignored. With an async I/O engine and an I/O depth > 1, it is possible for
1460 the same block to be overwritten, which can cause verification errors. Either
1461 do not use norandommap in this case, or also use the lfsr random generator.
1463 .. option:: softrandommap=bool
1465 See :option:`norandommap`. If fio runs with the random block map enabled and
1466 it fails to allocate the map, if this option is set it will continue without
1467 a random block map. As coverage will not be as complete as with random maps,
1468 this option is disabled by default.
1470 .. option:: random_generator=str
1472 Fio supports the following engines for generating I/O offsets for random I/O:
1475 Strong 2^88 cycle random number generator.
1477 Linear feedback shift register generator.
1479 Strong 64-bit 2^258 cycle random number generator.
1481 **tausworthe** is a strong random number generator, but it requires tracking
1482 on the side if we want to ensure that blocks are only read or written
1483 once. **lfsr** guarantees that we never generate the same offset twice, and
1484 it's also less computationally expensive. It's not a true random generator,
1485 however, though for I/O purposes it's typically good enough. **lfsr** only
1486 works with single block sizes, not with workloads that use multiple block
1487 sizes. If used with such a workload, fio may read or write some blocks
1488 multiple times. The default value is **tausworthe**, unless the required
1489 space exceeds 2^32 blocks. If it does, then **tausworthe64** is
1490 selected automatically.
1496 .. option:: blocksize=int[,int][,int], bs=int[,int][,int]
1498 The block size in bytes used for I/O units. Default: 4096. A single value
1499 applies to reads, writes, and trims. Comma-separated values may be
1500 specified for reads, writes, and trims. A value not terminated in a comma
1501 applies to subsequent types.
1506 means 256k for reads, writes and trims.
1509 means 8k for reads, 32k for writes and trims.
1512 means 8k for reads, 32k for writes, and default for trims.
1515 means default for reads, 8k for writes and trims.
1518 means default for reads, 8k for writes, and default for trims.
1520 .. option:: blocksize_range=irange[,irange][,irange], bsrange=irange[,irange][,irange]
1522 A range of block sizes in bytes for I/O units. The issued I/O unit will
1523 always be a multiple of the minimum size, unless
1524 :option:`blocksize_unaligned` is set.
1526 Comma-separated ranges may be specified for reads, writes, and trims as
1527 described in :option:`blocksize`.
1529 Example: ``bsrange=1k-4k,2k-8k``.
1531 .. option:: bssplit=str[,str][,str]
1533 Sometimes you want even finer grained control of the block sizes
1534 issued, not just an even split between them. This option allows you to
1535 weight various block sizes, so that you are able to define a specific
1536 amount of block sizes issued. The format for this option is::
1538 bssplit=blocksize/percentage:blocksize/percentage
1540 for as many block sizes as needed. So if you want to define a workload
1541 that has 50% 64k blocks, 10% 4k blocks, and 40% 32k blocks, you would
1544 bssplit=4k/10:64k/50:32k/40
1546 Ordering does not matter. If the percentage is left blank, fio will
1547 fill in the remaining values evenly. So a bssplit option like this one::
1549 bssplit=4k/50:1k/:32k/
1551 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages always
1552 add up to 100, if bssplit is given a range that adds up to more, it
1555 Comma-separated values may be specified for reads, writes, and trims as
1556 described in :option:`blocksize`.
1558 If you want a workload that has 50% 2k reads and 50% 4k reads, while
1559 having 90% 4k writes and 10% 8k writes, you would specify::
1561 bssplit=2k/50:4k/50,4k/90:8k/10
1563 Fio supports defining up to 64 different weights for each data
1566 .. option:: blocksize_unaligned, bs_unaligned
1568 If set, fio will issue I/O units with any size within
1569 :option:`blocksize_range`, not just multiples of the minimum size. This
1570 typically won't work with direct I/O, as that normally requires sector
1573 .. option:: bs_is_seq_rand=bool
1575 If this option is set, fio will use the normal read,write blocksize settings
1576 as sequential,random blocksize settings instead. Any random read or write
1577 will use the WRITE blocksize settings, and any sequential read or write will
1578 use the READ blocksize settings.
1580 .. option:: blockalign=int[,int][,int], ba=int[,int][,int]
1582 Boundary to which fio will align random I/O units. Default:
1583 :option:`blocksize`. Minimum alignment is typically 512b for using direct
1584 I/O, though it usually depends on the hardware block size. This option is
1585 mutually exclusive with using a random map for files, so it will turn off
1586 that option. Comma-separated values may be specified for reads, writes, and
1587 trims as described in :option:`blocksize`.
1593 .. option:: zero_buffers
1595 Initialize buffers with all zeros. Default: fill buffers with random data.
1597 .. option:: refill_buffers
1599 If this option is given, fio will refill the I/O buffers on every
1600 submit. Only makes sense if :option:`zero_buffers` isn't specified,
1601 naturally. Defaults to being unset i.e., the buffer is only filled at
1602 init time and the data in it is reused when possible but if any of
1603 :option:`verify`, :option:`buffer_compress_percentage` or
1604 :option:`dedupe_percentage` are enabled then `refill_buffers` is also
1605 automatically enabled.
1607 .. option:: scramble_buffers=bool
1609 If :option:`refill_buffers` is too costly and the target is using data
1610 deduplication, then setting this option will slightly modify the I/O buffer
1611 contents to defeat normal de-dupe attempts. This is not enough to defeat
1612 more clever block compression attempts, but it will stop naive dedupe of
1613 blocks. Default: true.
1615 .. option:: buffer_compress_percentage=int
1617 If this is set, then fio will attempt to provide I/O buffer content
1618 (on WRITEs) that compresses to the specified level. Fio does this by
1619 providing a mix of random data followed by fixed pattern data. The
1620 fixed pattern is either zeros, or the pattern specified by
1621 :option:`buffer_pattern`. If the `buffer_pattern` option is used, it
1622 might skew the compression ratio slightly. Setting
1623 `buffer_compress_percentage` to a value other than 100 will also
1624 enable :option:`refill_buffers` in order to reduce the likelihood that
1625 adjacent blocks are so similar that they over compress when seen
1626 together. See :option:`buffer_compress_chunk` for how to set a finer or
1627 coarser granularity for the random/fixed data region. Defaults to unset
1628 i.e., buffer data will not adhere to any compression level.
1630 .. option:: buffer_compress_chunk=int
1632 This setting allows fio to manage how big the random/fixed data region
1633 is when using :option:`buffer_compress_percentage`. When
1634 `buffer_compress_chunk` is set to some non-zero value smaller than the
1635 block size, fio can repeat the random/fixed region throughout the I/O
1636 buffer at the specified interval (which particularly useful when
1637 bigger block sizes are used for a job). When set to 0, fio will use a
1638 chunk size that matches the block size resulting in a single
1639 random/fixed region within the I/O buffer. Defaults to 512. When the
1640 unit is omitted, the value is interpreted in bytes.
1642 .. option:: buffer_pattern=str
1644 If set, fio will fill the I/O buffers with this pattern or with the contents
1645 of a file. If not set, the contents of I/O buffers are defined by the other
1646 options related to buffer contents. The setting can be any pattern of bytes,
1647 and can be prefixed with 0x for hex values. It may also be a string, where
1648 the string must then be wrapped with ``""``. Or it may also be a filename,
1649 where the filename must be wrapped with ``''`` in which case the file is
1650 opened and read. Note that not all the file contents will be read if that
1651 would cause the buffers to overflow. So, for example::
1653 buffer_pattern='filename'
1657 buffer_pattern="abcd"
1665 buffer_pattern=0xdeadface
1667 Also you can combine everything together in any order::
1669 buffer_pattern=0xdeadface"abcd"-12'filename'
1671 .. option:: dedupe_percentage=int
1673 If set, fio will generate this percentage of identical buffers when
1674 writing. These buffers will be naturally dedupable. The contents of the
1675 buffers depend on what other buffer compression settings have been set. It's
1676 possible to have the individual buffers either fully compressible, or not at
1677 all -- this option only controls the distribution of unique buffers. Setting
1678 this option will also enable :option:`refill_buffers` to prevent every buffer
1681 .. option:: invalidate=bool
1683 Invalidate the buffer/page cache parts of the files to be used prior to
1684 starting I/O if the platform and file type support it. Defaults to true.
1685 This will be ignored if :option:`pre_read` is also specified for the
1688 .. option:: sync=str
1690 Whether, and what type, of synchronous I/O to use for writes. The allowed
1694 Do not use synchronous IO, the default.
1700 Use synchronous file IO. For the majority of I/O engines,
1701 this means using O_SYNC.
1707 Use synchronous data IO. For the majority of I/O engines,
1708 this means using O_DSYNC.
1711 .. option:: iomem=str, mem=str
1713 Fio can use various types of memory as the I/O unit buffer. The allowed
1717 Use memory from :manpage:`malloc(3)` as the buffers. Default memory
1721 Use shared memory as the buffers. Allocated through
1722 :manpage:`shmget(2)`.
1725 Same as shm, but use huge pages as backing.
1728 Use :manpage:`mmap(2)` to allocate buffers. May either be anonymous memory, or can
1729 be file backed if a filename is given after the option. The format
1730 is `mem=mmap:/path/to/file`.
1733 Use a memory mapped huge file as the buffer backing. Append filename
1734 after mmaphuge, ala `mem=mmaphuge:/hugetlbfs/file`.
1737 Same as mmap, but use a MMAP_SHARED mapping.
1740 Use GPU memory as the buffers for GPUDirect RDMA benchmark.
1741 The :option:`ioengine` must be `rdma`.
1743 The area allocated is a function of the maximum allowed bs size for the job,
1744 multiplied by the I/O depth given. Note that for **shmhuge** and
1745 **mmaphuge** to work, the system must have free huge pages allocated. This
1746 can normally be checked and set by reading/writing
1747 :file:`/proc/sys/vm/nr_hugepages` on a Linux system. Fio assumes a huge page
1748 is 4MiB in size. So to calculate the number of huge pages you need for a
1749 given job file, add up the I/O depth of all jobs (normally one unless
1750 :option:`iodepth` is used) and multiply by the maximum bs set. Then divide
1751 that number by the huge page size. You can see the size of the huge pages in
1752 :file:`/proc/meminfo`. If no huge pages are allocated by having a non-zero
1753 number in `nr_hugepages`, using **mmaphuge** or **shmhuge** will fail. Also
1754 see :option:`hugepage-size`.
1756 **mmaphuge** also needs to have hugetlbfs mounted and the file location
1757 should point there. So if it's mounted in :file:`/huge`, you would use
1758 `mem=mmaphuge:/huge/somefile`.
1760 .. option:: iomem_align=int, mem_align=int
1762 This indicates the memory alignment of the I/O memory buffers. Note that
1763 the given alignment is applied to the first I/O unit buffer, if using
1764 :option:`iodepth` the alignment of the following buffers are given by the
1765 :option:`bs` used. In other words, if using a :option:`bs` that is a
1766 multiple of the page sized in the system, all buffers will be aligned to
1767 this value. If using a :option:`bs` that is not page aligned, the alignment
1768 of subsequent I/O memory buffers is the sum of the :option:`iomem_align` and
1771 .. option:: hugepage-size=int
1773 Defines the size of a huge page. Must at least be equal to the system
1774 setting, see :file:`/proc/meminfo`. Defaults to 4MiB. Should probably
1775 always be a multiple of megabytes, so using ``hugepage-size=Xm`` is the
1776 preferred way to set this to avoid setting a non-pow-2 bad value.
1778 .. option:: lockmem=int
1780 Pin the specified amount of memory with :manpage:`mlock(2)`. Can be used to
1781 simulate a smaller amount of memory. The amount specified is per worker.
1787 .. option:: size=int
1789 The total size of file I/O for each thread of this job. Fio will run until
1790 this many bytes has been transferred, unless runtime is limited by other options
1791 (such as :option:`runtime`, for instance, or increased/decreased by :option:`io_size`).
1792 Fio will divide this size between the available files determined by options
1793 such as :option:`nrfiles`, :option:`filename`, unless :option:`filesize` is
1794 specified by the job. If the result of division happens to be 0, the size is
1795 set to the physical size of the given files or devices if they exist.
1796 If this option is not specified, fio will use the full size of the given
1797 files or devices. If the files do not exist, size must be given. It is also
1798 possible to give size as a percentage between 1 and 100. If ``size=20%`` is
1799 given, fio will use 20% of the full size of the given files or devices.
1800 Can be combined with :option:`offset` to constrain the start and end range
1801 that I/O will be done within.
1803 .. option:: io_size=int, io_limit=int
1805 Normally fio operates within the region set by :option:`size`, which means
1806 that the :option:`size` option sets both the region and size of I/O to be
1807 performed. Sometimes that is not what you want. With this option, it is
1808 possible to define just the amount of I/O that fio should do. For instance,
1809 if :option:`size` is set to 20GiB and :option:`io_size` is set to 5GiB, fio
1810 will perform I/O within the first 20GiB but exit when 5GiB have been
1811 done. The opposite is also possible -- if :option:`size` is set to 20GiB,
1812 and :option:`io_size` is set to 40GiB, then fio will do 40GiB of I/O within
1813 the 0..20GiB region.
1815 .. option:: filesize=irange(int)
1817 Individual file sizes. May be a range, in which case fio will select sizes
1818 for files at random within the given range and limited to :option:`size` in
1819 total (if that is given). If not given, each created file is the same size.
1820 This option overrides :option:`size` in terms of file size, which means
1821 this value is used as a fixed size or possible range of each file.
1823 .. option:: file_append=bool
1825 Perform I/O after the end of the file. Normally fio will operate within the
1826 size of a file. If this option is set, then fio will append to the file
1827 instead. This has identical behavior to setting :option:`offset` to the size
1828 of a file. This option is ignored on non-regular files.
1830 .. option:: fill_device=bool, fill_fs=bool
1832 Sets size to something really large and waits for ENOSPC (no space left on
1833 device) as the terminating condition. Only makes sense with sequential
1834 write. For a read workload, the mount point will be filled first then I/O
1835 started on the result. This option doesn't make sense if operating on a raw
1836 device node, since the size of that is already known by the file system.
1837 Additionally, writing beyond end-of-device will not return ENOSPC there.
1843 .. option:: ioengine=str
1845 Defines how the job issues I/O to the file. The following types are defined:
1848 Basic :manpage:`read(2)` or :manpage:`write(2)`
1849 I/O. :manpage:`lseek(2)` is used to position the I/O location.
1850 See :option:`fsync` and :option:`fdatasync` for syncing write I/Os.
1853 Basic :manpage:`pread(2)` or :manpage:`pwrite(2)` I/O. Default on
1854 all supported operating systems except for Windows.
1857 Basic :manpage:`readv(2)` or :manpage:`writev(2)` I/O. Will emulate
1858 queuing by coalescing adjacent I/Os into a single submission.
1861 Basic :manpage:`preadv(2)` or :manpage:`pwritev(2)` I/O.
1864 Basic :manpage:`preadv2(2)` or :manpage:`pwritev2(2)` I/O.
1867 Fast Linux native asynchronous I/O. Supports async IO
1868 for both direct and buffered IO.
1869 This engine defines engine specific options.
1872 Linux native asynchronous I/O. Note that Linux may only support
1873 queued behavior with non-buffered I/O (set ``direct=1`` or
1875 This engine defines engine specific options.
1878 POSIX asynchronous I/O using :manpage:`aio_read(3)` and
1879 :manpage:`aio_write(3)`.
1882 Solaris native asynchronous I/O.
1885 Windows native asynchronous I/O. Default on Windows.
1888 File is memory mapped with :manpage:`mmap(2)` and data copied
1889 to/from using :manpage:`memcpy(3)`.
1892 :manpage:`splice(2)` is used to transfer the data and
1893 :manpage:`vmsplice(2)` to transfer data from user space to the
1897 SCSI generic sg v3 I/O. May either be synchronous using the SG_IO
1898 ioctl, or if the target is an sg character device we use
1899 :manpage:`read(2)` and :manpage:`write(2)` for asynchronous
1900 I/O. Requires :option:`filename` option to specify either block or
1901 character devices. This engine supports trim operations.
1902 The sg engine includes engine specific options.
1905 Doesn't transfer any data, just pretends to. This is mainly used to
1906 exercise fio itself and for debugging/testing purposes.
1909 Transfer over the network to given ``host:port``. Depending on the
1910 :option:`protocol` used, the :option:`hostname`, :option:`port`,
1911 :option:`listen` and :option:`filename` options are used to specify
1912 what sort of connection to make, while the :option:`protocol` option
1913 determines which protocol will be used. This engine defines engine
1917 Like **net**, but uses :manpage:`splice(2)` and
1918 :manpage:`vmsplice(2)` to map data and send/receive.
1919 This engine defines engine specific options.
1922 Doesn't transfer any data, but burns CPU cycles according to the
1923 :option:`cpuload`, :option:`cpuchunks` and :option:`cpumode` options.
1924 Setting :option:`cpuload`\=85 will cause that job to do nothing but burn 85%
1925 of the CPU. In case of SMP machines, use :option:`numjobs`\=<nr_of_cpu>
1926 to get desired CPU usage, as the cpuload only loads a
1927 single CPU at the desired rate. A job never finishes unless there is
1928 at least one non-cpuio job.
1929 Setting :option:`cpumode`\=qsort replace the default noop instructions loop
1930 by a qsort algorithm to consume more energy.
1933 The RDMA I/O engine supports both RDMA memory semantics
1934 (RDMA_WRITE/RDMA_READ) and channel semantics (Send/Recv) for the
1935 InfiniBand, RoCE and iWARP protocols. This engine defines engine
1939 I/O engine that does regular fallocate to simulate data transfer as
1943 does fallocate(,mode = FALLOC_FL_KEEP_SIZE,).
1946 does fallocate(,mode = 0).
1949 does fallocate(,mode = FALLOC_FL_KEEP_SIZE|FALLOC_FL_PUNCH_HOLE).
1952 I/O engine that sends :manpage:`ftruncate(2)` operations in response
1953 to write (DDIR_WRITE) events. Each ftruncate issued sets the file's
1954 size to the current block offset. :option:`blocksize` is ignored.
1957 I/O engine that does regular EXT4_IOC_MOVE_EXT ioctls to simulate
1958 defragment activity in request to DDIR_WRITE event.
1961 I/O engine supporting direct access to Ceph Reliable Autonomic
1962 Distributed Object Store (RADOS) via librados. This ioengine
1963 defines engine specific options.
1966 I/O engine supporting direct access to Ceph Rados Block Devices
1967 (RBD) via librbd without the need to use the kernel rbd driver. This
1968 ioengine defines engine specific options.
1971 I/O engine supporting GET/PUT requests over HTTP(S) with libcurl to
1972 a WebDAV or S3 endpoint. This ioengine defines engine specific options.
1974 This engine only supports direct IO of iodepth=1; you need to scale this
1975 via numjobs. blocksize defines the size of the objects to be created.
1977 TRIM is translated to object deletion.
1980 Using GlusterFS libgfapi sync interface to direct access to
1981 GlusterFS volumes without having to go through FUSE. This ioengine
1982 defines engine specific options.
1985 Using GlusterFS libgfapi async interface to direct access to
1986 GlusterFS volumes without having to go through FUSE. This ioengine
1987 defines engine specific options.
1990 Read and write through Hadoop (HDFS). The :option:`filename` option
1991 is used to specify host,port of the hdfs name-node to connect. This
1992 engine interprets offsets a little differently. In HDFS, files once
1993 created cannot be modified so random writes are not possible. To
1994 imitate this the libhdfs engine expects a bunch of small files to be
1995 created over HDFS and will randomly pick a file from them
1996 based on the offset generated by fio backend (see the example
1997 job file to create such files, use ``rw=write`` option). Please
1998 note, it may be necessary to set environment variables to work
1999 with HDFS/libhdfs properly. Each job uses its own connection to
2003 Read, write and erase an MTD character device (e.g.,
2004 :file:`/dev/mtd0`). Discards are treated as erases. Depending on the
2005 underlying device type, the I/O may have to go in a certain pattern,
2006 e.g., on NAND, writing sequentially to erase blocks and discarding
2007 before overwriting. The `trimwrite` mode works well for this
2011 Read and write using filesystem DAX to a file on a filesystem
2012 mounted with DAX on a persistent memory device through the PMDK
2016 Read and write using device DAX to a persistent memory device (e.g.,
2017 /dev/dax0.0) through the PMDK libpmem library.
2020 Prefix to specify loading an external I/O engine object file. Append
2021 the engine filename, e.g. ``ioengine=external:/tmp/foo.o`` to load
2022 ioengine :file:`foo.o` in :file:`/tmp`. The path can be either
2023 absolute or relative. See :file:`engines/skeleton_external.c` for
2024 details of writing an external I/O engine.
2027 Simply create the files and do no I/O to them. You still need to
2028 set `filesize` so that all the accounting still occurs, but no
2029 actual I/O will be done other than creating the file.
2032 Simply do stat() and do no I/O to the file. You need to set 'filesize'
2033 and 'nrfiles', so that files will be created.
2034 This engine is to measure file lookup and meta data access.
2037 Read and write using mmap I/O to a file on a filesystem
2038 mounted with DAX on a persistent memory device through the PMDK
2042 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2043 This engine is very basic and issues calls to IME whenever an IO is
2047 Synchronous read and write using DDN's Infinite Memory Engine (IME).
2048 This engine uses iovecs and will try to stack as much IOs as possible
2049 (if the IOs are "contiguous" and the IO depth is not exceeded)
2050 before issuing a call to IME.
2053 Asynchronous read and write using DDN's Infinite Memory Engine (IME).
2054 This engine will try to stack as much IOs as possible by creating
2055 requests for IME. FIO will then decide when to commit these requests.
2057 Read and write iscsi lun with libiscsi.
2059 Read and write a Network Block Device (NBD).
2062 I/O engine supporting libcufile synchronous access to nvidia-fs and a
2063 GPUDirect Storage-supported filesystem. This engine performs
2064 I/O without transferring buffers between user-space and the kernel,
2065 unless :option:`verify` is set or :option:`cuda_io` is `posix`.
2066 :option:`iomem` must not be `cudamalloc`. This ioengine defines
2067 engine specific options.
2069 I/O engine specific parameters
2070 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2072 In addition, there are some parameters which are only valid when a specific
2073 :option:`ioengine` is in use. These are used identically to normal parameters,
2074 with the caveat that when used on the command line, they must come after the
2075 :option:`ioengine` that defines them is selected.
2077 .. option:: cmdprio_percentage=int : [io_uring] [libaio]
2079 Set the percentage of I/O that will be issued with higher priority by setting
2080 the priority bit. Non-read I/O is likely unaffected by ``cmdprio_percentage``.
2081 This option cannot be used with the `prio` or `prioclass` options. For this
2082 option to set the priority bit properly, NCQ priority must be supported and
2083 enabled and :option:`direct`\=1 option must be used. fio must also be run as
2086 .. option:: fixedbufs : [io_uring]
2088 If fio is asked to do direct IO, then Linux will map pages for each
2089 IO call, and release them when IO is done. If this option is set, the
2090 pages are pre-mapped before IO is started. This eliminates the need to
2091 map and release for each IO. This is more efficient, and reduces the
2094 .. option:: hipri : [io_uring]
2096 If this option is set, fio will attempt to use polled IO completions.
2097 Normal IO completions generate interrupts to signal the completion of
2098 IO, polled completions do not. Hence they are require active reaping
2099 by the application. The benefits are more efficient IO for high IOPS
2100 scenarios, and lower latencies for low queue depth IO.
2102 .. option:: registerfiles : [io_uring]
2104 With this option, fio registers the set of files being used with the
2105 kernel. This avoids the overhead of managing file counts in the kernel,
2106 making the submission and completion part more lightweight. Required
2107 for the below :option:`sqthread_poll` option.
2109 .. option:: sqthread_poll : [io_uring]
2111 Normally fio will submit IO by issuing a system call to notify the
2112 kernel of available items in the SQ ring. If this option is set, the
2113 act of submitting IO will be done by a polling thread in the kernel.
2114 This frees up cycles for fio, at the cost of using more CPU in the
2117 .. option:: sqthread_poll_cpu : [io_uring]
2119 When :option:`sqthread_poll` is set, this option provides a way to
2120 define which CPU should be used for the polling thread.
2122 .. option:: userspace_reap : [libaio]
2124 Normally, with the libaio engine in use, fio will use the
2125 :manpage:`io_getevents(2)` system call to reap newly returned events. With
2126 this flag turned on, the AIO ring will be read directly from user-space to
2127 reap events. The reaping mode is only enabled when polling for a minimum of
2128 0 events (e.g. when :option:`iodepth_batch_complete` `=0`).
2130 .. option:: hipri : [pvsync2]
2132 Set RWF_HIPRI on I/O, indicating to the kernel that it's of higher priority
2135 .. option:: hipri_percentage : [pvsync2]
2137 When hipri is set this determines the probability of a pvsync2 I/O being high
2138 priority. The default is 100%.
2140 .. option:: nowait : [pvsync2] [libaio] [io_uring]
2142 By default if a request cannot be executed immediately (e.g. resource starvation,
2143 waiting on locks) it is queued and the initiating process will be blocked until
2144 the required resource becomes free.
2146 This option sets the RWF_NOWAIT flag (supported from the 4.14 Linux kernel) and
2147 the call will return instantly with EAGAIN or a partial result rather than waiting.
2149 It is useful to also use ignore_error=EAGAIN when using this option.
2151 Note: glibc 2.27, 2.28 have a bug in syscall wrappers preadv2, pwritev2.
2152 They return EOPNOTSUP instead of EAGAIN.
2154 For cached I/O, using this option usually means a request operates only with
2155 cached data. Currently the RWF_NOWAIT flag does not supported for cached write.
2157 For direct I/O, requests will only succeed if cache invalidation isn't required,
2158 file blocks are fully allocated and the disk request could be issued immediately.
2160 .. option:: cpuload=int : [cpuio]
2162 Attempt to use the specified percentage of CPU cycles. This is a mandatory
2163 option when using cpuio I/O engine.
2165 .. option:: cpuchunks=int : [cpuio]
2167 Split the load into cycles of the given time. In microseconds.
2169 .. option:: exit_on_io_done=bool : [cpuio]
2171 Detect when I/O threads are done, then exit.
2173 .. option:: namenode=str : [libhdfs]
2175 The hostname or IP address of a HDFS cluster namenode to contact.
2177 .. option:: port=int
2181 The listening port of the HFDS cluster namenode.
2185 The TCP or UDP port to bind to or connect to. If this is used with
2186 :option:`numjobs` to spawn multiple instances of the same job type, then
2187 this will be the starting port number since fio will use a range of
2192 The port to use for RDMA-CM communication. This should be the same value
2193 on the client and the server side.
2195 .. option:: hostname=str : [netsplice] [net] [rdma]
2197 The hostname or IP address to use for TCP, UDP or RDMA-CM based I/O. If the job
2198 is a TCP listener or UDP reader, the hostname is not used and must be omitted
2199 unless it is a valid UDP multicast address.
2201 .. option:: interface=str : [netsplice] [net]
2203 The IP address of the network interface used to send or receive UDP
2206 .. option:: ttl=int : [netsplice] [net]
2208 Time-to-live value for outgoing UDP multicast packets. Default: 1.
2210 .. option:: nodelay=bool : [netsplice] [net]
2212 Set TCP_NODELAY on TCP connections.
2214 .. option:: protocol=str, proto=str : [netsplice] [net]
2216 The network protocol to use. Accepted values are:
2219 Transmission control protocol.
2221 Transmission control protocol V6.
2223 User datagram protocol.
2225 User datagram protocol V6.
2229 When the protocol is TCP or UDP, the port must also be given, as well as the
2230 hostname if the job is a TCP listener or UDP reader. For unix sockets, the
2231 normal :option:`filename` option should be used and the port is invalid.
2233 .. option:: listen : [netsplice] [net]
2235 For TCP network connections, tell fio to listen for incoming connections
2236 rather than initiating an outgoing connection. The :option:`hostname` must
2237 be omitted if this option is used.
2239 .. option:: pingpong : [netsplice] [net]
2241 Normally a network writer will just continue writing data, and a network
2242 reader will just consume packages. If ``pingpong=1`` is set, a writer will
2243 send its normal payload to the reader, then wait for the reader to send the
2244 same payload back. This allows fio to measure network latencies. The
2245 submission and completion latencies then measure local time spent sending or
2246 receiving, and the completion latency measures how long it took for the
2247 other end to receive and send back. For UDP multicast traffic
2248 ``pingpong=1`` should only be set for a single reader when multiple readers
2249 are listening to the same address.
2251 .. option:: window_size : [netsplice] [net]
2253 Set the desired socket buffer size for the connection.
2255 .. option:: mss : [netsplice] [net]
2257 Set the TCP maximum segment size (TCP_MAXSEG).
2259 .. option:: donorname=str : [e4defrag]
2261 File will be used as a block donor (swap extents between files).
2263 .. option:: inplace=int : [e4defrag]
2265 Configure donor file blocks allocation strategy:
2268 Default. Preallocate donor's file on init.
2270 Allocate space immediately inside defragment event, and free right
2273 .. option:: clustername=str : [rbd,rados]
2275 Specifies the name of the Ceph cluster.
2277 .. option:: rbdname=str : [rbd]
2279 Specifies the name of the RBD.
2281 .. option:: pool=str : [rbd,rados]
2283 Specifies the name of the Ceph pool containing RBD or RADOS data.
2285 .. option:: clientname=str : [rbd,rados]
2287 Specifies the username (without the 'client.' prefix) used to access the
2288 Ceph cluster. If the *clustername* is specified, the *clientname* shall be
2289 the full *type.id* string. If no type. prefix is given, fio will add
2290 'client.' by default.
2292 .. option:: busy_poll=bool : [rbd,rados]
2294 Poll store instead of waiting for completion. Usually this provides better
2295 throughput at cost of higher(up to 100%) CPU utilization.
2297 .. option:: skip_bad=bool : [mtd]
2299 Skip operations against known bad blocks.
2301 .. option:: hdfsdirectory : [libhdfs]
2303 libhdfs will create chunk in this HDFS directory.
2305 .. option:: chunk_size : [libhdfs]
2307 The size of the chunk to use for each file.
2309 .. option:: verb=str : [rdma]
2311 The RDMA verb to use on this side of the RDMA ioengine connection. Valid
2312 values are write, read, send and recv. These correspond to the equivalent
2313 RDMA verbs (e.g. write = rdma_write etc.). Note that this only needs to be
2314 specified on the client side of the connection. See the examples folder.
2316 .. option:: bindname=str : [rdma]
2318 The name to use to bind the local RDMA-CM connection to a local RDMA device.
2319 This could be a hostname or an IPv4 or IPv6 address. On the server side this
2320 will be passed into the rdma_bind_addr() function and on the client site it
2321 will be used in the rdma_resolve_add() function. This can be useful when
2322 multiple paths exist between the client and the server or in certain loopback
2325 .. option:: stat_type=str : [filestat]
2327 Specify stat system call type to measure lookup/getattr performance.
2328 Default is **stat** for :manpage:`stat(2)`.
2330 .. option:: readfua=bool : [sg]
2332 With readfua option set to 1, read operations include
2333 the force unit access (fua) flag. Default is 0.
2335 .. option:: writefua=bool : [sg]
2337 With writefua option set to 1, write operations include
2338 the force unit access (fua) flag. Default is 0.
2340 .. option:: sg_write_mode=str : [sg]
2342 Specify the type of write commands to issue. This option can take three values:
2345 This is the default where write opcodes are issued as usual.
2347 Issue WRITE AND VERIFY commands. The BYTCHK bit is set to 0. This
2348 directs the device to carry out a medium verification with no data
2349 comparison. The writefua option is ignored with this selection.
2351 Issue WRITE SAME commands. This transfers a single block to the device
2352 and writes this same block of data to a contiguous sequence of LBAs
2353 beginning at the specified offset. fio's block size parameter specifies
2354 the amount of data written with each command. However, the amount of data
2355 actually transferred to the device is equal to the device's block
2356 (sector) size. For a device with 512 byte sectors, blocksize=8k will
2357 write 16 sectors with each command. fio will still generate 8k of data
2358 for each command but only the first 512 bytes will be used and
2359 transferred to the device. The writefua option is ignored with this
2362 .. option:: hipri : [sg]
2364 If this option is set, fio will attempt to use polled IO completions.
2365 This will have a similar effect as (io_uring)hipri. Only SCSI READ and
2366 WRITE commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor
2367 VERIFY). Older versions of the Linux sg driver that do not support
2368 hipri will simply ignore this flag and do normal IO. The Linux SCSI
2369 Low Level Driver (LLD) that "owns" the device also needs to support
2370 hipri (also known as iopoll and mq_poll). The MegaRAID driver is an
2371 example of a SCSI LLD. Default: clear (0) which does normal
2372 (interrupted based) IO.
2374 .. option:: http_host=str : [http]
2376 Hostname to connect to. For S3, this could be the bucket hostname.
2377 Default is **localhost**
2379 .. option:: http_user=str : [http]
2381 Username for HTTP authentication.
2383 .. option:: http_pass=str : [http]
2385 Password for HTTP authentication.
2387 .. option:: https=str : [http]
2389 Enable HTTPS instead of http. *on* enables HTTPS; *insecure*
2390 will enable HTTPS, but disable SSL peer verification (use with
2391 caution!). Default is **off**
2393 .. option:: http_mode=str : [http]
2395 Which HTTP access mode to use: *webdav*, *swift*, or *s3*.
2396 Default is **webdav**
2398 .. option:: http_s3_region=str : [http]
2400 The S3 region/zone string.
2401 Default is **us-east-1**
2403 .. option:: http_s3_key=str : [http]
2407 .. option:: http_s3_keyid=str : [http]
2409 The S3 key/access id.
2411 .. option:: http_swift_auth_token=str : [http]
2413 The Swift auth token. See the example configuration file on how
2416 .. option:: http_verbose=int : [http]
2418 Enable verbose requests from libcurl. Useful for debugging. 1
2419 turns on verbose logging from libcurl, 2 additionally enables
2420 HTTP IO tracing. Default is **0**
2422 .. option:: uri=str : [nbd]
2424 Specify the NBD URI of the server to test. The string
2425 is a standard NBD URI
2426 (see https://github.com/NetworkBlockDevice/nbd/tree/master/doc).
2427 Example URIs: nbd://localhost:10809
2428 nbd+unix:///?socket=/tmp/socket
2429 nbds://tlshost/exportname
2431 .. option:: gpu_dev_ids=str : [libcufile]
2433 Specify the GPU IDs to use with CUDA. This is a colon-separated list of
2434 int. GPUs are assigned to workers roundrobin. Default is 0.
2436 .. option:: cuda_io=str : [libcufile]
2438 Specify the type of I/O to use with CUDA. Default is **cufile**.
2441 Use libcufile and nvidia-fs. This option performs I/O directly
2442 between a GPUDirect Storage filesystem and GPU buffers,
2443 avoiding use of a bounce buffer. If :option:`verify` is set,
2444 cudaMemcpy is used to copy verificaton data between RAM and GPU.
2445 Verification data is copied from RAM to GPU before a write
2446 and from GPU to RAM after a read. :option:`direct` must be 1.
2448 Use POSIX to perform I/O with a RAM buffer, and use cudaMemcpy
2449 to transfer data between RAM and the GPUs. Data is copied from
2450 GPU to RAM before a write and copied from RAM to GPU after a
2451 read. :option:`verify` does not affect use of cudaMemcpy.
2456 .. option:: iodepth=int
2458 Number of I/O units to keep in flight against the file. Note that
2459 increasing *iodepth* beyond 1 will not affect synchronous ioengines (except
2460 for small degrees when :option:`verify_async` is in use). Even async
2461 engines may impose OS restrictions causing the desired depth not to be
2462 achieved. This may happen on Linux when using libaio and not setting
2463 :option:`direct`\=1, since buffered I/O is not async on that OS. Keep an
2464 eye on the I/O depth distribution in the fio output to verify that the
2465 achieved depth is as expected. Default: 1.
2467 .. option:: iodepth_batch_submit=int, iodepth_batch=int
2469 This defines how many pieces of I/O to submit at once. It defaults to 1
2470 which means that we submit each I/O as soon as it is available, but can be
2471 raised to submit bigger batches of I/O at the time. If it is set to 0 the
2472 :option:`iodepth` value will be used.
2474 .. option:: iodepth_batch_complete_min=int, iodepth_batch_complete=int
2476 This defines how many pieces of I/O to retrieve at once. It defaults to 1
2477 which means that we'll ask for a minimum of 1 I/O in the retrieval process
2478 from the kernel. The I/O retrieval will go on until we hit the limit set by
2479 :option:`iodepth_low`. If this variable is set to 0, then fio will always
2480 check for completed events before queuing more I/O. This helps reduce I/O
2481 latency, at the cost of more retrieval system calls.
2483 .. option:: iodepth_batch_complete_max=int
2485 This defines maximum pieces of I/O to retrieve at once. This variable should
2486 be used along with :option:`iodepth_batch_complete_min`\=int variable,
2487 specifying the range of min and max amount of I/O which should be
2488 retrieved. By default it is equal to the :option:`iodepth_batch_complete_min`
2493 iodepth_batch_complete_min=1
2494 iodepth_batch_complete_max=<iodepth>
2496 which means that we will retrieve at least 1 I/O and up to the whole
2497 submitted queue depth. If none of I/O has been completed yet, we will wait.
2501 iodepth_batch_complete_min=0
2502 iodepth_batch_complete_max=<iodepth>
2504 which means that we can retrieve up to the whole submitted queue depth, but
2505 if none of I/O has been completed yet, we will NOT wait and immediately exit
2506 the system call. In this example we simply do polling.
2508 .. option:: iodepth_low=int
2510 The low water mark indicating when to start filling the queue
2511 again. Defaults to the same as :option:`iodepth`, meaning that fio will
2512 attempt to keep the queue full at all times. If :option:`iodepth` is set to
2513 e.g. 16 and *iodepth_low* is set to 4, then after fio has filled the queue of
2514 16 requests, it will let the depth drain down to 4 before starting to fill
2517 .. option:: serialize_overlap=bool
2519 Serialize in-flight I/Os that might otherwise cause or suffer from data races.
2520 When two or more I/Os are submitted simultaneously, there is no guarantee that
2521 the I/Os will be processed or completed in the submitted order. Further, if
2522 two or more of those I/Os are writes, any overlapping region between them can
2523 become indeterminate/undefined on certain storage. These issues can cause
2524 verification to fail erratically when at least one of the racing I/Os is
2525 changing data and the overlapping region has a non-zero size. Setting
2526 ``serialize_overlap`` tells fio to avoid provoking this behavior by explicitly
2527 serializing in-flight I/Os that have a non-zero overlap. Note that setting
2528 this option can reduce both performance and the :option:`iodepth` achieved.
2530 This option only applies to I/Os issued for a single job except when it is
2531 enabled along with :option:`io_submit_mode`\=offload. In offload mode, fio
2532 will check for overlap among all I/Os submitted by offload jobs with :option:`serialize_overlap`
2537 .. option:: io_submit_mode=str
2539 This option controls how fio submits the I/O to the I/O engine. The default
2540 is `inline`, which means that the fio job threads submit and reap I/O
2541 directly. If set to `offload`, the job threads will offload I/O submission
2542 to a dedicated pool of I/O threads. This requires some coordination and thus
2543 has a bit of extra overhead, especially for lower queue depth I/O where it
2544 can increase latencies. The benefit is that fio can manage submission rates
2545 independently of the device completion rates. This avoids skewed latency
2546 reporting if I/O gets backed up on the device side (the coordinated omission
2547 problem). Note that this option cannot reliably be used with async IO
2554 .. option:: thinktime=time
2556 Stall the job for the specified period of time after an I/O has completed before issuing the
2557 next. May be used to simulate processing being done by an application.
2558 When the unit is omitted, the value is interpreted in microseconds. See
2559 :option:`thinktime_blocks` and :option:`thinktime_spin`.
2561 .. option:: thinktime_spin=time
2563 Only valid if :option:`thinktime` is set - pretend to spend CPU time doing
2564 something with the data received, before falling back to sleeping for the
2565 rest of the period specified by :option:`thinktime`. When the unit is
2566 omitted, the value is interpreted in microseconds.
2568 .. option:: thinktime_blocks=int
2570 Only valid if :option:`thinktime` is set - control how many blocks to issue,
2571 before waiting :option:`thinktime` usecs. If not set, defaults to 1 which will make
2572 fio wait :option:`thinktime` usecs after every block. This effectively makes any
2573 queue depth setting redundant, since no more than 1 I/O will be queued
2574 before we have to complete it and do our :option:`thinktime`. In other words, this
2575 setting effectively caps the queue depth if the latter is larger.
2577 .. option:: thinktime_blocks_type=str
2579 Only valid if :option:`thinktime` is set - control how :option:`thinktime_blocks`
2580 triggers. The default is `complete`, which triggers thinktime when fio completes
2581 :option:`thinktime_blocks` blocks. If this is set to `issue`, then the trigger happens
2584 .. option:: rate=int[,int][,int]
2586 Cap the bandwidth used by this job. The number is in bytes/sec, the normal
2587 suffix rules apply. Comma-separated values may be specified for reads,
2588 writes, and trims as described in :option:`blocksize`.
2590 For example, using `rate=1m,500k` would limit reads to 1MiB/sec and writes to
2591 500KiB/sec. Capping only reads or writes can be done with `rate=,500k` or
2592 `rate=500k,` where the former will only limit writes (to 500KiB/sec) and the
2593 latter will only limit reads.
2595 .. option:: rate_min=int[,int][,int]
2597 Tell fio to do whatever it can to maintain at least this bandwidth. Failing
2598 to meet this requirement will cause the job to exit. Comma-separated values
2599 may be specified for reads, writes, and trims as described in
2600 :option:`blocksize`.
2602 .. option:: rate_iops=int[,int][,int]
2604 Cap the bandwidth to this number of IOPS. Basically the same as
2605 :option:`rate`, just specified independently of bandwidth. If the job is
2606 given a block size range instead of a fixed value, the smallest block size
2607 is used as the metric. Comma-separated values may be specified for reads,
2608 writes, and trims as described in :option:`blocksize`.
2610 .. option:: rate_iops_min=int[,int][,int]
2612 If fio doesn't meet this rate of I/O, it will cause the job to exit.
2613 Comma-separated values may be specified for reads, writes, and trims as
2614 described in :option:`blocksize`.
2616 .. option:: rate_process=str
2618 This option controls how fio manages rated I/O submissions. The default is
2619 `linear`, which submits I/O in a linear fashion with fixed delays between
2620 I/Os that gets adjusted based on I/O completion rates. If this is set to
2621 `poisson`, fio will submit I/O based on a more real world random request
2622 flow, known as the Poisson process
2623 (https://en.wikipedia.org/wiki/Poisson_point_process). The lambda will be
2624 10^6 / IOPS for the given workload.
2626 .. option:: rate_ignore_thinktime=bool
2628 By default, fio will attempt to catch up to the specified rate setting,
2629 if any kind of thinktime setting was used. If this option is set, then
2630 fio will ignore the thinktime and continue doing IO at the specified
2631 rate, instead of entering a catch-up mode after thinktime is done.
2637 .. option:: latency_target=time
2639 If set, fio will attempt to find the max performance point that the given
2640 workload will run at while maintaining a latency below this target. When
2641 the unit is omitted, the value is interpreted in microseconds. See
2642 :option:`latency_window` and :option:`latency_percentile`.
2644 .. option:: latency_window=time
2646 Used with :option:`latency_target` to specify the sample window that the job
2647 is run at varying queue depths to test the performance. When the unit is
2648 omitted, the value is interpreted in microseconds.
2650 .. option:: latency_percentile=float
2652 The percentage of I/Os that must fall within the criteria specified by
2653 :option:`latency_target` and :option:`latency_window`. If not set, this
2654 defaults to 100.0, meaning that all I/Os must be equal or below to the value
2655 set by :option:`latency_target`.
2657 .. option:: latency_run=bool
2659 Used with :option:`latency_target`. If false (default), fio will find
2660 the highest queue depth that meets :option:`latency_target` and exit. If
2661 true, fio will continue running and try to meet :option:`latency_target`
2662 by adjusting queue depth.
2664 .. option:: max_latency=time
2666 If set, fio will exit the job with an ETIMEDOUT error if it exceeds this
2667 maximum latency. When the unit is omitted, the value is interpreted in
2670 .. option:: rate_cycle=int
2672 Average bandwidth for :option:`rate` and :option:`rate_min` over this number
2673 of milliseconds. Defaults to 1000.
2679 .. option:: write_iolog=str
2681 Write the issued I/O patterns to the specified file. See
2682 :option:`read_iolog`. Specify a separate file for each job, otherwise the
2683 iologs will be interspersed and the file may be corrupt.
2685 .. option:: read_iolog=str
2687 Open an iolog with the specified filename and replay the I/O patterns it
2688 contains. This can be used to store a workload and replay it sometime
2689 later. The iolog given may also be a blktrace binary file, which allows fio
2690 to replay a workload captured by :command:`blktrace`. See
2691 :manpage:`blktrace(8)` for how to capture such logging data. For blktrace
2692 replay, the file needs to be turned into a blkparse binary data file first
2693 (``blkparse <device> -o /dev/null -d file_for_fio.bin``).
2694 You can specify a number of files by separating the names with a ':'
2695 character. See the :option:`filename` option for information on how to
2696 escape ':' characters within the file names. These files will
2697 be sequentially assigned to job clones created by :option:`numjobs`.
2698 '-' is a reserved name, meaning read from stdin, notably if
2699 :option:`filename` is set to '-' which means stdin as well, then
2700 this flag can't be set to '-'.
2702 .. option:: read_iolog_chunked=bool
2704 Determines how iolog is read. If false(default) entire :option:`read_iolog`
2705 will be read at once. If selected true, input from iolog will be read
2706 gradually. Useful when iolog is very large, or it is generated.
2708 .. option:: merge_blktrace_file=str
2710 When specified, rather than replaying the logs passed to :option:`read_iolog`,
2711 the logs go through a merge phase which aggregates them into a single
2712 blktrace. The resulting file is then passed on as the :option:`read_iolog`
2713 parameter. The intention here is to make the order of events consistent.
2714 This limits the influence of the scheduler compared to replaying multiple
2715 blktraces via concurrent jobs.
2717 .. option:: merge_blktrace_scalars=float_list
2719 This is a percentage based option that is index paired with the list of
2720 files passed to :option:`read_iolog`. When merging is performed, scale
2721 the time of each event by the corresponding amount. For example,
2722 ``--merge_blktrace_scalars="50:100"`` runs the first trace in halftime
2723 and the second trace in realtime. This knob is separately tunable from
2724 :option:`replay_time_scale` which scales the trace during runtime and
2725 does not change the output of the merge unlike this option.
2727 .. option:: merge_blktrace_iters=float_list
2729 This is a whole number option that is index paired with the list of files
2730 passed to :option:`read_iolog`. When merging is performed, run each trace
2731 for the specified number of iterations. For example,
2732 ``--merge_blktrace_iters="2:1"`` runs the first trace for two iterations
2733 and the second trace for one iteration.
2735 .. option:: replay_no_stall=bool
2737 When replaying I/O with :option:`read_iolog` the default behavior is to
2738 attempt to respect the timestamps within the log and replay them with the
2739 appropriate delay between IOPS. By setting this variable fio will not
2740 respect the timestamps and attempt to replay them as fast as possible while
2741 still respecting ordering. The result is the same I/O pattern to a given
2742 device, but different timings.
2744 .. option:: replay_time_scale=int
2746 When replaying I/O with :option:`read_iolog`, fio will honor the
2747 original timing in the trace. With this option, it's possible to scale
2748 the time. It's a percentage option, if set to 50 it means run at 50%
2749 the original IO rate in the trace. If set to 200, run at twice the
2750 original IO rate. Defaults to 100.
2752 .. option:: replay_redirect=str
2754 While replaying I/O patterns using :option:`read_iolog` the default behavior
2755 is to replay the IOPS onto the major/minor device that each IOP was recorded
2756 from. This is sometimes undesirable because on a different machine those
2757 major/minor numbers can map to a different device. Changing hardware on the
2758 same system can also result in a different major/minor mapping.
2759 ``replay_redirect`` causes all I/Os to be replayed onto the single specified
2760 device regardless of the device it was recorded
2761 from. i.e. :option:`replay_redirect`\= :file:`/dev/sdc` would cause all I/O
2762 in the blktrace or iolog to be replayed onto :file:`/dev/sdc`. This means
2763 multiple devices will be replayed onto a single device, if the trace
2764 contains multiple devices. If you want multiple devices to be replayed
2765 concurrently to multiple redirected devices you must blkparse your trace
2766 into separate traces and replay them with independent fio invocations.
2767 Unfortunately this also breaks the strict time ordering between multiple
2770 .. option:: replay_align=int
2772 Force alignment of the byte offsets in a trace to this value. The value
2773 must be a power of 2.
2775 .. option:: replay_scale=int
2777 Scale byte offsets down by this factor when replaying traces. Should most
2778 likely use :option:`replay_align` as well.
2780 .. option:: replay_skip=str
2782 Sometimes it's useful to skip certain IO types in a replay trace.
2783 This could be, for instance, eliminating the writes in the trace.
2784 Or not replaying the trims/discards, if you are redirecting to
2785 a device that doesn't support them. This option takes a comma
2786 separated list of read, write, trim, sync.
2789 Threads, processes and job synchronization
2790 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2794 Fio defaults to creating jobs by using fork, however if this option is
2795 given, fio will create jobs by using POSIX Threads' function
2796 :manpage:`pthread_create(3)` to create threads instead.
2798 .. option:: wait_for=str
2800 If set, the current job won't be started until all workers of the specified
2801 waitee job are done.
2803 ``wait_for`` operates on the job name basis, so there are a few
2804 limitations. First, the waitee must be defined prior to the waiter job
2805 (meaning no forward references). Second, if a job is being referenced as a
2806 waitee, it must have a unique name (no duplicate waitees).
2808 .. option:: nice=int
2810 Run the job with the given nice value. See man :manpage:`nice(2)`.
2812 On Windows, values less than -15 set the process class to "High"; -1 through
2813 -15 set "Above Normal"; 1 through 15 "Below Normal"; and above 15 "Idle"
2816 .. option:: prio=int
2818 Set the I/O priority value of this job. Linux limits us to a positive value
2819 between 0 and 7, with 0 being the highest. See man
2820 :manpage:`ionice(1)`. Refer to an appropriate manpage for other operating
2821 systems since meaning of priority may differ. For per-command priority
2822 setting, see I/O engine specific `cmdprio_percentage` and `hipri_percentage`
2825 .. option:: prioclass=int
2827 Set the I/O priority class. See man :manpage:`ionice(1)`. For per-command
2828 priority setting, see I/O engine specific `cmdprio_percentage` and
2829 `hipri_percentage` options.
2831 .. option:: cpus_allowed=str
2833 Controls the same options as :option:`cpumask`, but accepts a textual
2834 specification of the permitted CPUs instead and CPUs are indexed from 0. So
2835 to use CPUs 0 and 5 you would specify ``cpus_allowed=0,5``. This option also
2836 allows a range of CPUs to be specified -- say you wanted a binding to CPUs
2837 0, 5, and 8 to 15, you would set ``cpus_allowed=0,5,8-15``.
2839 On Windows, when ``cpus_allowed`` is unset only CPUs from fio's current
2840 processor group will be used and affinity settings are inherited from the
2841 system. An fio build configured to target Windows 7 makes options that set
2842 CPUs processor group aware and values will set both the processor group
2843 and a CPU from within that group. For example, on a system where processor
2844 group 0 has 40 CPUs and processor group 1 has 32 CPUs, ``cpus_allowed``
2845 values between 0 and 39 will bind CPUs from processor group 0 and
2846 ``cpus_allowed`` values between 40 and 71 will bind CPUs from processor
2847 group 1. When using ``cpus_allowed_policy=shared`` all CPUs specified by a
2848 single ``cpus_allowed`` option must be from the same processor group. For
2849 Windows fio builds not built for Windows 7, CPUs will only be selected from
2850 (and be relative to) whatever processor group fio happens to be running in
2851 and CPUs from other processor groups cannot be used.
2853 .. option:: cpus_allowed_policy=str
2855 Set the policy of how fio distributes the CPUs specified by
2856 :option:`cpus_allowed` or :option:`cpumask`. Two policies are supported:
2859 All jobs will share the CPU set specified.
2861 Each job will get a unique CPU from the CPU set.
2863 **shared** is the default behavior, if the option isn't specified. If
2864 **split** is specified, then fio will assign one cpu per job. If not
2865 enough CPUs are given for the jobs listed, then fio will roundrobin the CPUs
2868 .. option:: cpumask=int
2870 Set the CPU affinity of this job. The parameter given is a bit mask of
2871 allowed CPUs the job may run on. So if you want the allowed CPUs to be 1
2872 and 5, you would pass the decimal value of (1 << 1 | 1 << 5), or 34. See man
2873 :manpage:`sched_setaffinity(2)`. This may not work on all supported
2874 operating systems or kernel versions. This option doesn't work well for a
2875 higher CPU count than what you can store in an integer mask, so it can only
2876 control cpus 1-32. For boxes with larger CPU counts, use
2877 :option:`cpus_allowed`.
2879 .. option:: numa_cpu_nodes=str
2881 Set this job running on specified NUMA nodes' CPUs. The arguments allow
2882 comma delimited list of cpu numbers, A-B ranges, or `all`. Note, to enable
2883 NUMA options support, fio must be built on a system with libnuma-dev(el)
2886 .. option:: numa_mem_policy=str
2888 Set this job's memory policy and corresponding NUMA nodes. Format of the
2893 ``mode`` is one of the following memory policies: ``default``, ``prefer``,
2894 ``bind``, ``interleave`` or ``local``. For ``default`` and ``local`` memory
2895 policies, no node needs to be specified. For ``prefer``, only one node is
2896 allowed. For ``bind`` and ``interleave`` the ``nodelist`` may be as
2897 follows: a comma delimited list of numbers, A-B ranges, or `all`.
2899 .. option:: cgroup=str
2901 Add job to this control group. If it doesn't exist, it will be created. The
2902 system must have a mounted cgroup blkio mount point for this to work. If
2903 your system doesn't have it mounted, you can do so with::
2905 # mount -t cgroup -o blkio none /cgroup
2907 .. option:: cgroup_weight=int
2909 Set the weight of the cgroup to this value. See the documentation that comes
2910 with the kernel, allowed values are in the range of 100..1000.
2912 .. option:: cgroup_nodelete=bool
2914 Normally fio will delete the cgroups it has created after the job
2915 completion. To override this behavior and to leave cgroups around after the
2916 job completion, set ``cgroup_nodelete=1``. This can be useful if one wants
2917 to inspect various cgroup files after job completion. Default: false.
2919 .. option:: flow_id=int
2921 The ID of the flow. If not specified, it defaults to being a global
2922 flow. See :option:`flow`.
2924 .. option:: flow=int
2926 Weight in token-based flow control. If this value is used, then there is a
2927 'flow counter' which is used to regulate the proportion of activity between
2928 two or more jobs. Fio attempts to keep this flow counter near zero. The
2929 ``flow`` parameter stands for how much should be added or subtracted to the
2930 flow counter on each iteration of the main I/O loop. That is, if one job has
2931 ``flow=8`` and another job has ``flow=-1``, then there will be a roughly 1:8
2932 ratio in how much one runs vs the other.
2934 .. option:: flow_sleep=int
2936 The period of time, in microseconds, to wait after the flow counter
2937 has exceeded its proportion before retrying operations.
2939 .. option:: stonewall, wait_for_previous
2941 Wait for preceding jobs in the job file to exit, before starting this
2942 one. Can be used to insert serialization points in the job file. A stone
2943 wall also implies starting a new reporting group, see
2944 :option:`group_reporting`.
2948 By default, fio will continue running all other jobs when one job finishes.
2949 Sometimes this is not the desired action. Setting ``exitall`` will instead
2950 make fio terminate all jobs in the same group, as soon as one job of that
2953 .. option:: exit_what
2955 By default, fio will continue running all other jobs when one job finishes.
2956 Sometimes this is not the desired action. Setting ``exit_all`` will
2957 instead make fio terminate all jobs in the same group. The option
2958 ``exit_what`` allows to control which jobs get terminated when ``exitall`` is
2959 enabled. The default is ``group`` and does not change the behaviour of
2960 ``exitall``. The setting ``all`` terminates all jobs. The setting ``stonewall``
2961 terminates all currently running jobs across all groups and continues execution
2962 with the next stonewalled group.
2964 .. option:: exec_prerun=str
2966 Before running this job, issue the command specified through
2967 :manpage:`system(3)`. Output is redirected in a file called
2968 :file:`jobname.prerun.txt`.
2970 .. option:: exec_postrun=str
2972 After the job completes, issue the command specified though
2973 :manpage:`system(3)`. Output is redirected in a file called
2974 :file:`jobname.postrun.txt`.
2978 Instead of running as the invoking user, set the user ID to this value
2979 before the thread/process does any work.
2983 Set group ID, see :option:`uid`.
2989 .. option:: verify_only
2991 Do not perform specified workload, only verify data still matches previous
2992 invocation of this workload. This option allows one to check data multiple
2993 times at a later date without overwriting it. This option makes sense only
2994 for workloads that write data, and does not support workloads with the
2995 :option:`time_based` option set.
2997 .. option:: do_verify=bool
2999 Run the verify phase after a write phase. Only valid if :option:`verify` is
3002 .. option:: verify=str
3004 If writing to a file, fio can verify the file contents after each iteration
3005 of the job. Each verification method also implies verification of special
3006 header, which is written to the beginning of each block. This header also
3007 includes meta information, like offset of the block, block number, timestamp
3008 when block was written, etc. :option:`verify` can be combined with
3009 :option:`verify_pattern` option. The allowed values are:
3012 Use an md5 sum of the data area and store it in the header of
3016 Use an experimental crc64 sum of the data area and store it in the
3017 header of each block.
3020 Use a crc32c sum of the data area and store it in the header of
3021 each block. This will automatically use hardware acceleration
3022 (e.g. SSE4.2 on an x86 or CRC crypto extensions on ARM64) but will
3023 fall back to software crc32c if none is found. Generally the
3024 fastest checksum fio supports when hardware accelerated.
3030 Use a crc32 sum of the data area and store it in the header of each
3034 Use a crc16 sum of the data area and store it in the header of each
3038 Use a crc7 sum of the data area and store it in the header of each
3042 Use xxhash as the checksum function. Generally the fastest software
3043 checksum that fio supports.
3046 Use sha512 as the checksum function.
3049 Use sha256 as the checksum function.
3052 Use optimized sha1 as the checksum function.
3055 Use optimized sha3-224 as the checksum function.
3058 Use optimized sha3-256 as the checksum function.
3061 Use optimized sha3-384 as the checksum function.
3064 Use optimized sha3-512 as the checksum function.
3067 This option is deprecated, since now meta information is included in
3068 generic verification header and meta verification happens by
3069 default. For detailed information see the description of the
3070 :option:`verify` setting. This option is kept because of
3071 compatibility's sake with old configurations. Do not use it.
3074 Verify a strict pattern. Normally fio includes a header with some
3075 basic information and checksumming, but if this option is set, only
3076 the specific pattern set with :option:`verify_pattern` is verified.
3079 Only pretend to verify. Useful for testing internals with
3080 :option:`ioengine`\=null, not for much else.
3082 This option can be used for repeated burn-in tests of a system to make sure
3083 that the written data is also correctly read back. If the data direction
3084 given is a read or random read, fio will assume that it should verify a
3085 previously written file. If the data direction includes any form of write,
3086 the verify will be of the newly written data.
3088 To avoid false verification errors, do not use the norandommap option when
3089 verifying data with async I/O engines and I/O depths > 1. Or use the
3090 norandommap and the lfsr random generator together to avoid writing to the
3091 same offset with muliple outstanding I/Os.
3093 .. option:: verify_offset=int
3095 Swap the verification header with data somewhere else in the block before
3096 writing. It is swapped back before verifying.
3098 .. option:: verify_interval=int
3100 Write the verification header at a finer granularity than the
3101 :option:`blocksize`. It will be written for chunks the size of
3102 ``verify_interval``. :option:`blocksize` should divide this evenly.
3104 .. option:: verify_pattern=str
3106 If set, fio will fill the I/O buffers with this pattern. Fio defaults to
3107 filling with totally random bytes, but sometimes it's interesting to fill
3108 with a known pattern for I/O verification purposes. Depending on the width
3109 of the pattern, fio will fill 1/2/3/4 bytes of the buffer at the time (it can
3110 be either a decimal or a hex number). The ``verify_pattern`` if larger than
3111 a 32-bit quantity has to be a hex number that starts with either "0x" or
3112 "0X". Use with :option:`verify`. Also, ``verify_pattern`` supports %o
3113 format, which means that for each block offset will be written and then
3114 verified back, e.g.::
3118 Or use combination of everything::
3120 verify_pattern=0xff%o"abcd"-12
3122 .. option:: verify_fatal=bool
3124 Normally fio will keep checking the entire contents before quitting on a
3125 block verification failure. If this option is set, fio will exit the job on
3126 the first observed failure. Default: false.
3128 .. option:: verify_dump=bool
3130 If set, dump the contents of both the original data block and the data block
3131 we read off disk to files. This allows later analysis to inspect just what
3132 kind of data corruption occurred. Off by default.
3134 .. option:: verify_async=int
3136 Fio will normally verify I/O inline from the submitting thread. This option
3137 takes an integer describing how many async offload threads to create for I/O
3138 verification instead, causing fio to offload the duty of verifying I/O
3139 contents to one or more separate threads. If using this offload option, even
3140 sync I/O engines can benefit from using an :option:`iodepth` setting higher
3141 than 1, as it allows them to have I/O in flight while verifies are running.
3142 Defaults to 0 async threads, i.e. verification is not asynchronous.
3144 .. option:: verify_async_cpus=str
3146 Tell fio to set the given CPU affinity on the async I/O verification
3147 threads. See :option:`cpus_allowed` for the format used.
3149 .. option:: verify_backlog=int
3151 Fio will normally verify the written contents of a job that utilizes verify
3152 once that job has completed. In other words, everything is written then
3153 everything is read back and verified. You may want to verify continually
3154 instead for a variety of reasons. Fio stores the meta data associated with
3155 an I/O block in memory, so for large verify workloads, quite a bit of memory
3156 would be used up holding this meta data. If this option is enabled, fio will
3157 write only N blocks before verifying these blocks.
3159 .. option:: verify_backlog_batch=int
3161 Control how many blocks fio will verify if :option:`verify_backlog` is
3162 set. If not set, will default to the value of :option:`verify_backlog`
3163 (meaning the entire queue is read back and verified). If
3164 ``verify_backlog_batch`` is less than :option:`verify_backlog` then not all
3165 blocks will be verified, if ``verify_backlog_batch`` is larger than
3166 :option:`verify_backlog`, some blocks will be verified more than once.
3168 .. option:: verify_state_save=bool
3170 When a job exits during the write phase of a verify workload, save its
3171 current state. This allows fio to replay up until that point, if the verify
3172 state is loaded for the verify read phase. The format of the filename is,
3175 <type>-<jobname>-<jobindex>-verify.state.
3177 <type> is "local" for a local run, "sock" for a client/server socket
3178 connection, and "ip" (192.168.0.1, for instance) for a networked
3179 client/server connection. Defaults to true.
3181 .. option:: verify_state_load=bool
3183 If a verify termination trigger was used, fio stores the current write state
3184 of each thread. This can be used at verification time so that fio knows how
3185 far it should verify. Without this information, fio will run a full
3186 verification pass, according to the settings in the job file used. Default
3189 .. option:: trim_percentage=int
3191 Number of verify blocks to discard/trim.
3193 .. option:: trim_verify_zero=bool
3195 Verify that trim/discarded blocks are returned as zeros.
3197 .. option:: trim_backlog=int
3199 Trim after this number of blocks are written.
3201 .. option:: trim_backlog_batch=int
3203 Trim this number of I/O blocks.
3205 .. option:: experimental_verify=bool
3207 Enable experimental verification.
3212 .. option:: steadystate=str:float, ss=str:float
3214 Define the criterion and limit for assessing steady state performance. The
3215 first parameter designates the criterion whereas the second parameter sets
3216 the threshold. When the criterion falls below the threshold for the
3217 specified duration, the job will stop. For example, `iops_slope:0.1%` will
3218 direct fio to terminate the job when the least squares regression slope
3219 falls below 0.1% of the mean IOPS. If :option:`group_reporting` is enabled
3220 this will apply to all jobs in the group. Below is the list of available
3221 steady state assessment criteria. All assessments are carried out using only
3222 data from the rolling collection window. Threshold limits can be expressed
3223 as a fixed value or as a percentage of the mean in the collection window.
3225 When using this feature, most jobs should include the :option:`time_based`
3226 and :option:`runtime` options or the :option:`loops` option so that fio does not
3227 stop running after it has covered the full size of the specified file(s) or device(s).
3230 Collect IOPS data. Stop the job if all individual IOPS measurements
3231 are within the specified limit of the mean IOPS (e.g., ``iops:2``
3232 means that all individual IOPS values must be within 2 of the mean,
3233 whereas ``iops:0.2%`` means that all individual IOPS values must be
3234 within 0.2% of the mean IOPS to terminate the job).
3237 Collect IOPS data and calculate the least squares regression
3238 slope. Stop the job if the slope falls below the specified limit.
3241 Collect bandwidth data. Stop the job if all individual bandwidth
3242 measurements are within the specified limit of the mean bandwidth.
3245 Collect bandwidth data and calculate the least squares regression
3246 slope. Stop the job if the slope falls below the specified limit.
3248 .. option:: steadystate_duration=time, ss_dur=time
3250 A rolling window of this duration will be used to judge whether steady state
3251 has been reached. Data will be collected once per second. The default is 0
3252 which disables steady state detection. When the unit is omitted, the
3253 value is interpreted in seconds.
3255 .. option:: steadystate_ramp_time=time, ss_ramp=time
3257 Allow the job to run for the specified duration before beginning data
3258 collection for checking the steady state job termination criterion. The
3259 default is 0. When the unit is omitted, the value is interpreted in seconds.
3262 Measurements and reporting
3263 ~~~~~~~~~~~~~~~~~~~~~~~~~~
3265 .. option:: per_job_logs=bool
3267 If set, this generates bw/clat/iops log with per file private filenames. If
3268 not set, jobs with identical names will share the log filename. Default:
3271 .. option:: group_reporting
3273 It may sometimes be interesting to display statistics for groups of jobs as
3274 a whole instead of for each individual job. This is especially true if
3275 :option:`numjobs` is used; looking at individual thread/process output
3276 quickly becomes unwieldy. To see the final report per-group instead of
3277 per-job, use :option:`group_reporting`. Jobs in a file will be part of the
3278 same reporting group, unless if separated by a :option:`stonewall`, or by
3279 using :option:`new_group`.
3281 .. option:: new_group
3283 Start a new reporting group. See: :option:`group_reporting`. If not given,
3284 all jobs in a file will be part of the same reporting group, unless
3285 separated by a :option:`stonewall`.
3287 .. option:: stats=bool
3289 By default, fio collects and shows final output results for all jobs
3290 that run. If this option is set to 0, then fio will ignore it in
3291 the final stat output.
3293 .. option:: write_bw_log=str
3295 If given, write a bandwidth log for this job. Can be used to store data of
3296 the bandwidth of the jobs in their lifetime.
3298 If no str argument is given, the default filename of
3299 :file:`jobname_type.x.log` is used. Even when the argument is given, fio
3300 will still append the type of log. So if one specifies::
3304 The actual log name will be :file:`foo_bw.x.log` where `x` is the index
3305 of the job (`1..N`, where `N` is the number of jobs). If
3306 :option:`per_job_logs` is false, then the filename will not include the
3309 The included :command:`fio_generate_plots` script uses :command:`gnuplot` to turn these
3310 text files into nice graphs. See `Log File Formats`_ for how data is
3311 structured within the file.
3313 .. option:: write_lat_log=str
3315 Same as :option:`write_bw_log`, except this option creates I/O
3316 submission (e.g., :file:`name_slat.x.log`), completion (e.g.,
3317 :file:`name_clat.x.log`), and total (e.g., :file:`name_lat.x.log`)
3318 latency files instead. See :option:`write_bw_log` for details about
3319 the filename format and `Log File Formats`_ for how data is structured
3322 .. option:: write_hist_log=str
3324 Same as :option:`write_bw_log` but writes an I/O completion latency
3325 histogram file (e.g., :file:`name_hist.x.log`) instead. Note that this
3326 file will be empty unless :option:`log_hist_msec` has also been set.
3327 See :option:`write_bw_log` for details about the filename format and
3328 `Log File Formats`_ for how data is structured within the file.
3330 .. option:: write_iops_log=str
3332 Same as :option:`write_bw_log`, but writes an IOPS file (e.g.
3333 :file:`name_iops.x.log`) instead. Because fio defaults to individual
3334 I/O logging, the value entry in the IOPS log will be 1 unless windowed
3335 logging (see :option:`log_avg_msec`) has been enabled. See
3336 :option:`write_bw_log` for details about the filename format and `Log
3337 File Formats`_ for how data is structured within the file.
3339 .. option:: log_avg_msec=int
3341 By default, fio will log an entry in the iops, latency, or bw log for every
3342 I/O that completes. When writing to the disk log, that can quickly grow to a
3343 very large size. Setting this option makes fio average the each log entry
3344 over the specified period of time, reducing the resolution of the log. See
3345 :option:`log_max_value` as well. Defaults to 0, logging all entries.
3346 Also see `Log File Formats`_.
3348 .. option:: log_hist_msec=int
3350 Same as :option:`log_avg_msec`, but logs entries for completion latency
3351 histograms. Computing latency percentiles from averages of intervals using
3352 :option:`log_avg_msec` is inaccurate. Setting this option makes fio log
3353 histogram entries over the specified period of time, reducing log sizes for
3354 high IOPS devices while retaining percentile accuracy. See
3355 :option:`log_hist_coarseness` and :option:`write_hist_log` as well.
3356 Defaults to 0, meaning histogram logging is disabled.
3358 .. option:: log_hist_coarseness=int
3360 Integer ranging from 0 to 6, defining the coarseness of the resolution of
3361 the histogram logs enabled with :option:`log_hist_msec`. For each increment
3362 in coarseness, fio outputs half as many bins. Defaults to 0, for which
3363 histogram logs contain 1216 latency bins. See :option:`write_hist_log`
3364 and `Log File Formats`_.
3366 .. option:: log_max_value=bool
3368 If :option:`log_avg_msec` is set, fio logs the average over that window. If
3369 you instead want to log the maximum value, set this option to 1. Defaults to
3370 0, meaning that averaged values are logged.
3372 .. option:: log_offset=bool
3374 If this is set, the iolog options will include the byte offset for the I/O
3375 entry as well as the other data values. Defaults to 0 meaning that
3376 offsets are not present in logs. Also see `Log File Formats`_.
3378 .. option:: log_compression=int
3380 If this is set, fio will compress the I/O logs as it goes, to keep the
3381 memory footprint lower. When a log reaches the specified size, that chunk is
3382 removed and compressed in the background. Given that I/O logs are fairly
3383 highly compressible, this yields a nice memory savings for longer runs. The
3384 downside is that the compression will consume some background CPU cycles, so
3385 it may impact the run. This, however, is also true if the logging ends up
3386 consuming most of the system memory. So pick your poison. The I/O logs are
3387 saved normally at the end of a run, by decompressing the chunks and storing
3388 them in the specified log file. This feature depends on the availability of
3391 .. option:: log_compression_cpus=str
3393 Define the set of CPUs that are allowed to handle online log compression for
3394 the I/O jobs. This can provide better isolation between performance
3395 sensitive jobs, and background compression work. See
3396 :option:`cpus_allowed` for the format used.
3398 .. option:: log_store_compressed=bool
3400 If set, fio will store the log files in a compressed format. They can be
3401 decompressed with fio, using the :option:`--inflate-log` command line
3402 parameter. The files will be stored with a :file:`.fz` suffix.
3404 .. option:: log_unix_epoch=bool
3406 If set, fio will log Unix timestamps to the log files produced by enabling
3407 write_type_log for each log type, instead of the default zero-based
3410 .. option:: block_error_percentiles=bool
3412 If set, record errors in trim block-sized units from writes and trims and
3413 output a histogram of how many trims it took to get to errors, and what kind
3414 of error was encountered.
3416 .. option:: bwavgtime=int
3418 Average the calculated bandwidth over the given time. Value is specified in
3419 milliseconds. If the job also does bandwidth logging through
3420 :option:`write_bw_log`, then the minimum of this option and
3421 :option:`log_avg_msec` will be used. Default: 500ms.
3423 .. option:: iopsavgtime=int
3425 Average the calculated IOPS over the given time. Value is specified in
3426 milliseconds. If the job also does IOPS logging through
3427 :option:`write_iops_log`, then the minimum of this option and
3428 :option:`log_avg_msec` will be used. Default: 500ms.
3430 .. option:: disk_util=bool
3432 Generate disk utilization statistics, if the platform supports it.
3435 .. option:: disable_lat=bool
3437 Disable measurements of total latency numbers. Useful only for cutting back
3438 the number of calls to :manpage:`gettimeofday(2)`, as that does impact
3439 performance at really high IOPS rates. Note that to really get rid of a
3440 large amount of these calls, this option must be used with
3441 :option:`disable_slat` and :option:`disable_bw_measurement` as well.
3443 .. option:: disable_clat=bool
3445 Disable measurements of completion latency numbers. See
3446 :option:`disable_lat`.
3448 .. option:: disable_slat=bool
3450 Disable measurements of submission latency numbers. See
3451 :option:`disable_lat`.
3453 .. option:: disable_bw_measurement=bool, disable_bw=bool
3455 Disable measurements of throughput/bandwidth numbers. See
3456 :option:`disable_lat`.
3458 .. option:: slat_percentiles=bool
3460 Report submission latency percentiles. Submission latency is not recorded
3461 for synchronous ioengines.
3463 .. option:: clat_percentiles=bool
3465 Report completion latency percentiles.
3467 .. option:: lat_percentiles=bool
3469 Report total latency percentiles. Total latency is the sum of submission
3470 latency and completion latency.
3472 .. option:: percentile_list=float_list
3474 Overwrite the default list of percentiles for latencies and the block error
3475 histogram. Each number is a floating point number in the range (0,100], and
3476 the maximum length of the list is 20. Use ``:`` to separate the numbers. For
3477 example, ``--percentile_list=99.5:99.9`` will cause fio to report the
3478 latency durations below which 99.5% and 99.9% of the observed latencies fell,
3481 .. option:: significant_figures=int
3483 If using :option:`--output-format` of `normal`, set the significant
3484 figures to this value. Higher values will yield more precise IOPS and
3485 throughput units, while lower values will round. Requires a minimum
3486 value of 1 and a maximum value of 10. Defaults to 4.
3492 .. option:: exitall_on_error
3494 When one job finishes in error, terminate the rest. The default is to wait
3495 for each job to finish.
3497 .. option:: continue_on_error=str
3499 Normally fio will exit the job on the first observed failure. If this option
3500 is set, fio will continue the job when there is a 'non-fatal error' (EIO or
3501 EILSEQ) until the runtime is exceeded or the I/O size specified is
3502 completed. If this option is used, there are two more stats that are
3503 appended, the total error count and the first error. The error field given
3504 in the stats is the first error that was hit during the run.
3506 The allowed values are:
3509 Exit on any I/O or verify errors.
3512 Continue on read errors, exit on all others.
3515 Continue on write errors, exit on all others.
3518 Continue on any I/O error, exit on all others.
3521 Continue on verify errors, exit on all others.
3524 Continue on all errors.
3527 Backward-compatible alias for 'none'.
3530 Backward-compatible alias for 'all'.
3532 .. option:: ignore_error=str
3534 Sometimes you want to ignore some errors during test in that case you can
3535 specify error list for each error type, instead of only being able to
3536 ignore the default 'non-fatal error' using :option:`continue_on_error`.
3537 ``ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST`` errors for
3538 given error type is separated with ':'. Error may be symbol ('ENOSPC',
3539 'ENOMEM') or integer. Example::
3541 ignore_error=EAGAIN,ENOSPC:122
3543 This option will ignore EAGAIN from READ, and ENOSPC and 122(EDQUOT) from
3544 WRITE. This option works by overriding :option:`continue_on_error` with
3545 the list of errors for each error type if any.
3547 .. option:: error_dump=bool
3549 If set dump every error even if it is non fatal, true by default. If
3550 disabled only fatal error will be dumped.
3552 Running predefined workloads
3553 ----------------------------
3555 Fio includes predefined profiles that mimic the I/O workloads generated by
3558 .. option:: profile=str
3560 The predefined workload to run. Current profiles are:
3563 Threaded I/O bench (tiotest/tiobench) like workload.
3566 Aerospike Certification Tool (ACT) like workload.
3568 To view a profile's additional options use :option:`--cmdhelp` after specifying
3569 the profile. For example::
3571 $ fio --profile=act --cmdhelp
3576 .. option:: device-names=str
3581 .. option:: load=int
3584 ACT load multiplier. Default: 1.
3586 .. option:: test-duration=time
3589 How long the entire test takes to run. When the unit is omitted, the value
3590 is given in seconds. Default: 24h.
3592 .. option:: threads-per-queue=int
3595 Number of read I/O threads per device. Default: 8.
3597 .. option:: read-req-num-512-blocks=int
3600 Number of 512B blocks to read at the time. Default: 3.
3602 .. option:: large-block-op-kbytes=int
3605 Size of large block ops in KiB (writes). Default: 131072.
3610 Set to run ACT prep phase.
3612 Tiobench profile options
3613 ~~~~~~~~~~~~~~~~~~~~~~~~
3615 .. option:: size=str
3620 .. option:: block=int
3623 Block size in bytes. Default: 4096.
3625 .. option:: numruns=int
3635 .. option:: threads=int
3640 Interpreting the output
3641 -----------------------
3644 Example output was based on the following:
3645 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --time_based \
3646 --rate=1256k --bs=14K --name=quick --runtime=1s --name=mixed \
3647 --runtime=2m --rw=rw
3649 Fio spits out a lot of output. While running, fio will display the status of the
3650 jobs created. An example of that would be::
3652 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]
3654 The characters inside the first set of square brackets denote the current status of
3655 each thread. The first character is the first job defined in the job file, and so
3656 forth. The possible values (in typical life cycle order) are:
3658 +------+-----+-----------------------------------------------------------+
3660 +======+=====+===========================================================+
3661 | P | | Thread setup, but not started. |
3662 +------+-----+-----------------------------------------------------------+
3663 | C | | Thread created. |
3664 +------+-----+-----------------------------------------------------------+
3665 | I | | Thread initialized, waiting or generating necessary data. |
3666 +------+-----+-----------------------------------------------------------+
3667 | | p | Thread running pre-reading file(s). |
3668 +------+-----+-----------------------------------------------------------+
3669 | | / | Thread is in ramp period. |
3670 +------+-----+-----------------------------------------------------------+
3671 | | R | Running, doing sequential reads. |
3672 +------+-----+-----------------------------------------------------------+
3673 | | r | Running, doing random reads. |
3674 +------+-----+-----------------------------------------------------------+
3675 | | W | Running, doing sequential writes. |
3676 +------+-----+-----------------------------------------------------------+
3677 | | w | Running, doing random writes. |
3678 +------+-----+-----------------------------------------------------------+
3679 | | M | Running, doing mixed sequential reads/writes. |
3680 +------+-----+-----------------------------------------------------------+
3681 | | m | Running, doing mixed random reads/writes. |
3682 +------+-----+-----------------------------------------------------------+
3683 | | D | Running, doing sequential trims. |
3684 +------+-----+-----------------------------------------------------------+
3685 | | d | Running, doing random trims. |
3686 +------+-----+-----------------------------------------------------------+
3687 | | F | Running, currently waiting for :manpage:`fsync(2)`. |
3688 +------+-----+-----------------------------------------------------------+
3689 | | V | Running, doing verification of written data. |
3690 +------+-----+-----------------------------------------------------------+
3691 | f | | Thread finishing. |
3692 +------+-----+-----------------------------------------------------------+
3693 | E | | Thread exited, not reaped by main thread yet. |
3694 +------+-----+-----------------------------------------------------------+
3695 | _ | | Thread reaped. |
3696 +------+-----+-----------------------------------------------------------+
3697 | X | | Thread reaped, exited with an error. |
3698 +------+-----+-----------------------------------------------------------+
3699 | K | | Thread reaped, exited due to signal. |
3700 +------+-----+-----------------------------------------------------------+
3703 Example output was based on the following:
3704 TZ=UTC fio --iodepth=8 --ioengine=null --size=100M --runtime=58m \
3705 --time_based --rate=2512k --bs=256K --numjobs=10 \
3706 --name=readers --rw=read --name=writers --rw=write
3708 Fio will condense the thread string as not to take up more space on the command
3709 line than needed. For instance, if you have 10 readers and 10 writers running,
3710 the output would look like this::
3712 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]
3714 Note that the status string is displayed in order, so it's possible to tell which of
3715 the jobs are currently doing what. In the example above this means that jobs 1--10
3716 are readers and 11--20 are writers.
3718 The other values are fairly self explanatory -- number of threads currently
3719 running and doing I/O, the number of currently open files (f=), the estimated
3720 completion percentage, the rate of I/O since last check (read speed listed first,
3721 then write speed and optionally trim speed) in terms of bandwidth and IOPS,
3722 and time to completion for the current running group. It's impossible to estimate
3723 runtime of the following groups (if any).
3726 Example output was based on the following:
3727 TZ=UTC fio --iodepth=16 --ioengine=posixaio --filename=/tmp/fiofile \
3728 --direct=1 --size=100M --time_based --runtime=50s --rate_iops=89 \
3729 --bs=7K --name=Client1 --rw=write
3731 When fio is done (or interrupted by :kbd:`Ctrl-C`), it will show the data for
3732 each thread, group of threads, and disks in that order. For each overall thread (or
3733 group) the output looks like::
3735 Client1: (groupid=0, jobs=1): err= 0: pid=16109: Sat Jun 24 12:07:54 2017
3736 write: IOPS=88, BW=623KiB/s (638kB/s)(30.4MiB/50032msec)
3737 slat (nsec): min=500, max=145500, avg=8318.00, stdev=4781.50
3738 clat (usec): min=170, max=78367, avg=4019.02, stdev=8293.31
3739 lat (usec): min=174, max=78375, avg=4027.34, stdev=8291.79
3740 clat percentiles (usec):
3741 | 1.00th=[ 302], 5.00th=[ 326], 10.00th=[ 343], 20.00th=[ 363],
3742 | 30.00th=[ 392], 40.00th=[ 404], 50.00th=[ 416], 60.00th=[ 445],
3743 | 70.00th=[ 816], 80.00th=[ 6718], 90.00th=[12911], 95.00th=[21627],
3744 | 99.00th=[43779], 99.50th=[51643], 99.90th=[68682], 99.95th=[72877],
3746 bw ( KiB/s): min= 532, max= 686, per=0.10%, avg=622.87, stdev=24.82, samples= 100
3747 iops : min= 76, max= 98, avg=88.98, stdev= 3.54, samples= 100
3748 lat (usec) : 250=0.04%, 500=64.11%, 750=4.81%, 1000=2.79%
3749 lat (msec) : 2=4.16%, 4=1.84%, 10=4.90%, 20=11.33%, 50=5.37%
3750 lat (msec) : 100=0.65%
3751 cpu : usr=0.27%, sys=0.18%, ctx=12072, majf=0, minf=21
3752 IO depths : 1=85.0%, 2=13.1%, 4=1.8%, 8=0.1%, 16=0.0%, 32=0.0%, >=64=0.0%
3753 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3754 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
3755 issued rwt: total=0,4450,0, short=0,0,0, dropped=0,0,0
3756 latency : target=0, window=0, percentile=100.00%, depth=8
3758 The job name (or first job's name when using :option:`group_reporting`) is printed,
3759 along with the group id, count of jobs being aggregated, last error id seen (which
3760 is 0 when there are no errors), pid/tid of that thread and the time the job/group
3761 completed. Below are the I/O statistics for each data direction performed (showing
3762 writes in the example above). In the order listed, they denote:
3765 The string before the colon shows the I/O direction the statistics
3766 are for. **IOPS** is the average I/Os performed per second. **BW**
3767 is the average bandwidth rate shown as: value in power of 2 format
3768 (value in power of 10 format). The last two values show: (**total
3769 I/O performed** in power of 2 format / **runtime** of that thread).
3772 Submission latency (**min** being the minimum, **max** being the
3773 maximum, **avg** being the average, **stdev** being the standard
3774 deviation). This is the time it took to submit the I/O. For
3775 sync I/O this row is not displayed as the slat is really the
3776 completion latency (since queue/complete is one operation there).
3777 This value can be in nanoseconds, microseconds or milliseconds ---
3778 fio will choose the most appropriate base and print that (in the
3779 example above nanoseconds was the best scale). Note: in :option:`--minimal` mode
3780 latencies are always expressed in microseconds.
3783 Completion latency. Same names as slat, this denotes the time from
3784 submission to completion of the I/O pieces. For sync I/O, clat will
3785 usually be equal (or very close) to 0, as the time from submit to
3786 complete is basically just CPU time (I/O has already been done, see slat
3790 Total latency. Same names as slat and clat, this denotes the time from
3791 when fio created the I/O unit to completion of the I/O operation.
3794 Bandwidth statistics based on samples. Same names as the xlat stats,
3795 but also includes the number of samples taken (**samples**) and an
3796 approximate percentage of total aggregate bandwidth this thread
3797 received in its group (**per**). This last value is only really
3798 useful if the threads in this group are on the same disk, since they
3799 are then competing for disk access.
3802 IOPS statistics based on samples. Same names as bw.
3804 **lat (nsec/usec/msec)**
3805 The distribution of I/O completion latencies. This is the time from when
3806 I/O leaves fio and when it gets completed. Unlike the separate
3807 read/write/trim sections above, the data here and in the remaining
3808 sections apply to all I/Os for the reporting group. 250=0.04% means that
3809 0.04% of the I/Os completed in under 250us. 500=64.11% means that 64.11%
3810 of the I/Os required 250 to 499us for completion.
3813 CPU usage. User and system time, along with the number of context
3814 switches this thread went through, usage of system and user time, and
3815 finally the number of major and minor page faults. The CPU utilization
3816 numbers are averages for the jobs in that reporting group, while the
3817 context and fault counters are summed.
3820 The distribution of I/O depths over the job lifetime. The numbers are
3821 divided into powers of 2 and each entry covers depths from that value
3822 up to those that are lower than the next entry -- e.g., 16= covers
3823 depths from 16 to 31. Note that the range covered by a depth
3824 distribution entry can be different to the range covered by the
3825 equivalent submit/complete distribution entry.
3828 How many pieces of I/O were submitting in a single submit call. Each
3829 entry denotes that amount and below, until the previous entry -- e.g.,
3830 16=100% means that we submitted anywhere between 9 to 16 I/Os per submit
3831 call. Note that the range covered by a submit distribution entry can
3832 be different to the range covered by the equivalent depth distribution
3836 Like the above submit number, but for completions instead.
3839 The number of read/write/trim requests issued, and how many of them were
3843 These values are for :option:`latency_target` and related options. When
3844 these options are engaged, this section describes the I/O depth required
3845 to meet the specified latency target.
3848 Example output was based on the following:
3849 TZ=UTC fio --ioengine=null --iodepth=2 --size=100M --numjobs=2 \
3850 --rate_process=poisson --io_limit=32M --name=read --bs=128k \
3851 --rate=11M --name=write --rw=write --bs=2k --rate=700k
3853 After each client has been listed, the group statistics are printed. They
3854 will look like this::
3856 Run status group 0 (all jobs):
3857 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
3858 WRITE: bw=1231KiB/s (1261kB/s), 616KiB/s-621KiB/s (630kB/s-636kB/s), io=64.0MiB (67.1MB), run=52747-53223msec
3860 For each data direction it prints:
3863 Aggregate bandwidth of threads in this group followed by the
3864 minimum and maximum bandwidth of all the threads in this group.
3865 Values outside of brackets are power-of-2 format and those
3866 within are the equivalent value in a power-of-10 format.
3868 Aggregate I/O performed of all threads in this group. The
3869 format is the same as bw.
3871 The smallest and longest runtimes of the threads in this group.
3873 And finally, the disk statistics are printed. This is Linux specific. They will look like this::
3875 Disk stats (read/write):
3876 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
3878 Each value is printed for both reads and writes, with reads first. The
3882 Number of I/Os performed by all groups.
3884 Number of merges performed by the I/O scheduler.
3886 Number of ticks we kept the disk busy.
3888 Total time spent in the disk queue.
3890 The disk utilization. A value of 100% means we kept the disk
3891 busy constantly, 50% would be a disk idling half of the time.
3893 It is also possible to get fio to dump the current output while it is running,
3894 without terminating the job. To do that, send fio the **USR1** signal. You can
3895 also get regularly timed dumps by using the :option:`--status-interval`
3896 parameter, or by creating a file in :file:`/tmp` named
3897 :file:`fio-dump-status`. If fio sees this file, it will unlink it and dump the
3898 current output status.
3904 For scripted usage where you typically want to generate tables or graphs of the
3905 results, fio can output the results in a semicolon separated format. The format
3906 is one long line of values, such as::
3908 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%
3909 A description of this job goes here.
3911 The job description (if provided) follows on a second line for terse v2.
3912 It appears on the same line for other terse versions.
3914 To enable terse output, use the :option:`--minimal` or
3915 :option:`--output-format`\=terse command line options. The
3916 first value is the version of the terse output format. If the output has to be
3917 changed for some reason, this number will be incremented by 1 to signify that
3920 Split up, the format is as follows (comments in brackets denote when a
3921 field was introduced or whether it's specific to some terse version):
3925 terse version, fio version [v3], jobname, groupid, error
3929 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3930 Submission latency: min, max, mean, stdev (usec)
3931 Completion latency: min, max, mean, stdev (usec)
3932 Completion latency percentiles: 20 fields (see below)
3933 Total latency: min, max, mean, stdev (usec)
3934 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3935 IOPS [v5]: min, max, mean, stdev, number of samples
3941 Total IO (KiB), bandwidth (KiB/sec), IOPS, runtime (msec)
3942 Submission latency: min, max, mean, stdev (usec)
3943 Completion latency: min, max, mean, stdev (usec)
3944 Completion latency percentiles: 20 fields (see below)
3945 Total latency: min, max, mean, stdev (usec)
3946 Bw (KiB/s): min, max, aggregate percentage of total, mean, stdev, number of samples [v5]
3947 IOPS [v5]: min, max, mean, stdev, number of samples
3949 TRIM status [all but version 3]:
3951 Fields are similar to READ/WRITE status.
3955 user, system, context switches, major faults, minor faults
3959 <=1, 2, 4, 8, 16, 32, >=64
3961 I/O latencies microseconds::
3963 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
3965 I/O latencies milliseconds::
3967 <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
3969 Disk utilization [v3]::
3971 disk name, read ios, write ios, read merges, write merges, read ticks, write ticks,
3972 time spent in queue, disk utilization percentage
3974 Additional Info (dependent on continue_on_error, default off)::
3976 total # errors, first error code
3978 Additional Info (dependent on description being set)::
3982 Completion latency percentiles can be a grouping of up to 20 sets, so for the
3983 terse output fio writes all of them. Each field will look like this::
3987 which is the Xth percentile, and the `usec` latency associated with it.
3989 For `Disk utilization`, all disks used by fio are shown. So for each disk there
3990 will be a disk utilization section.
3992 Below is a single line containing short names for each of the fields in the
3993 minimal output v3, separated by semicolons::
3995 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
3997 In client/server mode terse output differs from what appears when jobs are run
3998 locally. Disk utilization data is omitted from the standard terse output and
3999 for v3 and later appears on its own separate line at the end of each terse
4006 The `json` output format is intended to be both human readable and convenient
4007 for automated parsing. For the most part its sections mirror those of the
4008 `normal` output. The `runtime` value is reported in msec and the `bw` value is
4009 reported in 1024 bytes per second units.
4015 The `json+` output format is identical to the `json` output format except that it
4016 adds a full dump of the completion latency bins. Each `bins` object contains a
4017 set of (key, value) pairs where keys are latency durations and values count how
4018 many I/Os had completion latencies of the corresponding duration. For example,
4021 "bins" : { "87552" : 1, "89600" : 1, "94720" : 1, "96768" : 1, "97792" : 1, "99840" : 1, "100864" : 2, "103936" : 6, "104960" : 534, "105984" : 5995, "107008" : 7529, ... }
4023 This data indicates that one I/O required 87,552ns to complete, two I/Os required
4024 100,864ns to complete, and 7529 I/Os required 107,008ns to complete.
4026 Also included with fio is a Python script `fio_jsonplus_clat2csv` that takes
4027 json+ output and generates CSV-formatted latency data suitable for plotting.
4029 The latency durations actually represent the midpoints of latency intervals.
4030 For details refer to :file:`stat.h`.
4036 There are two trace file format that you can encounter. The older (v1) format is
4037 unsupported since version 1.20-rc3 (March 2008). It will still be described
4038 below in case that you get an old trace and want to understand it.
4040 In any case the trace is a simple text file with a single action per line.
4043 Trace file format v1
4044 ~~~~~~~~~~~~~~~~~~~~
4046 Each line represents a single I/O action in the following format::
4050 where `rw=0/1` for read/write, and the `offset` and `length` entries being in bytes.
4052 This format is not supported in fio versions >= 1.20-rc3.
4055 Trace file format v2
4056 ~~~~~~~~~~~~~~~~~~~~
4058 The second version of the trace file format was added in fio version 1.17. It
4059 allows to access more then one file per trace and has a bigger set of possible
4062 The first line of the trace file has to be::
4066 Following this can be lines in two different formats, which are described below.
4068 The file management format::
4072 The `filename` is given as an absolute path. The `action` can be one of these:
4075 Add the given `filename` to the trace.
4077 Open the file with the given `filename`. The `filename` has to have
4078 been added with the **add** action before.
4080 Close the file with the given `filename`. The file has to have been
4084 The file I/O action format::
4086 filename action offset length
4088 The `filename` is given as an absolute path, and has to have been added and
4089 opened before it can be used with this format. The `offset` and `length` are
4090 given in bytes. The `action` can be one of these:
4093 Wait for `offset` microseconds. Everything below 100 is discarded.
4094 The time is relative to the previous `wait` statement.
4096 Read `length` bytes beginning from `offset`.
4098 Write `length` bytes beginning from `offset`.
4100 :manpage:`fsync(2)` the file.
4102 :manpage:`fdatasync(2)` the file.
4104 Trim the given file from the given `offset` for `length` bytes.
4107 I/O Replay - Merging Traces
4108 ---------------------------
4110 Colocation is a common practice used to get the most out of a machine.
4111 Knowing which workloads play nicely with each other and which ones don't is
4112 a much harder task. While fio can replay workloads concurrently via multiple
4113 jobs, it leaves some variability up to the scheduler making results harder to
4114 reproduce. Merging is a way to make the order of events consistent.
4116 Merging is integrated into I/O replay and done when a
4117 :option:`merge_blktrace_file` is specified. The list of files passed to
4118 :option:`read_iolog` go through the merge process and output a single file
4119 stored to the specified file. The output file is passed on as if it were the
4120 only file passed to :option:`read_iolog`. An example would look like::
4122 $ fio --read_iolog="<file1>:<file2>" --merge_blktrace_file="<output_file>"
4124 Creating only the merged file can be done by passing the command line argument
4125 :option:`--merge-blktrace-only`.
4127 Scaling traces can be done to see the relative impact of any particular trace
4128 being slowed down or sped up. :option:`merge_blktrace_scalars` takes in a colon
4129 separated list of percentage scalars. It is index paired with the files passed
4130 to :option:`read_iolog`.
4132 With scaling, it may be desirable to match the running time of all traces.
4133 This can be done with :option:`merge_blktrace_iters`. It is index paired with
4134 :option:`read_iolog` just like :option:`merge_blktrace_scalars`.
4136 In an example, given two traces, A and B, each 60s long. If we want to see
4137 the impact of trace A issuing IOs twice as fast and repeat trace A over the
4138 runtime of trace B, the following can be done::
4140 $ fio --read_iolog="<trace_a>:"<trace_b>" --merge_blktrace_file"<output_file>" --merge_blktrace_scalars="50:100" --merge_blktrace_iters="2:1"
4142 This runs trace A at 2x the speed twice for approximately the same runtime as
4143 a single run of trace B.
4146 CPU idleness profiling
4147 ----------------------
4149 In some cases, we want to understand CPU overhead in a test. For example, we
4150 test patches for the specific goodness of whether they reduce CPU usage.
4151 Fio implements a balloon approach to create a thread per CPU that runs at idle
4152 priority, meaning that it only runs when nobody else needs the cpu.
4153 By measuring the amount of work completed by the thread, idleness of each CPU
4154 can be derived accordingly.
4156 An unit work is defined as touching a full page of unsigned characters. Mean and
4157 standard deviation of time to complete an unit work is reported in "unit work"
4158 section. Options can be chosen to report detailed percpu idleness or overall
4159 system idleness by aggregating percpu stats.
4162 Verification and triggers
4163 -------------------------
4165 Fio is usually run in one of two ways, when data verification is done. The first
4166 is a normal write job of some sort with verify enabled. When the write phase has
4167 completed, fio switches to reads and verifies everything it wrote. The second
4168 model is running just the write phase, and then later on running the same job
4169 (but with reads instead of writes) to repeat the same I/O patterns and verify
4170 the contents. Both of these methods depend on the write phase being completed,
4171 as fio otherwise has no idea how much data was written.
4173 With verification triggers, fio supports dumping the current write state to
4174 local files. Then a subsequent read verify workload can load this state and know
4175 exactly where to stop. This is useful for testing cases where power is cut to a
4176 server in a managed fashion, for instance.
4178 A verification trigger consists of two things:
4180 1) Storing the write state of each job.
4181 2) Executing a trigger command.
4183 The write state is relatively small, on the order of hundreds of bytes to single
4184 kilobytes. It contains information on the number of completions done, the last X
4187 A trigger is invoked either through creation ('touch') of a specified file in
4188 the system, or through a timeout setting. If fio is run with
4189 :option:`--trigger-file`\= :file:`/tmp/trigger-file`, then it will continually
4190 check for the existence of :file:`/tmp/trigger-file`. When it sees this file, it
4191 will fire off the trigger (thus saving state, and executing the trigger
4194 For client/server runs, there's both a local and remote trigger. If fio is
4195 running as a server backend, it will send the job states back to the client for
4196 safe storage, then execute the remote trigger, if specified. If a local trigger
4197 is specified, the server will still send back the write state, but the client
4198 will then execute the trigger.
4200 Verification trigger example
4201 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4203 Let's say we want to run a powercut test on the remote Linux machine 'server'.
4204 Our write workload is in :file:`write-test.fio`. We want to cut power to 'server' at
4205 some point during the run, and we'll run this test from the safety or our local
4206 machine, 'localbox'. On the server, we'll start the fio backend normally::
4208 server# fio --server
4210 and on the client, we'll fire off the workload::
4212 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
4214 We set :file:`/tmp/my-trigger` as the trigger file, and we tell fio to execute::
4216 echo b > /proc/sysrq-trigger
4218 on the server once it has received the trigger and sent us the write state. This
4219 will work, but it's not **really** cutting power to the server, it's merely
4220 abruptly rebooting it. If we have a remote way of cutting power to the server
4221 through IPMI or similar, we could do that through a local trigger command
4222 instead. Let's assume we have a script that does IPMI reboot of a given hostname,
4223 ipmi-reboot. On localbox, we could then have run fio with a local trigger
4226 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
4228 For this case, fio would wait for the server to send us the write state, then
4229 execute ``ipmi-reboot server`` when that happened.
4231 Loading verify state
4232 ~~~~~~~~~~~~~~~~~~~~
4234 To load stored write state, a read verification job file must contain the
4235 :option:`verify_state_load` option. If that is set, fio will load the previously
4236 stored state. For a local fio run this is done by loading the files directly,
4237 and on a client/server run, the server backend will ask the client to send the
4238 files over and load them from there.
4244 Fio supports a variety of log file formats, for logging latencies, bandwidth,
4245 and IOPS. The logs share a common format, which looks like this:
4247 *time* (`msec`), *value*, *data direction*, *block size* (`bytes`),
4248 *offset* (`bytes`), *command priority*
4250 *Time* for the log entry is always in milliseconds. The *value* logged depends
4251 on the type of log, it will be one of the following:
4254 Value is latency in nsecs
4260 *Data direction* is one of the following:
4269 The entry's *block size* is always in bytes. The *offset* is the position in bytes
4270 from the start of the file for that particular I/O. The logging of the offset can be
4271 toggled with :option:`log_offset`.
4273 *Command priority* is 0 for normal priority and 1 for high priority. This is controlled
4274 by the ioengine specific :option:`cmdprio_percentage`.
4276 Fio defaults to logging every individual I/O but when windowed logging is set
4277 through :option:`log_avg_msec`, either the average (by default) or the maximum
4278 (:option:`log_max_value` is set) *value* seen over the specified period of time
4279 is recorded. Each *data direction* seen within the window period will aggregate
4280 its values in a separate row. Further, when using windowed logging the *block
4281 size* and *offset* entries will always contain 0.
4287 Normally fio is invoked as a stand-alone application on the machine where the
4288 I/O workload should be generated. However, the backend and frontend of fio can
4289 be run separately i.e., the fio server can generate an I/O workload on the "Device
4290 Under Test" while being controlled by a client on another machine.
4292 Start the server on the machine which has access to the storage DUT::
4296 where `args` defines what fio listens to. The arguments are of the form
4297 ``type,hostname`` or ``IP,port``. *type* is either ``ip`` (or ip4) for TCP/IP
4298 v4, ``ip6`` for TCP/IP v6, or ``sock`` for a local unix domain socket.
4299 *hostname* is either a hostname or IP address, and *port* is the port to listen
4300 to (only valid for TCP/IP, not a local socket). Some examples:
4304 Start a fio server, listening on all interfaces on the default port (8765).
4306 2) ``fio --server=ip:hostname,4444``
4308 Start a fio server, listening on IP belonging to hostname and on port 4444.
4310 3) ``fio --server=ip6:::1,4444``
4312 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
4314 4) ``fio --server=,4444``
4316 Start a fio server, listening on all interfaces on port 4444.
4318 5) ``fio --server=1.2.3.4``
4320 Start a fio server, listening on IP 1.2.3.4 on the default port.
4322 6) ``fio --server=sock:/tmp/fio.sock``
4324 Start a fio server, listening on the local socket :file:`/tmp/fio.sock`.
4326 Once a server is running, a "client" can connect to the fio server with::
4328 fio <local-args> --client=<server> <remote-args> <job file(s)>
4330 where `local-args` are arguments for the client where it is running, `server`
4331 is the connect string, and `remote-args` and `job file(s)` are sent to the
4332 server. The `server` string follows the same format as it does on the server
4333 side, to allow IP/hostname/socket and port strings.
4335 Fio can connect to multiple servers this way::
4337 fio --client=<server1> <job file(s)> --client=<server2> <job file(s)>
4339 If the job file is located on the fio server, then you can tell the server to
4340 load a local file as well. This is done by using :option:`--remote-config` ::
4342 fio --client=server --remote-config /path/to/file.fio
4344 Then fio will open this local (to the server) job file instead of being passed
4345 one from the client.
4347 If you have many servers (example: 100 VMs/containers), you can input a pathname
4348 of a file containing host IPs/names as the parameter value for the
4349 :option:`--client` option. For example, here is an example :file:`host.list`
4350 file containing 2 hostnames::
4352 host1.your.dns.domain
4353 host2.your.dns.domain
4355 The fio command would then be::
4357 fio --client=host.list <job file(s)>
4359 In this mode, you cannot input server-specific parameters or job files -- all
4360 servers receive the same job file.
4362 In order to let ``fio --client`` runs use a shared filesystem from multiple
4363 hosts, ``fio --client`` now prepends the IP address of the server to the
4364 filename. For example, if fio is using the directory :file:`/mnt/nfs/fio` and is
4365 writing filename :file:`fileio.tmp`, with a :option:`--client` `hostfile`
4366 containing two hostnames ``h1`` and ``h2`` with IP addresses 192.168.10.120 and
4367 192.168.10.121, then fio will create two files::
4369 /mnt/nfs/fio/192.168.10.120.fileio.tmp
4370 /mnt/nfs/fio/192.168.10.121.fileio.tmp
4372 Terse output in client/server mode will differ slightly from what is produced
4373 when fio is run in stand-alone mode. See the terse output section for details.